Search

Purpose: Artificial intelligence (AI)-aided methods have made significant progress in the auto-delineation of normal tissues. However, these approaches struggle with the auto-contouring of radiotherapy target volume. Our goal is to model the delineation of target volume as a clinical decision-making problem, resolved by leveraging large language model-aided multimodal learning approaches., Methods and Materials: A vision-language model, termed Medformer, has been developed, employing the hierarchical vision transformer as its backbone, and incorporating large language models to extract text-rich features. The contextually embedded linguistic features are seamlessly integrated into visual features for language-aware visual encoding through the visual language attention module. Metrics, including Dice similarity coefficient (DSC), intersection over union (IOU), and 95th percentile Hausdorff distance (HD95), were used to quantitatively evaluate the performance of our model. The evaluation was conducted on an in-house prostate cancer dataset and a public oropharyngeal carcinoma (OPC) dataset, totaling 668 subjects., Results: Our Medformer achieved a DSC of 0.81 ± 0.10 versus 0.72 ± 0.10, IOU of 0.73 ± 0.12 versus 0.65 ± 0.09, and HD95 of 9.86 ± 9.77 mm versus 19.13 ± 12.96 mm for delineation of gross tumor volume (GTV) on the prostate cancer dataset. Similarly, on the OPC dataset, it achieved a DSC of 0.77 ± 0.11 versus 0.72 ± 0.09, IOU of 0.70 ± 0.09 versus 0.65 ± 0.07, and HD95 of 7.52 ± 4.8 mm versus 13.63 ± 7.13 mm, representing significant improvements (p < 0.05). For delineating the clinical target volume (CTV), Medformer achieved a DSC of 0.91 ± 0.04, IOU of 0.85 ± 0.05, and HD95 of 2.98 ± 1.60 mm, comparable to other state-of-the-art algorithms., Conclusions: Auto-delineation of the treatment target based on multimodal learning outperforms conventional approaches that rely purely on visual features. Our method could be adopted into routine practice to rapidly contour CTV/GTV., Competing Interests: Declaration of competing interest None, (Copyright © 2024. Published by Elsevier Inc.)

Background: Surface-guided imaging (SGI) is increasingly utilized to monitor patient motion during deep inspiration breath hold (DIBH) in radiotherapy. Understanding the association between surface and internal motion is crucial for effective monitoring., Purpose: To investigate the relation between motion detected by SGI using surface-guided radiotherapy (SGRT) and internal motion measured through diaphragm tracking on kV projections acquired with DIBH for online CBCT., Methods: Both SGI and kV were simultaneously acquired for ten patients over a total of 200 breath holds (BH). Diaphragm tracking was performed using second-degree polynomial curve fitting on the derivative images for each kV projection and high-pass filtering at 1/30 Hz to remove rotational effects. The superior-inferior (SI) and anterior-posterior (AP) motions of SGI were then compared to kV tracking using various statistical measures., Results: The correlation (individuals' median: -0.07 to 0.73) was a suboptimal metric for the BH data. The median and 95th percentile absolute differences between SGI-SI and kV were 0.73 mm and 3.46 mm, respectively, during DIBH. For SGI-AP, the corresponding values were 0.55 mm and 2.80 mm. For inter-BH measurements, the contingency table based on a 3 mm threshold indicated surface/diaphragm motion agreement for SGI-SI/kV and SGI-AP/kV was 61 % and 56 %, respectively., Conclusion: Both intra- and inter-BH measurements indicated a limited association between surface and diaphragm motion, with certain constraints noted due to kV tracking and DIBH data. These findings warrant further investigation into the association between surface and internal motion., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Associazione Italiana di Fisica Medica e Sanitaria. Published by Elsevier Ltd. All rights reserved.)

Background and Objectives: Ependymoma is commonly classified as World Health Organization grade 2 with the anaplastic variant categorized as grade 3. Incomplete resection or anaplastic features can result in unfavorable outcomes. Stereotactic radiosurgery (SRS) provides a minimally invasive approach for recurrent ependymomas. Our study investigates the efficacy and safety of SRS for grade 2 and 3 ependymomas in pediatric and adult populations., Methods: We conducted a retrospective analysis on 34 patients with 75 ependymomas after CyberKnife SRS between 1998 and 2023. Fourteen were pediatric (3-18 years), and 20 were adult (19-75 years) patients. The median age was 21 years, and the median tumor volume was 0.64 cc. The median single-fraction equivalent dose was 16.6 Gy, with SRS administered at 77% of the median isodose line., Results: After a median follow-up of 42.7 months (range: 3.8-438.3), 22.7% of ependymomas progressed. The 5-year local tumor control rate was 78.1%, varying between 59.6% and 90.2% for children and adults, with grade 2 at 85.9% compared with 58.5% for grade 3 tumors. The 5-year overall survival rate was 73.6%, notably higher in adults (94.7%) than in children (41%), and 100% for grade 2 but decreased to 35.9% for grade 3 patients. The 5-year progression-free survival rate was 68.5%, with 78.3% and 49.2% for adults and children, respectively, and a favorable 88.8% for grade 2, contrasting with 32.6% for grade 3 patients. Symptom improvement was observed in 85.3% of patients. Adverse radiation effects occurred in 21.4% of pediatric patients., Conclusion: Our study supports SRS as a viable modality for pediatric and adult patients with grade 2 and 3 ependymomas. Despite lower local tumor control in pediatric and grade 3 cases, integrating SRS holds promise for improved outcomes. Emphasizing careful patient selection, personalized treatment planning, and long-term follow-up is crucial for optimal neurosurgical outcomes., (Copyright © 2024 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the Congress of Neurological Surgeons.)

Background and Objectives: Hemangiopericytomas are infrequent vascular tumors originating from Zimmermann pericytes. The conventional treatment involves gross total resection, followed by adjuvant radiotherapy. Nevertheless, their tendency to infiltrate dural sinuses, high vascularity, and anatomic complexity pose challenges for radical resection, leading to a significant risk of recurrence. Stereotactic radiosurgery (SRS) has emerged as a promising adjuvant therapy to address these challenges. Our study provides the largest single-institutional retrospective, aiming to evaluate the effectiveness and safety of SRS as a treatment modality for residual, recurrent, and metastatic hemangiopericytomas., Methods: From 1998 to 2023, 27 patients with 101 tumors underwent CyberKnife SRS at Stanford University Medical Center. The median age was 51 years at the time of treatment. The median follow-up period from SRS was 103 months (range: 6-250). All patients underwent upfront surgical resection. The median tumor volume was 1.5 cc. The median single-fraction equivalent dose was 19 Gy. The SRS was administered at the 76% of the median isodose line (range: 64-89)., Results: Of the 101 treated tumors, 24 (23.8%) progressed with a median time to recurrence of 30 months. At 10 years, the rates of local tumor control (LTC), overall survival (OS), and progression-free survival (PFS) were 74.3%, 80.8%, and 67%, respectively. In patients with metastatic lesions, the LTC rates were significantly greater when compared with those with residual or recurrent tumors. There was no significant difference between patients with residual, recurrent, and metastatic hemangiopericytomas in OS and PFS. Notably, no cases of radiation-induced adverse events were detected., Conclusion: SRS leads to excellent LTC, PFS, and OS at 10 years with negligible risk for adverse events. Therefore, it is an effective and safe management modality for patients with residual, recurrent, and metastatic hemangiopericytomas., (Copyright © Congress of Neurological Surgeons 2024. All rights reserved.)

Photoionization and dissociative photoionization of acetaldehyde (CH 3 CHO) in the 10.0-13.7 eV energy range are studied by using synchrotron radiation double imaging photoelectron photoion coincidence spectroscopy (i 2 PEPICO). The X 2 A' and A 2 A" electronic states of CH 3 CHO + as well as the Franck-Condon gap region between these two states have been populated with several vibrational sequences and assigned in the high-resolution slow photoelectron spectrum (SPES). The adiabatic ionization energies (AIEs) of the X 2 A' and A 2 A" states are measured at 10.228±0.006 and 12.52±0.05 eV, respectively. The present results show that the X 2 A' state is a stable state while the A 2 A" state is fully dissociative to produce CH 3 CO + , CHO + and CH 4 + fragment ions. The 0 K appearance energies (AE 0K ) of CH 3 CO + and CHO + fragment ions are determined through the modeling of the breakdown diagram, i. e., AE 0K (CH 3 CO + )=10.89±0.01 eV (including a reverse barrier of ~0.19 eV) and AE 0K (CHO + )=11.54±0.05 eV. In addition, the dissociation mechanisms of CH 3 CHO + including statistical dissociation, direct bond breaking and isomerization are discussed with the support of the calculated dissociation limits and transition state energies., (© 2024 Wiley-VCH GmbH.)

