Search

Qualitative observer-based and quantitative radiomics-based analyses of T1w contrast-enhanced magnetic resonance imaging (T1w-CE MRI) have both been shown to predict the outcomes of brain metastasis (BM) stereotactic radiosurgery (SRS). Comparison of these methods and interpretation of radiomics-based machine learning (ML) models remains limited. To address this need, we collected a dataset of n = 123 BMs from 99 patients including 12 clinical features, 107 pre-treatment T1w-CE MRI radiomic features, and BM post-SRS progression scores. A previously published outcome model using SRS dose prescription and five-way BM qualitative appearance scoring was evaluated. We found high qualitative scoring interobserver variability across five observers that negatively impacted the model's risk stratification. Radiomics-based ML models trained to replicate the qualitative scoring did so with high accuracy (bootstrap-corrected AUC = 0.84-0.94), but risk stratification using these replicated qualitative scores remained poor. Radiomics-based ML models trained to directly predict post-SRS progression offered enhanced risk stratification (Kaplan-Meier rank-sum p = 0.0003) compared to using qualitative appearance. The qualitative appearance scoring enabled interpretation of the progression radiomics-based ML model, with necrotic BMs and a subset of heterogeneous BMs predicted as being at high-risk of post-SRS progression, in agreement with current radiobiological understanding. Our study's results show that while radiomics-based SRS outcome models out-perform qualitative appearance analysis, qualitative appearance still provides critical insight into ML model operation., (© 2023. The Author(s).)

Background: MRI radiomic features and machine learning have been used to predict brain metastasis (BM) stereotactic radiosurgery (SRS) outcomes. Previous studies used only single-center datasets, representing a significant barrier to clinical translation and further research. This study, therefore, presents the first dual-center validation of these techniques., Methods: SRS datasets were acquired from 2 centers ( n = 123 BMs and n = 117 BMs). Each dataset contained 8 clinical features, 107 pretreatment T1w contrast-enhanced MRI radiomic features, and post-SRS BM progression endpoints determined from follow-up MRI. Random decision forest models were used with clinical and/or radiomic features to predict progression. 250 bootstrap repetitions were used for single-center experiments., Results: Training a model with one center's dataset and testing it with the other center's dataset required using a set of features important for outcome prediction at both centers, and achieved area under the receiver operating characteristic curve (AUC) values up to 0.70. A model training methodology developed using the first center's dataset was locked and externally validated with the second center's dataset, achieving a bootstrap-corrected AUC of 0.80. Lastly, models trained on pooled data from both centers offered balanced accuracy across centers with an overall bootstrap-corrected AUC of 0.78., Conclusions: Using the presented validated methodology, radiomic models trained at a single center can be used externally, though they must utilize features important across all centers. These models' accuracies are inferior to those of models trained using each individual center's data. Pooling data across centers shows accurate and balanced performance, though further validation is required., Competing Interests: All authors: None., (© The Author(s) 2023. Published by Oxford University Press, the Society for Neuro-Oncology and the European Association of Neuro-Oncology.)

Recent studies have used T1w contrast-enhanced (T1w-CE) magnetic resonance imaging (MRI) radiomic features and machine learning to predict post-stereotactic radiosurgery (SRS) brain metastasis (BM) progression, but have not examined the effects of combining clinical and radiomic features, BM primary cancer, BM volume effects, and using multiple scanner models. To investigate these effects, a dataset of n = 123 BMs from 99 SRS patients with 12 clinical features, 107 pre-treatment T1w-CE radiomic features, and BM progression determined by follow-up MRI was used with a random decision forest model and 250 bootstrapped repetitions. Repeat experiments assessed the relative accuracy across primary cancer sites, BM volume groups, and scanner model pairings. Correction for accuracy imbalances across volume groups was investigated by removing volume-correlated features. We found that using clinical and radiomic features together produced the most accurate model with a bootstrap-corrected area under the receiver operating characteristic curve of 0.77. Accuracy also varied by primary cancer site, BM volume, and scanner model pairings. The effect of BM volume was eliminated by removing features at a volume-correlation coefficient threshold of 0.25. These results show that feature type, primary cancer, volume, and scanner model are all critical factors in the accuracy of radiomics-based prognostic models for BM SRS that must be characterised and controlled for before clinical translation., (© 2022. The Author(s).)