Purpose: Ultrahigh-dose-rate (FLASH) irradiation has been reported to reduce normal tissue damage compared with conventional dose rate (CONV) irradiation without compromising tumor control. This proof-of-concept study aims to develop a deep learning (DL) approach to quantify the FLASH isoeffective dose (dose of CONV that would be required to produce the same effect as the given physical FLASH dose) with postirradiation mouse intestinal histology images., Methods and Materials: Eighty-four healthy C57BL/6J female mice underwent 16 MeV electron CONV (0.12 Gy/s; n = 41) or FLASH (200 Gy/s; n = 43) single fraction whole abdominal irradiation. Physical dose ranged from 12 to 16 Gy for FLASH and 11 to 15 Gy for CONV in 1 Gy increments. Four days after irradiation, 9 jejunum cross-sections from each mouse were hematoxylin and eosin stained and digitized for histological analysis. CONV data set was randomly split into training (n = 33) and testing (n = 8) data sets. ResNet101-based DL models were retrained using the CONV training data set to estimate the dose based on histological features. The classical manual crypt counting (CC) approach was implemented for model comparison. Cross-section-wise mean squared error was computed to evaluate the dose estimation accuracy of both approaches. The validated DL model was applied to the FLASH data set to map the physical FLASH dose into the isoeffective dose., Results: The DL model achieved a cross-section-wise mean squared error of 0.20 Gy 2 on the CONV testing data set compared with 0.40 Gy 2 of the CC approach. Isoeffective doses estimated by the DL model for FLASH doses of 12, 13, 14, 15, and 16 Gy were 12.19 ± 0.46, 12.54 ± 0.37, 12.69 ± 0.26, 12.84 ± 0.26, and 13.03 ± 0.28 Gy, respectively., Conclusions: Our proposed DL model achieved accurate CONV dose estimation. The DL model results indicate that in the physical dose range of 13 to 16 Gy, the biologic dose response of small intestinal tissue to FLASH irradiation is represented by a lower isoeffective dose compared with the physical dose. Our DL approach can be a tool for studying isoeffective doses of other radiation dose modifying interventions., (Copyright © 2023 Elsevier Inc. All rights reserved.)

Background/Objectives : Medulloblastoma is the most common malignant brain tumor in children. In recent decades, the therapeutic landscape has undergone significant changes, with stereotactic radiosurgery (SRS) emerging as a promising treatment for recurrent cases. Our study provides a comprehensive analysis of the long-term efficacy and safety of SRS in recurrent medulloblastomas across both pediatric and adult patients at a single institution. Methods : We retrospectively reviewed the clinical and radiological records of patients who underwent CyberKnife SRS for recurrent cranial medulloblastomas at our institution between 1998 and 2023. Follow-up data were available for 15 medulloblastomas in 10 patients. The cohort comprised eight pediatric patients (ages 3-18) and two adult patients (ages 19-75). The median age at the time of SRS was 13 years, the median tumor volume accounted for 1.9 cc, the median biologically equivalent dose (BED) was 126 Gy, and the single-fraction equivalent dose (SFED) was 18 Gy. The SRS was administered at 75% of the median isodose line. Results : Following a median follow-up of 39 months (range: 6-78), 53.3% of the medulloblastomas progressed, 13.3% regressed, and 33.3% remained stable. The 3-year local tumor control (LTC) rate for all medulloblastomas was 65%, with lower rates observed in the adult cohort (50%) and higher rates in pediatric patients (67%). The 3-year overall survival (OS) rate was 70%, with significantly higher rates in pediatric patients (75%) compared to adult patients (50%). The 3-year progression-free survival (PFS) rate was 58.3%, with higher rates in pediatric patients (60%) compared to adult patients (50%). Two pediatric patients developed radiation-induced edema, while two adult patients experienced radiation necrosis at the latest follow-up, with both adult patients passing away. Conclusions : Our study provides a complex perspective on the efficacy and safety of CyberKnife SRS in treating recurrent cranial medulloblastomas across pediatric and adult populations. The rarity of adverse radiation events (AREs) underscores the safety profile of SRS, reinforcing its role in enhancing treatment outcomes. The intricacies of symptomatic outcomes, intertwined with factors such as age, tumor location, and prior surgeries, emphasize the need for personalized treatment approaches. Our findings underscore the imperative for ongoing research and the development of more refined treatment strategies for recurrent medulloblastomas. Given the observed disparities in treatment outcomes, a more meticulous tailoring of treatment approaches becomes crucial.

The growing popularity of e-cigarettes and the associated risks of nicotine addiction present a new challenge to global public health security. Measuring the nicotine levels in e-cigarette aerosols is essential to assess the safety of e-cigarettes. In this study, a rapid in situ method was developed for online quantification of nicotine in e-cigarette aerosols by using a homemade vacuum ultraviolet photoionization aerosol mass spectrometer (VUV-AMS). E-cigarette liquids with different nicotine concentrations were prepared to generate aerosols containing different levels of nicotine, which were employed as the calibration sources for nicotine quantification by VUV-AMS. The results showed that the mass concentration of nicotine in e-cigarette aerosols has a good linear relationship with its signal intensity in the mass spectrum, and the limits of detection and quantitation of nicotine by VUV-AMS were found to be 2.0 and 6.2 μg per puff respectively. Then the online method was utilized to measure five commercial e-cigarettes, and their nicotine yields were determined to be between 31 and 188 μg per puff with the nicotine fluxes from 7.7 to 70 μg s -1 , agreeing with the results of the gas chromatography with a flame ionization detector (GC-FID). This study demonstrated the feasibility and advantages of VUV-AMS for quick quantification of nicotine in e-cigarette aerosols within seconds.

Detection and segmentation of brain metastases (BMs) play a pivotal role in diagnosis, treatment planning, and follow-up evaluations for effective BM management. Given the rising prevalence of BM cases and its predominantly multiple onsets, automated segmentation is becoming necessary in stereotactic radiosurgery. It not only alleviates the clinician's manual workload and improves clinical workflow efficiency but also ensures treatment safety, ultimately improving patient care. Recent strides in machine learning, particularly in deep learning (DL), have revolutionized medical image segmentation, achieving state-of-the-art results. This review aims to analyze auto-segmentation strategies, characterize the utilized data, and assess the performance of cutting-edge BM segmentation methodologies. Additionally, we delve into the challenges confronting BM segmentation and share insights gleaned from our algorithmic and clinical implementation experiences.

With the remarkably strong growth of the biopharmaceutical market, an increasing demand for self-administration and rising competitions attract substantial interest to the biologic-device combination products. The ease-of-use of biologic-device combination products can minimize dosing error, improve patient compliance and add value to the life-cycle management of biological products. As listed in the purple book issued by the U.S. Food and Drug Administration (FDA), a total of 98 brand biologic-device combination products have been approved with Biologic License Application from January 2000 to August 2023, where this review mainly focused on 63 products containing neither insulin nor vaccine. Prefilled syringes (PFS) and autoinjectors are the most widely adopted devices, whereas innovative modifications like needle safety guard and dual-chamber design and novel devices like on-body injector also emerged as promising presentations. All 16 insulin products employ pen injectors, while all 19 vaccine products are delivered by a PFS. This review provides a systematic summary of FDA-approved biologic-device combination products regarding their device configurations, routes of administration, formulations, instructions for use, etc. In addition, challenges and opportunities associated with biologic-device compatibility, regulatory complexity, and smart connected devices are also discussed. It is believed that evolving technologies will definitely move the boundaries of biologic-device combination product development even further., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024. Published by Elsevier Inc.)

Background and Objectives: Stereotactic radiosurgery (SRS) has been an attractive treatment modality for both cranial and spinal hemangioblastomas, especially for multiple lesions commonly associated with von Hippel-Lindau (VHL) disease. This study aims to provide the largest long-term analysis of treatment efficacy and adverse effects of SRS for cranial and spinal hemangioblastomas at a single institution., Methods: We evaluated the clinical and radiological outcomes of patients with hemangioblastomas treated with CyberKnife SRS at our institute from 1998 to 2022. The follow-up data were available for 135 hemangioblastomas in 35 patients. Twenty-eight patients had 123 hemangioblastomas associated with VHL, and 7 had 12 sporadic hemangioblastomas. The median age was 36 years, and the median tumor volume accounted for 0.4 cc. The SRS was administered with the median single-fraction equivalent dose of 18 Gy to the 77% median isodose line., Results: At a median follow-up of 57 months (range: 3-260), only 20 (16.2%) of the VHL-associated and 1 (8.3%) sporadic hemangioblastomas progressed. The 5-year local tumor control rate was 91.3% for all hemangioblastomas, 91.7% among the sporadic lesions, and 92.9% in patients with VHL. SRS improved tumor-associated symptoms of 98 (74.8%) of 131 symptomatic hemangioblastomas, including headache, neck pain, dizziness, visual disturbances, dysesthesia, ataxia, motor impairment, seizures, and dysphagia. Two patients developed radiation necrosis (5.7%), and 1 of them required surgical resection., Conclusion: SRS is a safe and effective treatment option for patients with hemangioblastomas in critical locations, such as the brainstem, cervicomedullary junction, and spinal cord, and in patients with multiple hemangioblastomas associated with VHL disease., (Copyright © Congress of Neurological Surgeons 2023. All rights reserved.)

Objective: Pregnant women experience enormous psychological pressure, particularly during the late trimester. Symptoms of depression in late pregnancy may persist postpartum, increasing the incidence of postpartum depression. This study is aimed to investigate the factors influencing depressive symptoms among pregnant women in their third trimester at a Chinese tertiary hospital and provide information for effective intervention., Methods: Pregnant women in their third trimester who visited the Ningbo Women and Children's Hospital between January 1, 2020 and June 30, 2022 participated in this study. A score of ≥ 13 on the Edinburgh Postnatal Depression Scale (EPDS) was considered as positive for depressive symptom. Potential influencing factors were examined by using an online questionnaire and analyzed using multivariate logistic regression., Results: A total of 1196 participants were recruited. The mean EPDS score was 7.12 ± 4.22. The positive screening rate for depressive symptom was 9.9%. Univariate analysis showed that living with partner, annual family income, planned pregnancy, sleep quality, and partner's drinking habits were related to positive screening for depression(P < 0.05). Furthermore, multivariate logistic regression analysis showed that living away from the partner (odds ratio [OR]: 2.054, 95% confidence interval [CI]: 1.094-3.696, P = 0.02), annual family income < 150,000 Chinese Yuan (CNY; OR: 1.762, 95% CI: 1.170-2.678, P = 0.007), poor sleep quality (OR: 4.123, 95% CI: 2.764-6.163, P < 0.001), and partner's frequent drinking habit (OR: 2.227, 95% CI: 1.129-4.323, P = 0.019) were independent influencing factors for positive depression screening (P < 0.05)., Conclusion: Family's economic condition, sleep quality, living with partner, and partner's drinking habits were related to positive depression screening in late pregnancy. Pregnant women with these risk factors should be given more attention and supported to avoid developing depression., (© 2023. The Author(s).)

Purpose: To develop an angular correction methodology and characterize a high-resolution complementary metal-oxide-semiconductor (CMOS) array for patient specific quality assurance on a robotic arm linear accelerator., Methods: Beam path files from the treatment planning software (TPS) were used to calculate the angle of radiation beam with respect to the detector plane. Beams from multiple discrete angles were delivered to the CMOS detector array and an angular dependency look up table (LUT) was created. The LUT was then used to correct for the angular dependency of the detector. An iso-centric 5 mm fixed cone, non iso-centric multi-target fixed cone, 10 mm Iris and a multi-leaf collimator (MLC) based collimated plan were delivered to the phantom and compared to the TPS with and without angular correction applied. Additionally, the CMOS array was compared to gafchromic film and a diode array., Results: Large errors of up to 30% were observed for oblique angles. When angular correction was applied, the gamma passing rate increased from 99.2% to 100% (average gamma value decreased from 0.29 to 0.14) for the 5-mm iso-centric cone plan. Similarly, the passing rate increased from 84.0% to 100% for the Iris plan and from 49.98% to 98.4% for the MLC plan when angular correction was applied. For the multi-target plan, applying angular correction improved the gamma passing rate from 94% to 99.6%. The 5 mm iso-centric fixed cone plan was also delivered to film, and the gamma passing rate was 91.3% when using gafchromic film as the reference dataset, whereas the diode array provided insufficient sampling for this plan., Conclusion: A methodology of calculating the beam angle based on the beam path files was developed and validated. The array was demonstrated to be superior to other quality assurance tools because of its sub-millimeter spatial resolution and immediate read out of the results., (© 2023 The Authors. Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.)

Deep learning (DL) models for medical image segmentation are highly influenced by intensity variations of input images and lack generalization due to primarily utilizing pixels' intensity information for inference. Acquiring sufficient training data is another challenge limiting models' applications. Here, we proposed to leverage the consistency of organs' anatomical position and shape information in medical images. We introduced a framework leveraging recurring anatomical patterns through global binary masks for organ segmentation. Two scenarios were studied: (1) global binary masks were the only input for the U-Net based model, forcing exclusively encoding organs' position and shape information for rough segmentation or localization. (2) Global binary masks were incorporated as an additional channel providing position/shape clues to mitigate training data scarcity. Two datasets of the brain and heart computed tomography (CT) images with their ground-truth were split into (26:10:10) and (12:3:5) for training, validation, and test respectively. The two scenarios were evaluated using full training split as well as reduced subsets of training data. In scenario (1), training exclusively on global binary masks led to Dice scores of 0.77 ± 0.06 and 0.85 ± 0.04 for the brain and heart structures respectively. Average Euclidian distance of 3.12 ± 1.43 mm and 2.5 ± 0.93 mm were obtained relative to the center of mass of the ground truth for the brain and heart structures respectively. The outcomes indicated encoding a surprising degree of position and shape information through global binary masks. In scenario (2), incorporating global binary masks led to significantly higher accuracy relative to the model trained on only CT images in small subsets of training data; the performance improved by 4.3%-125.3% and 1.3%-48.1% for 1-8 training cases of the brain and heart datasets respectively. The findings imply the advantages of utilizing global binary masks for building models that are robust to image intensity variations as well as an effective approach to boost performance when access to labeled training data is highly limited., (© 2023 Institute of Physics and Engineering in Medicine.)

We compiled a sampling of the treatment techniques of intensity-modulated total body irradiation, total marrow irradiation and total marrow and lymphoid irradiation utilized by several centers across North America and Europe. This manuscript does not serve as a consensus guideline, but rather is meant to serve as a convenient reference for centers that are considering starting an intensity-modulated program., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2023 Published by Elsevier B.V. on behalf of European Society for Radiotherapy and Oncology.)

Accurate and efficient delineation of the clinical target volume (CTV) is of utmost significance in post-operative breast cancer radiotherapy. However, CTV delineation is challenging as the exact extent of microscopic disease encompassed by CTV is not visualizable in radiological images and remains uncertain. We proposed to mimic physicians' contouring practice for CTV segmentation in stereotactic partial breast irradiation (S-PBI) where CTV is derived from tumor bed volume (TBV) via a margin expansion followed by correcting the extensions for anatomical barriers of tumor invasion (e.g. skin, chest wall). We proposed a deep-learning model, where CT images and the corresponding TBV masks formed a multi-channel input for a 3D U-Net based architecture. The design guided the model to encode the location-related image features and directed the network to focus on TBV to initiate CTV segmentation. Gradient weighted class activation map (Grad-CAM) visualizations of the model predictions revealed that the extension rules and geometric/anatomical boundaries were learnt during model training to assist the network to limit the expansion to a certain distance from the chest wall and the skin. We retrospectively collected 175 prone CT images from 35 post-operative breast cancer patients who received 5-fraction partial breast irradiation regimen on GammaPod. The 35 patients were randomly split into training (25), validation (5) and test (5) sets. Our model achieved mean (standard deviation) of 0.94 (±0.02), 2.46 (±0.5) mm, and 0.53 (±0.14) mm for Dice similarity coefficient, 95th percentile Hausdorff distance, and average symmetric surface distance respectively on the test set. The results are promising for improving the efficiency and accuracy of CTV delineation during on-line treatment planning procedure., (© 2023 Institute of Physics and Engineering in Medicine.)

Purpose: Deep inspiration breath-hold (DIBH) is crucial in reducing the lung and cardiac dose for treatment of left-sided breast cancer. We compared the stability and reproducibility of two DIBH techniques: Active Breathing Coordinator (ABC) and VisionRT (VRT)., Materials and Methods: We examined intra- and inter-fraction positional variation of the left lung. Eight left-sided breast cancer patients were monitored with electronic portal imaging during breath-hold (BH) at every fraction. For each patient, half of the fractions were treated using ABC and the other half with VRT, with an equal amount starting with either ABC or VRT. The lung in each portal image was delineated, and the variation of its area was evaluated. Intrafraction stability was evaluated as the mean coefficient of variation (CV) of the lung area for the supraclavicular (SCV) and left lateral (LLat) field over the course of treatment. Reproducibility was the CV for the first image of each fraction. Daily session time and total imaging monitor units (MU) used in patient positioning were recorded., Results: The mean intrafraction stability across all patients for the LLat field was 1.3 ± 0.7% and 1.5 ± 0.9% for VRT and ABC, respectively. Similarly, this was 1.5 ± 0.7% and 1.6 ± 0.8% for VRT and ABC, respectively, for the SCV field. The mean interfraction reproducibility for the LLat field was 11.0 ± 3.4% and 14.9 ± 6.0% for VRT and ABC, respectively. Similarly, this was 13.0 ± 2.5% and 14.8 ± 9% for VRT and ABC, respectively, for the SCV. No difference was observed in the number of verification images required for either technique., Conclusions: The stability and reproducibility were found to be comparable between ABC and VRT. ABC can have larger interfractional variation with less feedback to the treating therapist compared to VRT as shown in the increase in geometric misses at the matchline., (© 2023 The Authors. Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.)

Objective: To determine willingness to pay for a diabetic retinopathy screening, and its determinants, among people with diabetes mellitus in Qujiang District of Shaoguan City, rural Guangdong, southern China., Design: This cross-sectional study was conducted through a large-scale screening programme in 2019. We randomly selected 575 (21.5%) among 2677 people over 18 years old with known diabetes who attended the screening. Participants elected to pay nothing or RMB10-RMB120 (US$1.6-US$18.8), in RMB10 intervals, displayed on printed cards. One trained interviewer collected all the data., Setting: Ten primary health centres in Qujiang District of Shaoguan City, Guangdong., Participants: 545 from the 575 randomly selected people (94.8%) agreed to participate in the study., Outcome Measures: Proportion of participants willing to pay anything for screening, mean amount they were willing to pay and determinants of these figures., Results: Among 545 participants (mean age 64.6 years (SD±10.4), 40.7% men), 327 (60.0%) were willing to pay something for screening, of whom 273 (83.5%) would pay RMB10-RMB30 (US$1.6-US$4.7). People living in rural areas and those from lower-income families were more likely to be willing to pay anything, while men, urban residents and those covered by employer-linked insurance were willing to pay larger sums (p<0.05 for all)., Conclusion: Nearly two-thirds of participants were willing to pay for screening in this screening programme organised at the primary care level in rural China. This finding offers the potential that such activities can be sustained and scaled up through user fees., Competing Interests: Competing interests: None declared., (© Author(s) (or their employer(s)) 2023. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.)