Purpose/objectives: High dose rate (HDR) interstitial brachytherapy (ISBT) boost is integral for definitive radiation treatment of primary vaginal cancer. Technological advances with CT or MRI guidance provide improved precision and ability to treat more extensively invasive tumors over historical techniques, but reported experience is limited. We sought to provide updated outcome and toxicity data for women with primary vaginal cancer undergoing treatment with a modern ISBT technique., Material/methods: Databases of primary vaginal carcinoma patients treated at two Canadian academic cancer institutions were combined including patient, tumor and treatment characteristics, and survival outcomes and toxicity data. Descriptive statistics, survival estimates based on the Kaplan-Meier method, and univariable/multivariable Cox proportional hazards regression analyses are reported., Results: Between 2002 and 2017, 67 women with primary vaginal cancer were treated with 3D HDR ISBT. FIGO stage distribution was I (22.4%), II (50.8%), III (17.9%), IVa (9.0%). All patients received external beam radiotherapy and HDR ISBT of 500-750 cGy per fraction over 2-4 fractions. Median follow-up was 2.68 years (95% confidence interval: 2.04-6.04). Cumulative rate of grade 3-4 genitourinary/gastrointestinal toxicity was 10.4%. Four patients developed vaginal fistula. Progression-free survival at 2 and 3 years was 73.5% and 66.4% for all patients, 78.3% and 75.0% for stage I-II and 61.6% and 46.2% for stage III-IVa, respectively (log-rank p = 0.252)., Conclusions: Use of 3D image-guided HDR ISBT boost was safe and resulted in improved survival outcomes compared to historical rates in this series of primary vaginal cancer patients. Prospective study is warranted to better define clinical and dosimetric predictors of local control., Competing Interests: Declaration of Competing Interest AVL has received honoraria unrelated to this work, from Varian Medical Systems, RefleXion and AstraZeneca., (Copyright © 2020 Elsevier Inc. All rights reserved.)

Background and Purpose: This study aimed to determine the effects of reducing the dose of contrast agent (CA) in a DCE-MRI scan on inter- and intra-observer variability in the context of MRI-guided target volume delineation for stereotactic body radiation therapy of early stage breast cancer patients. This is in hopes of reducing risks to patients due to findings of residual CA in brain and bone., Materials and Methods: Twenty-three patients receiving neoadjuvant radiation therapy were enrolled. Five observers delineated the gross target volume (GTV) using DCE-MRI for guidance. 14/23 patients received the full clinical dose of CA and 9/23 received half. Clinical target volumes (CTV) were created through a 0.5 cm uniform expansion. Several metrics were used to quantify the inter and intra-observer reliability including differences in delineation volume and the reliability coefficient., Results: There were no significant differences in the volume, though half contrast patients had a lower median for both the GTV and CTV (difference of 0.26 cm 3 and 1.27 cm 3 , respectively). All indicated a high degree of agreement between and within observers for both dose groups. However, the full dose group had a greater inter-observer variability, most likely due to the full CA causing more pronounced enhancement in the periphery., Conclusions: Reducing the dose of contrast agent did not significantly alter inter- or intra-observer variability. These results have prompted our centre to reduce the dose of gadolinium in all patients enrolled in the SIGNAL trial., (Copyright © 2018 Elsevier B.V. All rights reserved.)

Background: Intensity modulated radiotherapy (IMRT) is increasingly being used to treat gynaecological malignancies in the postoperative setting. The purpose of this study was to evaluate the use of image-guided radiotherapy (IGRT) using cone-beam computed tomography (CBCT) with fiducial markers for daily localization., Material and Methods: A single institution study was performed of consecutive cervical or endometrial cancer patients receiving adjuvant external beam radiotherapy (n = 15). Patients were set up at treatment using daily CBCT and alignment of implanted fiducial markers. Image registration was retrospectively completed based on soft tissue matching and the resulting couch shifts from each IGRT method were compared (n = 122)., Results: The median shift between IGRT methods was 2 mm, 1 mm and 1 mm in the anterior-posterior (A-P), superior-inferior (S-I), and lateral directions, respectively. The largest deviations were observed in the A-P direction; however, more than 90% were within 5 mm and 63.9% were within 2.5 mm., Conclusions: IGRT based on soft tissue match provides a noninvasive convenient method for daily localization and is accurate within treatment uncertainty for the majority of cases.