Background: Active breathing coordinator (ABC)-assisted deep inspiration breath hold (DIBH) is an important organ sparing radiation therapy (RT) technique for left-sided breast cancer patients. Patients with advanced breast cancer undergoing chest wall and regional nodal irradiation often require a field matching technique. While field matching has been demonstrated to be safe and effective in free breathing patients, its safety and accuracy in DIBH/ABC use has not been previously reported., Purpose: To report the accuracy, feasibility, and safety of field matching with ABC/DIBH for patients receiving breast/chest wall irradiation with nodal irradiation using a three-field technique., Methods: From December 2012 to May 2018, breast cancer patients undergoing ABC/DIBH-based RT at a single institution were reviewed. For each fraction, the amount of overlap/gap between the supraclavicular and the tangential field were measured and recorded. Patient characteristics, including acute and delayed skin toxicities, were analyzed., Results: A total of 202 patients utilized ABC/DIBH and 4973 fractions had gap/overlap measurements available for analysis. The average gap/overlap measured at junction was 0.28 mm ± 0.99 mm. A total of 72% of fractions had no measurable gap/overlap (0 mm), while 5.6% had an overlap and 22.7% a gap. There was no significant trend for worsening or improvement of gap/overlap measurements with increasing fraction number per patient. OSLD measurements were compared to the planned dose. The median dose 1 cm above the junction was 106% ± 7% of planned dose (range 94%-116%). One centimeter below the junction, the median dose was 114% ± 11% of planned dose (range 95%-131%). At the junction, the median dose was 106% ± 16.3% of planned dose (range 86%-131%). Acute skin toxicity was similar to historically reported values (grade 3, 5.4%, grade 4, 0%)., Conclusion: ABC-assisted DIBH is a safe and technically feasible method of delivering RT in the setting of complex matching field technique for breast and regional nodal treatments., (© 2022 The Authors. Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.)

Objective : Gliomas are the most common primary brain tumors. Approximately 70% of the glioma patients diagnosed with glioblastoma have an averaged overall survival (OS) of only ∼16 months. Early survival prediction is essential for treatment decision-making in glioma patients. Here we proposed an ensemble learning approach to predict the post-operative OS of glioma patients using only pre-operative MRIs. Approach : Our dataset was from the Medical Image Computing and Computer Assisted Intervention Brain Tumor Segmentation challenge 2020, which consists of multimodal pre-operative MRI scans of 235 glioma patients with survival days recorded. The backbone of our approach was a Siamese network consisting of twinned ResNet-based feature extractors followed by a 3-layer classifier. During training, the feature extractors explored traits of intra and inter-class by minimizing contrastive loss of randomly paired 2D pre-operative MRIs, and the classifier utilized the extracted features to generate labels with cost defined by cross-entropy loss. During testing, the extracted features were also utilized to define distance between the test sample and the reference composed of training data, to generate an additional predictor via K-NN classification. The final label was the ensemble classification from both the Siamese model and the K-NN model. Main results : Our approach classifies the glioma patients into 3 OS classes: long-survivors (>15 months), mid-survivors (between 10 and 15 months) and short-survivors (<10 months). The performance is assessed by the accuracy (ACC) and the area under the curve (AUC) of 3-class classification. The final result achieved an ACC of 65.22% and AUC of 0.81. Significance : Our Siamese network based ensemble learning approach demonstrated promising ability in mining discriminative features with minimal manual processing and generalization requirement. This prediction strategy can be potentially applied to assist timely clinical decision-making., (© 2022 Institute of Physics and Engineering in Medicine.)

Purpose: Many radiotherapy modalities can deliver concentrated radiation in the form of spots, such as Gamma Knife (GK), GammaPod (GP), intensity-modulated proton therapy, and brachytherapy, and can be generalized as spot-based treatments. These treatments have a great therapeutic advantage of creating potent target dose while sparing the surrounding normal tissues. However, global optimization to determine the spot positions, shapes, and intensities is an intractable combinatorial problem for any real 3D problem. The conventional approach adopts heuristic spot selection and intensity optimization in a sequential manner to mitigate the problem complexity. In this work, we propose a novel framework that enables global optimization of spot-based treatment planning., Methods: The framework is based on kernel decomposition (KD) dose calculation, which models each spot dose as a scaled shift-invariant kernel, with the reference kernels and scales pre-calculated. During optimization, the framework incorporates Fast Fourier Transform (FFT) for objective and derivative evaluations and accommodates all spot candidates in optimization search with a temporal complexity of O(N 3 log N) as opposed to O(N 6 ) complexity in the conventional beamlet framework for volume dimensions of N × N × N. We demonstrated the FFT framework using simulations with different objectives. The framework's planning performance was illustrated using clinical GK and GP cases., Results: Pre-processing involves only a small number of reference kernels and a scale map for the KD model with marginal spatial and temporal overheads. For simulations with 512 × 512 image dimensions, plan optimization finished in ∼2 seconds with FFT, whereas it took 100× longer with the beamlet approach. For clinical cases, the FFT attained solutions within a minute with improved plan quality compared to clinical plans: better conformity and less integral dose because of using a global fine search space for optimal spots., Conclusions: The scaled shift-invariance and FFT framework opens a new paradigm for spot-based treatment planning, as it can substantially reduce both the spatial and temporal complexities. The framework makes global optimization for spot-based treatment planning clinically feasible., (© 2022 American Association of Physicists in Medicine.)

Online analysis of chemical composition of cigarette smoke of a heated tobacco product (HTP) was performed by using a home-made vacuum ultraviolet (VUV) lamp-based photoionization time-of-flight (TOF) mass spectrometer. A capillary inlet and an aerodynamic lens were utilized to sample the gas- and particulate-phase of the HTP smoke, without dilution and pretreatment, which can be switched from each other within minutes. A thermal desorption unit was installed to vaporize the particulate-phase into gas and its vaporization temperature was determined, based on an equilibrium between the evaporation efficiency and the thermal decomposition of organic compounds. Then these species were softly ionized by VUV photons and their ions were measured by a reflectron TOF mass analyzer. Meanwhile, the puff-by-puff resolved size distributions of the HTP smoke were probed with a commercial scanning mobility particle sizers (SMPS). The mean diameters of particles firstly increase with the puff number, mainly located in the range of 200 - 300 nm, and then approached a steady state. This method was validated to measure the physical-chemical characteristics of the HTP cigarette smoke.•A capillary inlet and an aerodynamic lens were utilized to sample the gas- and particulate-phase of the HTP smoke.•Chemical composition of the HTP smoke was measured by using a compact VUV photoionization mass spectrometer.•The particle size distribution of the HTP smoke without dilution was measured online., (© 2022 The Author(s).)

Purpose: To investigate the influence of keratometric astigmatism on visual outcomes following small incision lenticule extraction (SMILE)., Methods: Eighty eyes undergoing SMILE for myopia correction were classified into two groups based on preoperative keratometric astigmatism: low keratometric astigmatism (LA) and high keratometric astigmatism (HA) groups. Refractive outcomes, vector components, and changes in higher order aberrations (HOAs) were evaluated preoperatively and 6 months postoperatively., Results: At the postoperative 6-month visit, no significant difference was observed in the decentered distance between the HA and LA groups (HA: 0.17 ± 0.08 mm, LA: 0.16 ± 0.08 mm, P = 0.189). No significant differences in the correction index ( P = 0.481), absolute angle of error ( P = 0.104), or index of success ( P = 0.147) were observed between the two groups. There was no significant difference in the induction of corneal aberrations between the two groups. Furthermore, there were no significant associations between the decentered distance and the vector components of astigmatic correction or induction of higher-order aberration in the HA group ( P ≥ 0.294, P ≥ 0.112) or the LA group ( P ≥ 0.323, P ≥ 0.080)., Conclusions: SMILE for high keratometric astigmatism could achieve comparable treatment centration and visual quality to that of low keratometric astigmatism., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2022 Liu, Yu, Lu, Cheng, Gu, Liu and Zhou.)

Purpose: Adaptive radiotherapy (ART), especially online ART, effectively accounts for positioning errors and anatomical changes. One key component of online ART process is accurately and efficiently delineating organs at risk (OARs) and targets on online images, such as cone beam computed tomography (CBCT). Direct application of deep learning (DL)-based segmentation to CBCT images suffered from issues such as low image quality and limited available contour labels for training. To overcome these obstacles to online CBCT segmentation, we propose a registration-guided DL (RgDL) segmentation framework that integrates image registration algorithms and DL segmentation models., Methods: The RgDL framework is composed of two components: image registration and RgDL segmentation. The image registration algorithm transforms/deforms planning contours that were subsequently used as guidance by the DL model to obtain accurate final segmentations. We had two implementations of the proposed framework-Rig-RgDL (Rig for rigid body) and Def-RgDL (Def for deformable)-with rigid body (RB) registration or deformable image registration (DIR) as the registration algorithm, respectively, and U-Net as the DL model architecture. The two implementations of RgDL framework were trained and evaluated on seven OARs in an institutional clinical head-and-neck dataset., Results: Compared to the baseline approaches using the registration or the DL alone, RgDLs achieved more accurate segmentation, as measured by higher mean Dice similarity coefficients (DSCs) and other distance-based metrics. Rig-RgDL achieved a DSC of 84.5% on seven OARs on average, higher than RB or DL alone by 4.5% and 4.7%. The average DSC of Def-RgDL was 86.5%, higher than DIR or DL alone by 2.4% and 6.7%. The inference time required by the DL model component to generate final segmentations of seven OARs was less than 1 s in RgDL. By examining the contours from RgDLs and DL case by case, we found that RgDL was less susceptible to image artifacts. We also studied how the performances of RgDL and DL vary with the size of the training dataset. The DSC of DL dropped by 12.1% as the number of training data decreased from 22 to 5, whereas RgDL only dropped by 3.4%., Conclusion: By incorporating the patient-specific registration guidance to a population-based DL segmentation model, RgDL framework overcame the obstacles associated with online CBCT segmentation, including low image quality and insufficient training data, and achieved better segmentation accuracy than baseline methods. The resulting segmentation accuracy and efficiency show promise for applying this RgDL framework for online ART., (© 2022 American Association of Physicists in Medicine.)

In 2019, our institution became the second in the world to go live with GammaPod (Xcision Medical Systems, LLC, Columbia, MD), a device dedicated for stereotactic radiation therapy of breast cancer, with breast immobilization, real-time imaging, and highly-conformal dosimetry. At our institution, GammaPod is used for 5-fraction adjuvant partial breast irradiation, single-fraction tumor cavity boost before whole-breast irradiation, single-fraction preoperative radiation, and (in poor surgical candidates), single-fraction definitive radiation. Here, we describe our workflow, observed procedure step times, and homegrown techniques for improved efficiency in our institutional experience of 93 patients treated between 2019 and 2021., (Published by Elsevier Inc.)

This paper presents the data of chemical composition of the particles from OH oxidation reaction of 1,3,5-trimethylbenzene (1,3,5-TMB). The particle-phase compositions are measured on-line by using a vacuum ultraviolet (VUV) photoionization aerosol mass spectrometer. The assignments of the major peaks of photoionization mass spectrum, as well as their molecular structures, are presented. The optimized structures of the reactants, intermediates and transition states involved in the reaction of the bicyclic peroxy radical with HO 2 are shown. The reaction routes of the OH-initiated oxidation of the deuterated 1,3,5-TMB sample are also calculated and displayed for comparison. The data presented here is related to the paper "Direct observation of the particle-phase bicyclic products from OH-initiated oxidation of 1,3,5-trimethylbenzene under NOx-free conditions" by Lin et al. (2022)., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2022 The Author(s).)

Purpose: To investigate the achieved optical zone and functional optical zone in myopic eyes with high astigmatism after small incision lenticule extraction (SMILE)., Methods: Seventy-five eyes of 75 consecutive patients with myopia before SMILE were included and divided into two groups (no astigmatism group: without astigmatism vs high astigmatism group: with astigmatism > 2.00 diopters). The achieved optical zone, functional optical zone, and corneal aberrations were compared using Scheimpflug imaging at 6 months postoperatively. Correlations between corneal aberrations and functional optical zone were analyzed., Results: The mean achieved optical zone diameter was smaller in the no astigmatism group than in the high astigmatism group. There were no significant differences in achieved optical zone among the different meridians in each group. The mean functional optical zone diameter was significantly smaller in the no astigmatism group than in the high astigmatism group. Compared with the programmed optical zone, significant reduction of the functional optical zone was found in the two groups. The reduction of the functional optical zone was significantly smaller in the high astigmatism group than in the no astigmatism group, except for the 60°, 90°, and 120° meridians. There were significant correlations between the functional optical zone area and the induced corneal aberrations in the high astigmatism group, except for horizontal coma, whereas significant correlations of functional optical zone with total higher order aberrations and spherical aberration were detected in the no astigmatism group., Conclusions: Myopic eyes with high astigmatism following SMILE had a circular achieved optical zone, but the functional optical zone was oval and larger than in eyes without astigmatism. Moreover, a lower spherical aberration was observed in the patients with high astigmatism. [ J Refract Surg . 2022;38(4):243-249.] .

Purpose: We report on our early experience of our prospective multicenter phase 1 dose- escalation study of single-fraction stereotactic partial breast irradiation (S-PBI) for early stage breast cancer after partial mastectomy using a robotic stereotactic radiation system., Methods and Materials: Thirty women with in situ or invasive breast cancer stage 0, I, or II with tumor size <3 cm treated with lumpectomy were enrolled in this phase 1 single-fraction S-PBI dose-escalation trial. Women received either 22.5, 26.5, or 30 Gy in a single fraction using a robotic stereotactic radiation system. The primary outcome was to reach tumoricidal dose of 30 Gy in a single fraction to the lumpectomy cavity without exceeding the maximum tolerated dose. Secondary outcomes were to determine dose-limiting toxicity and cosmesis. Tertiary goals were ipsilateral breast recurrence rate, distant disease-free interval, recurrence-free survival, and overall survival., Results: From June 2016 to January 2021, 11, 8, and 10 patients were treated to doses of 22.5, 26.5, or 30 Gy in a single fraction, respectively, with median follow-up being 47.9, 25.1, and 16.2 months. No patients experienced acute (<90 days) grade 3 or higher treatment-related toxicity, and maximum tolerated dose was not reached. There were 2 delayed grade 3 toxicities. Four patients (13.8%) developed fat necrosis across all 3 cohorts, which compares favorably with results from other PBI trials. No dose cohort had a statistically significant cosmetic detriment from baseline to 12 months or 24 months follow-up by patient- or physician-reported global cosmetic scores. There were no reports of disease recurrence., Conclusions: This phase 1 trial demonstrates that S-PBI can be used to safely escalate dose to 30 Gy in a single fraction with low toxicity and without detriment in cosmesis relative to baseline., (Published by Elsevier Inc.)

Total body irradiation is an important part of the conditioning regimens frequently used to prepare patients for allogeneic hematopoietic stem cell transplantation (SCT). Volumetric-modulated arc therapy enabled total body irradiation (VMAT-TBI), an alternative to conventional TBI (cTBI), is a novel radiotherapy treatment technique that has been implemented and investigated in our institution. The purpose of this study is to (1) report our six-year clinical experience in terms of treatment planning strategy and delivery time and (2) evaluate the clinical outcomes and toxicities in our cohort of patients treated with VMAT-TBI. This is a retrospective single center study. Forty-four patients at our institution received VMAT-TBI and chemotherapy conditioning followed by allogeneic SCT between 2014 and 2020. Thirty-two patients (73%) received standard-dose TBI (12-13.2 Gy in 6-8 fractions twice daily), whereas 12 (27%) received low-dose TBI (2-4 Gy in one fraction). Treatment planning, delivery, and treatment outcome data including overall survival (OS), relapse-free survival (RFS), and toxicities were analyzed. The developed VMAT-TBI planning strategy consistently generated plans satisfying our dose constraints, with planning target volume coverage >90%, mean lung dose ∼50% to 75% of prescription dose, and minimal hotspots in critical organs. Most of the treatment deliveries were <100 minutes (range 33-147, mean 72). The median follow-up was 26 months. At the last follow-up, 34 of 44 (77%) of patients were alive, with 1- and 2-year OS of 90% and 79% and RFS of 88% and 71%, respectively. The most common grade 3+ toxicities observed were mucositis (31 patients [71%]) and nephrotoxicity (6 patients [13%]), both of which were deemed multifactorial in cause. Four patients (9%) in standard-dose cohort developed grade 3+ pneumonitis, with 3 cases in the setting of documented respiratory infection and only 1 (2%) deemed likely related to radiation alone. VMAT-TBI provides a safe alternative to cTBI. The dose modulation capability of VMAT-TBI may lead to new treatment strategies, such as simultaneous boost and further critical organ sparing, for better malignant cell eradication, immune suppression, and lower toxicities., (Published by Elsevier Inc.)

We present a vacuum ultraviolet (VUV) lamp-based photoionization time-of-flight (TOF) mass spectrometer coupled with a capillary inlet and an aerodynamic lens to online analyze the chemical compositions of the gas- and particulate-phase of cigarette smoke of a heated tobacco product (HTP). Both phase compositions of the fresh cigarette smoke, without dilution and pretreatment, are softly photoionized and their mass spectra are measured with a time resolution of 1 s. It is shown that the gas-phase compositions with low mass are volatile organic compounds (VOCs), and the particulate-phase compositions are also clearly identified and cover the full mass range of the mass spectrometer. The time- or puff-by-puff resolved dynamic data are obtained for each species and provide abundant information to unravel the chemistry of the HTP smoke. In addition, the present results show that besides thermal vaporization, a couple of chemical reactions including pyrolysis and degradation have also occurred in the HTP smoking process, although its operation temperature is less than 350 °C. Even if not done here, this study paves the way to analyze the gas- and particulate-phase chemical compositions of a complex system in real time, like the cigarette smoke presented here, by using advanced soft ionization mass spectrometry., (Copyright © 2021 Elsevier B.V. All rights reserved.)

Purpose: Pre-calculation of accurate dose deposition kernels for treatment planning of spot-based radiotherapies, such as Gamma Knife (GK) and Gamma Pod (GP), can be very time-consuming and may require large data storage with an enormous number of possible spots. We proposed a novel kernel decomposition (KD) model to address accurate and fast (real-time) dose calculation with reduced data storage requirements for spot-based treatment planning. The application of the KD model was demonstrated for clinical GK and GP radiotherapy platforms., Methods: The dose deposition kernel at each spot (shot position) is modeled as the product of a shift-invariant kernel based on a reference kernel and spatially variant scale factor. The reference kernel, one for each collimator, is defined at the center of the commissioning phantom for GK and at the center of the treatment target for GP and calculated using the Monte Carlo (MC) method. The spatially variant scale factor is defined as the ratio of the mean tissue maximum ratio (TMR) at the candidate shot position to that at the reference kernel position, and the mean TMR map is calculated within the entire volume through parallel beam ray tracing on the density image followed by averaging over all source directions. The proposed KD dose calculations were compared with the MC method and with the GK and GP treatment planning system (TPS) computations for various shot positions and collimator sizes utilizing a phantom and 14 and 12 clinical plans for GK and GP, respectively., Results: For the phantom study, the KD Gamma index (3%/1 mm) passing rates were greater than 99% (median 100%) relative to the MC doses, except for the shots close to the boundary. The passing rates dropped below 90% for 8 mm (16 mm) shots positioned within ∼1 cm (∼2 cm) of the boundary. For the clinical GK plans, the KD Gamma passing rates were greater than 99% (median 100%) compared to the MC and greater than 92% (median 99%) compared to the TPS. For the clinical GP plans, the KD Gamma passing rates were greater than 95% (median 98%) compared to the MC and greater than 91% (median 97%) compared to the TPS. The scale factors were calculated in sub-seconds with GPU implementation and only need to be calculated once before treatment plan optimization. The calculation of the dose kernel was also within sub-seconds without requiring beam-by-beam calculation commonly done in the TPS., Conclusion: The proposed model can provide an accurate dose and enables real-time dose and derivative calculations by kernel shifting and scaling without pre-calculating or requiring large data storage for GK and GP dose deposition kernels during treatment planning. This model could be useful for spot-based radiotherapy treatment planning by allowing an efficient global fine search for optimal spots., (© 2021 American Association of Physicists in Medicine.)