Purpose: When brachytherapy doses are reported or added, biologically effective dose (BED) minimum dose covering 90% of the volume (D90) is used as if dose is delivered uniformly to the target. Unlike BED(D90), equivalent uniform BED (EUBED) and generalized biologically equivalent uniform dose (gBEUD) are quantities that integrate dose inhomogeneity. Here we compared BED(D90) and equivalent uniform BED (EUBED)/gBEUD in 3 settings: (1) 2 sites using tandem and ovoid (T&O) but different styles of implants; (2) 2 sites using different devices-T&O and tandem and ring (T&R)-and different styles; and (3) the same site using T&O and T&R with the same style., Methods and Materials: EUBED and gBEUD were calculated for 260 fractions from 3 institutions using BED(α/β = 10 Gy). EUBED uses an extra parameter α with smaller values associated with radioresistant tumors. Similarly, gBEUD uses a, which places variable emphasis on hot/cold spots. Distributions were compared using the Kolmogorov-Smirnoff test at 5% significance., Results: For the 2 sites using T&O, the distribution of EUBED-BED(D90) was not different for values of α = 0.5 to 0.3 Gy -1 but was statistically different for values of α = 0.15 to 0.05 Gy -1 (P=.01, .002). The mean percentage differences between EUBED and BED(D90) ranged from 20% to 100% for α = 0.5 Gy -1 to 0.05 Gy -1 . Using gBEUD-BED(D90), the P values indicate the distributions to be similar for a = -10 but to be significantly different for other values of a (-5, -1, 1). Between sites and at the same site using T&O versus T&R, the distributions were statistically different with EUBED/gBEUD irrespective of parameter values at which these quantities were computed. These differences indicate that EUBED/gBEUD capture differences between the techniques and applicators that are not detected by the BED(D90)., Conclusions: BED(D90) is unable to distinguish between plans created by different devices or optimized differently. EUBED/gBEUD distinguish between dose distributions created by different devices and styles of implant and planning. This discrepancy is particularly important with the increased use of magnetic resonance imaging and hybrid devices, whereby one has the ability to create dose distributions that are significant departures from the classic pear., (Copyright © 2017 Elsevier Inc. All rights reserved.)

Background and Purpose: This study compares dosimetric parameters of planning target volume (PTV) coverage and organs at risk (OAR) sparing when postoperative radiotherapy for gynecologic cancers is delivered using volumetric modulated arc therapy (VMAT) versus a four-field (4FLD) box technique., Material and Methods: From July to December 2012, women requiring postoperative radiation for gynecologic cancers were treated with a standardized VMAT protocol. Two sets of optimized 4FLD plans were retrospectively generated: one based on standard anatomical borders (4FLD) and one based on the clinical target volume (CTV) created for VMAT with a 2 cm expansion guiding field border placement (4FLD+2). Ninety-five percent isodose curves were generated to evaluate PTV coverage., Results: VMAT significantly improved dose conformity compared with 4FLD and 4FLD+2 plans (p < 0.001) and provided additional coverage of the PTV posteriorly and superiorly, corresponding to coverage of the presacral and proximal iliac vessels. There was a significant reduction in dose to all OARs with VMAT, including a 58% reduction in the volume of the small bowel receiving more than 45 Gy (p=0.005)., Conclusions: Despite treating a larger volume, radiotherapy using a 4FLD technique is less homogenous and provides inferior coverage of the PTV compared with VMAT. With meticulous treatment planning and delivery, VMAT effectively encompasses the PTV and minimizes dose to OARs.