Stereotactic radiosurgery (SRS) is now the standard of care for brain metastases (BMs) patients. The SRS treatment planning process requires precise target delineation, which in clinical workflow for patients with multiple (>4) BMs (mBMs) could become a pronounced time bottleneck. Our group has developed an automated BMs segmentation platform to assist in this process. The accuracy of the auto-segmentation, however, is influenced by the presence of false-positive segmentations, mainly caused by the injected contrast during MRI acquisition. To address this problem and further improve the segmentation performance, a deep-learning and radiomics ensemble classifier was developed to reduce the false-positive rate in segmentations. The proposed model consists of a Siamese network and a radiomic-based support vector machine (SVM) classifier. The 2D-based Siamese network contains a pair of parallel feature extractors with shared weights followed by a single classifier. This architecture is designed to identify the inter-class difference. On the other hand, the SVM model takes the radiomic features extracted from 3D segmentation volumes as the input for twofold classification, either a false-positive segmentation or a true BM. Lastly, the outputs from both models create an ensemble to generate the final label. The performance of the proposed model in the segmented mBMs testing dataset reached the accuracy (ACC), sensitivity (SEN), specificity (SPE) and area under the curve of 0.91, 0.96, 0.90 and 0.93, respectively. After integrating the proposed model into the original segmentation platform, the average segmentation false negative rate (FNR) and the false positive over the union (FPoU) were 0.13 and 0.09, respectively, which preserved the initial FNR (0.07) and significantly improved the FPoU (0.55). The proposed method effectively reduced the false-positive rate in the BMs raw segmentations indicating that the integration of the proposed ensemble classifier into the BMs segmentation platform provides a beneficial tool for mBMs SRS management., (© 2022 Institute of Physics and Engineering in Medicine.)

We study the conformers of the ethyl peroxy radical (C 2 H 5 O 2 ), the simplest peroxy radical having more than one conformer, by combining synchrotron radiation vacuum ultraviolet (VUV) photoionization mass spectrometry with theoretical calculations. The ethyl peroxy radical is formed in a microwave discharge flow tube through the reaction of the ethyl radical (C 2 H 5 ) with oxygen molecules, where C 2 H 5 is generated via the hydrogen-abstraction reaction of ethane with fluorine atoms. Two kinds of C 2 H 5 + , originating from photoionization of C 2 H 5 and from dissociative photoionization of C 2 H 5 O 2 , whose cation is not stable, have been identified and separated in photoionization mass spectra. The photoionization spectrum corresponding to C 2 H 5 O 2 is obtained and assigned with Franck-Condon calculations. The present findings show that the gauche conformer ( G -C 2 H 5 O 2 ) of C 2 H 5 O 2 has favorable Franck-Condon factors in the ionization transitions, whereas the contribution of the trans conformer ( T -C 2 H 5 O 2 ) to the photoionization spectrum is minor or negligible due to its large geometric changes in the photoionization process. Moreover, the reason for the instability of C 2 H 5 O 2 + and its detailed dissociation mechanisms have been unraveled with the aid of the calculated potential energy curves.

Efficient, reliable and reproducible target volume delineation is a key step in the effective planning of breast radiotherapy. However, post-operative breast target delineation is challenging as the contrast between the tumor bed volume (TBV) and normal breast tissue is relatively low in CT images. In this study, we propose to mimic the marker-guidance procedure in manual target delineation. We developed a saliency-based deep learning segmentation (SDL-Seg) algorithm for accurate TBV segmentation in post-operative breast irradiation. The SDL-Seg algorithm incorporates saliency information in the form of markers' location cues into a U-Net model. The design forces the model to encode the location-related features, which underscores regions with high saliency levels and suppresses low saliency regions. The saliency maps were generated by identifying markers on CT images. Markers' location were then converted to probability maps using a distance transformation coupled with a Gaussian filter. Subsequently, the CT images and the corresponding saliency maps formed a multi-channel input for the SDL-Seg network. Our in-house dataset was comprised of 145 prone CT images from 29 post-operative breast cancer patients, who received 5-fraction partial breast irradiation (PBI) regimen on GammaPod. The 29 patients were randomly split into training (19), validation (5) and test (5) sets. The performance of the proposed method was compared against basic U-Net. Our model achieved mean (standard deviation) of 76.4(±2.7) %, 6.76(±1.83) mm, and 1.9(±0.66) mm for Dice similarity coefficient, 95 percentile Hausdorff distance, and average symmetric surface distance respectively on the test set with computation time of below 11 seconds per one CT volume. SDL-Seg showed superior performance relative to basic U-Net for all the evaluation metrics while preserving low computation cost. The findings demonstrate that SDL-Seg is a promising approach for improving the efficiency and accuracy of the on-line treatment planning procedure of PBI, such as GammaPod based PBI., (© 2021 Institute of Physics and Engineering in Medicine.)

Purpose: Our purpose was to evaluate cosmetic changes after 5-fraction adjuvant stereotactic partial breast irradiation (S-PBI)., Methods and Materials: Seventy-five women with in situ or invasive breast cancer stage 0, I, or II, with tumor size ≤3 cm, were enrolled after lumpectomy in a phase 1 dose escalation trial of S-PBI into cohorts receiving 30, 32.5, 35, 37.5, or 40 Gy in 5 fractions. Before S-PBI, 3 to 4 gold fiducial markers were placed in the lumpectomy cavity for tracking with the Synchrony respiratory tracking system. S-PBI was delivered with a CyberKnife robotic radiosurgery system. Patients and physicians evaluated global cosmesis using the Harvard Breast Cosmesis Scale. Eight independent panelists evaluated digital photography for global cosmesis and 10 subdomains at baseline and follow-up. McNemar tests were used to evaluate change in cosmesis, graded as excellent/good or fair/poor, from baseline to year 3. Wilcoxon signed rank tests were used to evaluate change in subdomains. Cohen's kappa (κ) statistic was used to estimate interobserver agreement (IOA) between raters, and Fleiss' κ was used to estimate IOA between panelists., Results: Median cosmetic follow-up was 5, 5, 5, 4, and 3 years for the 30, 32.5, 35, 37.5, and 40 Gy cohorts. Most patients reported excellent/good cosmesis at both baseline (86.3%) and year 3 (89.8%). No dose cohort had significantly worsened cosmesis by year 3 on McNemar analysis. No cosmetic subdomain had significant worsening by year 3. IOA was fair for patient-physician (κ = 0.300, P < .001), patient-panel (κ = 0.295, P < .001), physician-panel (κ = 0.256, P < .001), and individual panelists (Fleiss κ = 0.327, P < .001)., Conclusions: Dose escalation of S-PBI from 30 to 40 Gy in 5 fractions for early stage breast cancer was not associated with a detectable change in cosmesis by year 3. S-PBI is a promising modality for treatment of early stage breast cancer., (Published by Elsevier Inc.)

Purpose: Stereotactic radiosurgery (SRS) has become a primary treatment for multiple brain metastases (BM) but may require distribution of BMs over several sessions to make delivery time and radiation toxicity manageable. Contrasting to equal fraction dose in conventional fractionation, distributed SRS delivers full dose to a subset of BMs in each session while avoiding adjacent BMs in the same session to reduce toxicity from overlapping radiation. However, current clinical treatment planning for distributed SRS relies on manual BM assignment, which can be tedious and error prone. This work describes a novel approach to automate the distribution of BM in the Gamma Knife (GK) clinical workflow., Methods: We represent each BM as an electrostatic field of the same polarity that exerts repulsive forces on other BMs in the same session. This representation naturally leads to separation of close BMs into different sessions to lower the potential energy. Indeed, the BM distribution problem can be formulated as minimization of the total potential energy from all treatment sessions subject to delivery time constraints in mixed-integer quadratic programming (MIQP). We retrospectively studied eight clinical GK cases of multiple BM and compared the automated MIQP solution with clinically used BM distribution to demonstrate the efficacy of the proposed approach., Results: With the problem size equal to the number of BMs times the number of sessions, this MIQP can be solved in a minute on a personal workstation. The MIQP solution effectively separated BMs for a given number of treatment sessions and evened out the delivery time distribution among sessions. Compared to the clinically used manual BM distributions in paired t-test for a similar range of delivery time variation, the automated BM distributions had lower energy objectives (range of decrease: [11% 89%]; median: 25%; P = . 073 ), more uniformly distributed treatment volumes (range of decrease for the normalized standard deviation of volume distribution: [0.02 0.95]; median: 0.16; P = . 013 ), more scattered BMs in each treatment session (range of increase for the mean minimum BM distance: [0 14] mm; median: 6 mm; P = . 008 ), and lower overall V 12 (range of decrease: [0.0 1.6] cc; median: 0.2 cc; P = . 052 ). Moreover, without distribution, that is, with all BMs treated in the same session, V 12 was substantially larger compared to both manual and automated BM distributions; the increase ranged from 0.1 to 16.6 cc with a median of 1.3 cc., Conclusions: The proposed approach models the clinical practice and provides an efficient solution for optimal selection of BM subsets for distributed SRS. Further evaluations are underway to establish this approach as a tool for improving clinical workflow and to facilitate systematic study on the benefits of distributed SRS treatments., (© 2021 American Association of Physicists in Medicine.)

Long non-coding RNAs (lncRNAs) are transcripts longer than 200 nucleotides but not translated into proteins. LncRNAs regulate gene expressions at multiple levels, such as chromatin, transcription, and post-transcription. Further, lncRNAs participate in various biological processes such as cell differentiation, cell cycle regulation, and maintenance of stem cell pluripotency. We have previously reported that lncRNAs are closely related to programmed cell death (PCD), which includes apoptosis, autophagy, necroptosis, and ferroptosis. Overexpression of lncRNA can suppress the extrinsic apoptosis pathway by downregulating of membrane receptors and protect tumor cells by inhibiting the expression of necroptosis-related proteins. Some lncRNAs can also act as competitive endogenous RNA to prevent oxidation, thereby inhibiting ferroptosis, while some are known to activate autophagy. The relationship between lncRNA and PCD has promising implications in clinical research, and reports have highlighted this relationship in various cancers such as non-small cell lung cancer and gastric cancer. This review systematically summarizes the advances in the understanding of the molecular mechanisms through which lncRNAs impact PCD.

The two isomers of the propylperoxy radical 1-C 3 H 7 O 2 and 2-C 3 H 7 O 2 , together with their individual rotamers, are identified and assigned by threshold photoelectron spectroscopy with the aid of high-level theoretical computations, from which their accurate adiabatic ionization energies are derived. This study paves the way to probing elusive peroxy radicals and their isomers in advanced mass spectrometry analysis of combustion and atmospheric reactions.

Background and Purpose: Involuntary motion due to swallowing cause inaccurate dose delivery during larynx radiotherapy, a deviation that may be particularly problematic during stereotactic body radiation therapy (SBRT). The goal of this study was to develop a motion management solution for larynx SBRT using surface imaging., Material and Methods: Ten patients were recently treated on a phase II study of larynx SBRT on a LINAC equipped with a surface guidance system. A small region of the immobilization mask was manually cut open to allow surface tracking. Pre-treatment and intra-fractional CBCTs were acquired to verify internal anatomy. Patients were verbally instructed not to swallow during treatment. During treatment delivery, beam hold was initiated by the Motion Management Interface if surface motion exceeded a patient-specific threshold. Patient motion was recorded in log files and analyzed. We also performed phantom studies to assess the theoretical impact of gating on dose delivery., Results: The frequency (6.5 ± 5.2 times per fraction) and duration (3.9 ± 2.5 seconds per swallow) of swallowing varied both between patients and fractions. The magnitude of each swallow showed mean peak amplitude at 5.8 ± 3.8 mm above baseline, mostly in the longitudinal direction. Beam duty cycle was 95.0% ± 7.0% (absolute range: 76-100%), with inefficiency most prominent in the early fractions. The 95th percentile residual motion was reduced from 3.4 mm to 2.3 mm with both verbal instruction and gating. Phantom studies confirmed dose delivery accuracy represented by gamma pass rate was improved by 5% using this approach., Conclusions: Laryngeal motion management using surface imaging is feasible and efficacious. Uncontrolled movement of the larynx was not uncommon during treatment, with gating reducing potential for unplanned dose deviations. Additional research is needed to determine the clinical benefit with this system., (Copyright © 2020 Elsevier B.V. All rights reserved.)

We developed a fluence map prediction method that directly generates fluence maps for a given desired dose distribution without optimization for volumetric modulated arc therapy (VMAT) planning. The prediction consists of two steps. First, projections of the desired dose are calculated and then inversely mapped to fluence maps in the phantom geometry by a deep neural network. Second, a plan scaling technique is applied to scale fluence maps from phantom to patient geometry. We evaluated the performance of the proposed fluence map prediction method for 102 head and neck (H&N) and 14 prostate cancer VMAT plans by comparing the patient doses calculated from the predicted fluence maps with the given desired dose distributions. The mean dose differences were 1.42% ± 0.37%, 1.53% ± 0.44% and 1.25% ± 0.44% for the planning target volume (PTV), the region from the PTV boundary to the 50% isodose line, and the region from the 50% to the 20% isodose line, respectively. The gamma passing rate was 98.06% ± 2.64% with the 3 mm/3% criterion. The prediction time for a single VMAT plan was less than one second. In conclusion, we developed an inverse mapping-based method that predicts fluence maps for desired dose distributions with high accuracy. Our method is effectively an optimization-free inverse planning approach, which was orders of magnitude faster than fluence map optimization. Combining the proposed method with leaf sequencing has the potential to dramatically speed up VMAT treatment planning., (© 2020 Institute of Physics and Engineering in Medicine.)

Deep learning (DL)-based auto-segmentation has the potential for accurate organ delineation in radiotherapy applications but requires large amounts of clean labeled data to train a robust model. However, annotating medical images is extremely time-consuming and requires clinical expertise, especially for segmentation that demands voxel-wise labels. On the other hand, medical images without annotations are abundant and highly accessible. To alleviate the influence of the limited number of clean labels, we propose a weakly supervised DL training approach using deformable image registration (DIR)-based annotations, leveraging the abundance of unlabeled data. We generate pseudo-contours by utilizing DIR to propagate atlas contours onto abundant unlabeled images and train a robust DL-based segmentation model. With 10 labeled TCIA dataset and 50 unlabeled CT scans from our institution, our model achieved Dice similarity coefficient of 87.9%, 73.4%, 73.4%, 63.2% and 61.0% on mandible, left & right parotid glands and left & right submandibular glands of TCIA test set and competitive performance on our institutional clinical dataset and a third party (PDDCA) dataset. Experimental results demonstrated the proposed method outperformed traditional multi-atlas DIR methods and fully supervised limited data training and is promising for DL-based medical image segmentation application with limited annotated data., (© 2020 Institute of Physics and Engineering in Medicine.)

Purpose: This study reports predictive dosimetric and physiologic factors for fat necrosis after stereotactic-partial breast irradiation (S-PBI)., Methods and Materials: Seventy-five patients with ductal carcinoma-in situ or invasive nonlobular epithelial histologies stage 0, I, or II, with tumor size <3 cm were enrolled in a dose-escalation, phase I S-PBI trial between January 2011 and July 2015. Fat necrosis was evaluated clinically at each follow-up. Treatment data were extracted from the Multiplan Treatment Planning System (Cyberknife, Accuray). Univariate and stepwise logistic regression analyses were conducted to identify factors associated with palpable fat necrosis., Results: With a median follow-up of 61 months (range: 4.3-99.5 months), 11 patients experienced palpable fat necrosis, 5 cases of which were painful. The median time to development of fat necrosis was 12.7 months (range, 3-42 months). On univariate analyses, higher V32.5-47.5 Gy (P < .05) and larger breast volume (P < .01) were predictive of any fat necrosis; higher V35-50 Gy (P < .05), receiving 2 treatments on consecutive days (P = .02), and higher Dmax (P = .01) were predictive of painful fat necrosis. On multivariate analyses, breast volume larger than 1063 cm 3 remained a predictive factor for any fat necrosis; receiving 2 treatments on consecutive days and higher V45 Gy were predictive of painful fat necrosis. Breast laterality, planning target volume (PTV), race, body mass index, diabetic status, and tobacco or drug use were not significantly associated with fat necrosis on univariate analysis., Conclusions: Early-stage breast cancer patients treated with breast conserving surgery and S-PBI in our study had a fat necrosis rate comparable to other accelerated partial breast irradiation modalities, but S-PBI is less invasive. To reduce risk of painful fat necrosis, we recommend not delivering fractions on consecutive days; limiting V42.5 < 50 cm 3 , V45 < 20 cm 3 , V47.5 < 1 cm 3 , Dmax ≤ 48 Gy and PTV < 100 cm 3 when feasible; and counseling patients about the increased risk for fat necrosis when constraints are not met and for those with breast volume >1000 cm 3 ., (Published by Elsevier Inc.)

We report a synchrotron radiation vacuum ultraviolet photoionization study of the hydroperoxyl radical (HO 2 ), a key reaction intermediate in combustion and atmospheric chemistry as well as astrochemistry, using double imaging photoelectron photoion coincidence spectroscopy. The HO 2 radical is formed in a microwave discharge flow tube reactor through a set of reactions initiated by F atoms in a CH 4 /O 2 /He gas mixture. The high-resolution threshold photoelectron spectrum of HO 2 in the 11 eV-12 eV energy range is acquired without interferences from other species and assigned with the aid of theoretically calculated adiabatic ionization energies (AIEs) and Franck-Condon factors. The three vibrational modes of the radical cation HO 2 + , the H-O stretch, the H-O-O bend, and the O-O stretch, have been identified, and their individual frequencies are measured. In addition, the AIEs of the X 3 A″ ground state and the a 1 A' first excited electronic state of HO 2 + are experimentally determined at 11.359 ± 0.003 eV and 11.639 ± 0.005 eV, respectively, in agreement with high-level theoretically computed results. Furthermore, the former AIE value provides validation of thermochemical networks used to extract the enthalpy of formation of the HO 2 radical.