Stereotactic body radiation therapy (SBRT) has quickly become a preferred treatment option for early-stage lung cancer patients who are ineligible for surgery. This technique uses tightly conformed megavoltage (MV) x-ray beams to irradiate a tumour with ablative doses in only a few treatment fractions. Small high energy x-ray fields can cause lateral electron disequilibrium (LED) to occur within low density media, which can reduce tumour dose. These dose effects may be challenging to predict using analytic dose calculation algorithms, especially at higher beam energies. As a result, previous authors have suggested using low energy photons (<10 MV) and larger fields (>5 × 5 cm(2)) for lung cancer patients to avoid the negative dosimetric effects of LED. In this work, we propose a new form of SBRT, described as LED-optimized SBRT (LED-SBRT), which utilizes radiotherapy (RT) parameters designed to cause LED to advantage. It will be shown that LED-SBRT creates enhanced dose gradients at the tumour/lung interface, which can be used to manipulate tumour dose, and/or normal lung dose. To demonstrate the potential benefits of LED-SBRT, the DOSXYZnrc (National Research Council of Canada, Ottawa, ON) Monte Carlo (MC) software was used to calculate dose within a cylindrical phantom and a typical lung patient. 6 MV or 18 MV x-ray fields were focused onto a small tumour volume (diameter ∼1 cm). For the phantom, square fields of 1 × 1 cm(2), 3 × 3 cm(2), or 5 × 5 cm(2) were applied. However, in the patient, 3 × 1 cm(2), 3 × 2 cm(2), 3 × 2.5 cm(2), or 3 × 3 cm(2) field sizes were used in simulations to assure target coverage in the superior-inferior direction. To mimic a 180° SBRT arc in the (symmetric) phantom, a single beam profile was calculated, rotated, and beams were summed at 1° segments to accumulate an arc dose distribution. For the patient, a 360° arc was modelled with 36 equally weighted (and spaced) fields focused on the tumour centre. A planning target volume (PTV) was generated by considering the extent of tumour motion over the patient's breathing cycle and set-up uncertainties. All patient dose results were normalized such that at least 95% of the PTV received at least 54 Gy (i.e. D95 = 54 Gy). Further, we introduce 'LED maps' as a novel clinical tool to compare the magnitude of LED resulting from the various SBRT arc plans. Results from the phantom simulation suggest that the best lung sparing occurred for RT parameters that cause severe LED. For equal tumour dose coverage, normal lung dose (2 cm outside the target region) was reduced from 92% to 23%, comparing results between the 18 MV (5 × 5 cm(2)) and 18 MV (1 × 1 cm(2)) arc simulations. In addition to reduced lung dose for the 18 MV (1 × 1 cm(2)) arc, maximal tumour dose increased beyond 125%. Thus, LED can create steep dose gradients to spare normal lung, while increasing tumour dose levels (if desired). In the patient simulation, a LED-optimized arc plan was designed using either 18 MV (3 × 1 cm(2)) or 6 MV (3 × 3cm(2)) beams. Both plans met the D95 dose coverage requirement for the target. However, the LED-optimized plan increased the maximum, mean, and minimum dose within the PTV by as much as 80 Gy, 11 Gy, and 3 Gy, respectively. Despite increased tumour dose levels, the 18 MV (3 × 1 cm(2)) arc plan improved or maintained the V20, V5, and mean lung dose metrics compared to the 6 MV (3 × 3 cm(2)) simulation. We conclude that LED-SBRT has the potential to increase dose gradients, and dose levels within a small lung tumour. The magnitude of tumour dose increase or lung sparing can be optimized through manipulation of RT parameters (e.g. beam energy and field size).

Cone-beam computed tomography (CBCT) has rapidly become a clinically useful imaging modality for image-guided radiation therapy. Unfortunately, CBCT images of the thorax are susceptible to artefacts due to scattered photons, beam hardening, lag in data acquisition, and respiratory motion during a slow scan. These limitations cause dose errors when CBCT image data are used directly in dose computations for on-line, dose adaptive radiation therapy (DART). The purpose of this work is to assess the magnitude of errors in CBCT numbers (HU), and determine the resultant effects on derived tissue density and computed dose accuracy for stereotactic body radiation therapy (SBRT) of lung cancer. Planning CT (PCT) images of three lung patients were acquired using a Philips multi-slice helical CT simulator, while CBCT images were obtained with a Varian On-Board Imaging system. To account for erroneous CBCT data, three practical correction techniques were tested: (1) conversion of CBCT numbers to electron density using phantoms, (2) replacement of individual CBCT pixel values with bulk CT numbers, averaged from PCT images for tissue regions, and (3) limited replacement of CBCT lung pixels values (LCT) likely to produce artificial lateral electron disequilibrium. For each corrected CBCT data set, lung SBRT dose distributions were computed for a 6 MV volume modulated arc therapy (VMAT) technique within the Philips Pinnacle treatment planning system. The reference prescription dose was set such that 95% of the planning target volume (PTV) received at least 54 Gy (i.e. D95). Further, we used the relative depth dose factor as an a priori index to predict the effects of incorrect low tissue density on computed lung dose in regions of severe electron disequilibrium. CT number profiles from co-registered CBCT and PCT patient lung images revealed many reduced lung pixel values in CBCT data, with some pixels corresponding to vacuum (-1000 HU). Similarly, CBCT data in a plastic lung phantom were reduced by 200 HU compared with known CT number values. For the three patients, dose results using the CBCT number data registered with PCT showed a prescription dose reduction ranging from 4 to 13% (D95 = 47 Gy). Therefore, accurate determination of lung density, especially for very low lung density (<0.2 g cm(-3)) is essential, but difficult to achieve using the CBCT data. Applying corrective techniques (1) and (2) to CBCT patient data produced unacceptable dose differences. For one typical VMAT SBRT patient, the D95 for the corrected CBCT and BCT image-based plans differed by -4% (D95 = 52 Gy) and 9% (D95 = 59 Gy) compared to the co-registered PCT image-based plan. However, corrective technique (3) produced negligible dose differences comparing LCT and PCT image-based plans. With regard to implementing on-line DART, dose errors must be minimized because they affect re-optimization decisions, and prevent accurate accumulation of the dose distribution.

Modern radiation therapy techniques such as intensity-modulated radiation therapy (IMRT) and stereotactic body radiation therapy (SBRT) use tightly conformed megavoltage x-ray fields to irradiate a tumour within lung tissue. For these conditions, lateral electron disequilibrium (LED) may occur, which systematically perturbs the dose distribution within tumour and nearby lung tissues. The goal of this work is to determine the combination of beam and lung density parameters that cause significant LED within and near the tumour. The Monte Carlo code DOSXYZnrc (National Research Council of Canada, Ottawa, ON) was used to simulate four 20 × 20 × 25 cm(3) water-lung-water slab phantoms, which contained lung tissue only, or one of three different centrally located small tumours (sizes: 1 × 1 × 1, 3 × 3 × 3, 5 × 5 × 5 cm(3)). Dose calculations were performed using combinations of six beam energies (Co-60 up to 18 MV), five field sizes (1 × 1 cm(2) up to 15 × 15 cm(2)), and 12 lung densities (0.001 g cm(-3) up to 1 g cm(-3)) for a total of 1440 simulations. We developed the relative depth-dose factor (RDDF), which can be used to characterize the extent of LED (RDDF <1.0). For RDDF <0.7 severe LED occurred, and both lung and tumour dose were drastically reduced. For example, a 6 MV (3 × 3 cm(2)) field was used to irradiate a 1 cm(3) tumour embedded in lung with ultra-low density of 0.001 g cm(-3) (RDDF = 0.2). Dose in up-stream lung and tumour centre were reduced by as much as 80% with respect to the water density calculation. These reductions were worse for smaller tumours irradiated with high energy beams, small field sizes, and low lung density. In conclusion, SBRT trials based on dose calculations in homogeneous tissue are misleading as they do not reflect the actual dosimetric effects due to LED. Future clinical trials should only use dose calculation engines that can account for electron scatter, with special attention given to patients with low lung density (i.e. emphysema). In cases where tissue inhomogeneity corrections are applied, the nature of the correction used may be inadequate in predicting the correct level of LED. In either case, the dose to the tumour is not the prescribed dose and clinical response data are uncertain. The new information from this study can be used by radiation oncologists who wish to perform advanced radiation therapy techniques while avoiding the deleterious predictable dosimetric effects of LED., (© 2012 Institute of Physics and Engineering in Medicine)