Partly due to the use of exhaustive-annotated data, deep networks have achieved impressive performance on medical image segmentation. Medical imaging data paired with noisy annotation are, however, ubiquitous, but little is known about the effect of noisy annotation on deep learning based medical image segmentation. We studied the effect of noisy annotation in the context of mandible segmentation from CT images. First, 202 images of head and neck cancer patients were collected from our clinical database, where the organs-at-risk were annotated by one of twelve planning dosimetrists. The mandibles were roughly annotated as the planning avoiding structure. Then, mandible labels were checked and corrected by a head and neck specialist to get the reference standard. At last, by varying the ratios of noisy labels in the training set, deep networks were trained and tested for mandible segmentation. The trained models were further tested on other two public datasets. Experimental results indicated that the network trained with noisy labels had worse segmentation than that trained with reference standard, and in general, fewer noisy labels led to better performance. When using 20% or less noisy cases for training, no significant difference was found on the segmentation results between the models trained by noisy or reference annotation. Cross-dataset validation results verified that the models trained with noisy data achieved competitive performance to that trained with reference standard. This study suggests that the involved network is robust to noisy annotation to some extent in mandible segmentation from CT images. It also highlights the importance of labeling quality in deep learning. In the future work, extra attention should be paid to how to utilize a small number of reference standard samples to improve the performance of deep learning with noisy annotation.

Purpose: GammaPod, a stereotactic partial breast irradiator allowing highly conformal radiation dose delivery, has its unique mechanical design and treatment planning system (TPS). However, the uniqueness of the system poses challenges regarding initial GammaPod system commissioning and routine quality assurance (QA). In this study, we report POD-DOSI, a dedicated dosimetry system for accurate and efficient commissioning and QA of GammaPod., Materials and Methods: The POD-DOSI system consists of two subsystems, POD-Scanner and POD-Calculator. The POD-Scanner is an automatic ion-chamber positioning system driven by two translational stepper motors for anterior-posterior, longitudinal, and lateral beam scanning. The stepper motors are controlled by a microcomputer through an in-house-developed graphical user interface, which can be remotely accessed by a laptop via wireless connection. The POD-Calculator is a commissioned GPU-based Monte Carlo dose calculation engine, which calculates dose by transporting particles from phase space constructed for GammaPod. In our institution, the POD-DOSI system was used for GammaPod TPS commissioning and dose verification. The POD-Calculator was further developed as a secondary dose calculation tool performing patient-specific plan QA before each treatment., Results: The POD-DOSI system has been fully evaluated and tested, both mechanically and dosimetrically, and applied successfully to drive the commissioning of our GammaPod system. The POD-Scanner achieved 0.1 mm accuracy in ion-chamber positioning tests. The POD-Calculator generated dose profiles matched well with water phantom measurements and TPS calculations to <0.5 mm accuracy. For end-to-end test on 56 different treatment plans, in-water point dose measurements by POD-Scanner were within ±2.20% of the doses calculated by POD-Calculator (range: -2.01% to 2.20%, mean: 0.04%, std&#95;dev: 1.10%). Correspondingly, when switching the calculation medium from water to breast tissue, the point doses calculated by the POD-Calculator were within ±1.60% of the point doses calculated by the GammaPOD TPS (range: -1.59% to 1.51%, mean: -0.02%, std&#95;dev: 0.73%). The average three-dimensional gamma passing rate between the GammaPod TPS doses and the POD-Calculator doses was 97.10% under the 2%/1 mm gamma criteria. The POD-DOSI system substantially shortened the GammaPod dosimetry commissioning time from weeks to days., Conclusion: The developed POD-DOSI system resolves the challenges and streamlines the process of GammaPod commissioning and QA. It improves the efficiency and accuracy for both GammaPod commissioning and routine patient-specific QA., (© 2020 American Association of Physicists in Medicine.)

Purpose: Stereotactic radiosurgery (SRS) has become a standard of care for patients' with brain metastases (BMs). However, the manual multiple BMs delineation can be time-consuming and could create an efficiency bottleneck in SRS workflow. There is a clinical need for automatic delineation and quantitative evaluation tools. In this study, building on our previous developed deep learning-based segmentation algorithms, we developed a web-based automated BMs segmentation and labeling platform to assist the SRS clinical workflow., Method: This platform was developed based on the Django framework, including a web client and a back-end server. The web client enables interactions as database access, data import, and image viewing. The server performs the segmentation and labeling tasks including: skull stripping; deep learning-based BMs segmentation; and affine registration-based BMs labeling. Additionally, the client can display BMs contours with corresponding atlas labels, and allows further postprocessing tasks including: (a) adjusting window levels; (b) displaying/hiding specific contours; (c) removing false-positive contours; (d) exporting contours as DICOM RTStruct files; etc. RESULTS: We evaluated this platform on 10 clinical cases with BMs number varied from 12-81 per case. The overall operation took about 4-5 min per patient. The segmentation accuracy was evaluated between the manual contour and automatic segmentation with several metrics. The averaged center of mass shift was 1.55 ± 0.36 mm, the Hausdorff distance was 2.98 ± 0.63 mm, the mean of surface-to-surface distance (SSD) was 1.06 ± 0.31 mm, and the standard deviation of SSD was 0.80 ± 0.16 mm. In addition, the initial averaged false-positive over union (FPoU) and false-negative rate (FNR) were 0.43 ± 0.19 and 0.15 ± 0.10 respectively. After case-specific postprocessing, the averaged FPoU and FNR were 0.19 ± 0.10 and 0.15 ± 0.10 respectively., Conclusion: The evaluated web-based BMs segmentation and labeling platform can substantially improve the clinical efficiency compared to manual contouring. This platform can be a useful tool for assisting SRS treatment planning and treatment follow-up., (© 2020 American Association of Physicists in Medicine.)

We present here a synchrotron radiation vacuum ultraviolet photoionization study of the simplest alkoxy radical, CH 3 O, a key reaction intermediate in atmospheric and combustion chemistry. A microwave discharge fast flow tube connected to a molecular beam sampling system is employed as a chemical reactor to initiate reactions and generate radicals. The CH 3 O + cation from direct ionization of the CH 3 O radical is detected successfully in the photoionization mass spectrum close to its ionization threshold. In addition, after identifying and removing the contribution of the 13 C-isotopic formaldehyde H 2 13 CO with the same isobaric mass m/z = 31, the high-resolution threshold photoelectron spectrum of CH 3 O is obtained and assigned with the aid of calculated Franck-Condon factors. The adiabatic ionization energy of CH 3 O is determined at 10.701 eV with an accuracy of 0.005 eV.

Intracranial visual pathway is related to the effective transmission of visual signals to brain. It was not only the target organ of diseases but also the organs at risk in radiotherapy thus its delineation plays an important role in both diagnosis and treatment planning. Traditional manual segmentation method suffered from time- and labor- consuming as well as intra- and inter- variability. In order to overcome these problems, state-of-the-art segmentation models were designed and various features were extracted and utilized, but it's hard to tell their effectiveness on intracranial visual pathway delineation. It's because that these methods worked on different dataset and accompanied with different training tricks. This study aimed to research the contribution of global features and local features in delineating the intracranial visual pathway from MRI scans. The two typical segmentation models, 3D UNet and DeepMedic, were chosen since they focused on global features and local features respectively. We constructed the hybrid model through serially connecting the two mentioned models to validate the performance of combined global and local features. Validation results showed that the hybrid model outperformed the individual ones. It proved that multi scale feature fusion was important in improving the segmentation performance.

We propose a novel BIRADS-SSDL network that integrates clinically-approved breast lesion characteristics (BIRADS features) into task-oriented semi-supervised deep learning (SSDL) for accurate diagnosis of ultrasound (US) images with a small training dataset. Breast US images are converted to BIRADS-oriented feature maps (BFMs) using a distance-transformation coupled with a Gaussian filter. Then, the converted BFMs are used as the input of an SSDL network, which performs unsupervised stacked convolutional auto-encoder (SCAE) image reconstruction guided by lesion classification. This integrated multi-task learning allows SCAE to extract image features with the constraints from the lesion classification task, while the lesion classification is achieved by utilizing the SCAE encoder features with a convolutional network. We trained the BIRADS-SSDL network with an alternative learning strategy by balancing the reconstruction error and classification label prediction error. To demonstrate the effectiveness of our approach, we evaluated it using two breast US image datasets. We compared the performance of the BIRADS-SSDL network with conventional SCAE and SSDL methods that use the original images as inputs, as well as with an SCAE that use BFMs as inputs. The experimental results on two breast US datasets show that BIRADS-SSDL ranked the best among the four networks, with a classification accuracy of around 94.23 ± 3.33% and 84.38 ± 3.11% on two datasets. In the case of experiments across two datasets collected from two different institutions/and US devices, the developed BIRADS-SSDL is generalizable across the different US devices and institutions without overfitting to a single dataset and achieved satisfactory results. Furthermore, we investigate the performance of the proposed method by varying the model training strategies, lesion boundary accuracy, and Gaussian filter parameters. The experimental results showed that a pre-training strategy can help to speed up model convergence during training but with no improvement of the classification accuracy on the testing dataset. The classification accuracy decreases as the segmentation accuracy decreases. The proposed BIRADS-SSDL achieves the best results among the compared methods in each case and has the capacity to deal with multiple different datasets under one model. Compared with state-of-the-art methods, BIRADS-SSDL could be promising for effective breast US computer-aided diagnosis using small datasets.