Search

Predicting typhoon intensity accurately across space and time is crucial for issuing timely disaster warnings and facilitating emergency response. This has vast potential for minimizing life losses and property damages as well as reducing economic and environmental impacts. Leveraging satellite imagery for scenario analysis is effective but also introduces additional challenges due to the complex relations among clouds and the highly dynamic context. Existing deep learning methods in this domain rely on convolutional neural networks (CNNs), which suffer from limited per-layer receptive fields. This limitation hinders their ability to capture long-range dependencies and global contextual knowledge during inference. In response, we introduce a novel approach, namely "Typhoon Intensity Transformer" (Tint), which leverages self-attention mechanisms with global receptive fields per layer. Tint adopts a sequence-to-sequence feature representation learning perspective. It begins by cutting a given satellite image into a sequence of patches and recursively employs self-attention operations to extract both local and global contextual relations between all patch pairs simultaneously, thereby enhancing per-patch feature representation learning. Extensive experiments on a publicly available typhoon benchmark validate the efficacy of Tint in comparison with both state-of-the-art deep learning and conventional meteorological methods. Our code is available at https://github.com/chen-huanxin/Tint., Comment: 8 pages, 2 figures, accepted by Tackling Climate Change with Machine Learning: workshop at NeurIPS 2023

Cloud analysis is a critical component of weather and climate science, impacting various sectors like disaster management. However, achieving fine-grained cloud analysis, such as cloud segmentation, in remote sensing remains challenging due to the inherent difficulties in obtaining accurate labels, leading to significant labeling errors in training data. Existing methods often assume the availability of reliable segmentation annotations, limiting their overall performance. To address this inherent limitation, we introduce an innovative model-agnostic Cloud Adaptive-Labeling (CAL) approach, which operates iteratively to enhance the quality of training data annotations and consequently improve the performance of the learned model. Our methodology commences by training a cloud segmentation model using the original annotations. Subsequently, it introduces a trainable pixel intensity threshold for adaptively labeling the cloud training images on the fly. The newly generated labels are then employed to fine-tune the model. Extensive experiments conducted on multiple standard cloud segmentation benchmarks demonstrate the effectiveness of our approach in significantly boosting the performance of existing segmentation models. Our CAL method establishes new state-of-the-art results when compared to a wide array of existing alternatives., Comment: Accepted at the TCCML Workshop at NeurIPS 2023

Purpose:The recently proposed IPO-IMPT framework allows simultaneous optimization of dose, dose rate, and LET for FLASH treatment planning. Finding solutions to IPO-IMPT is difficult due to computational intensiveness. Nevertheless, an inverse solution that simultaneously specifies the geometry of a sparse filter and weights of a proton intensity map is desirable for both clinical and preclinical applications. Such solutions can reduce effective biological dose to organs at risk in cancer patients as well as reduce the number of animal irradiations needed to derive extra biological dose models in preclinical studies. Methods:Unlike our initial forward heuristic, this inverse IPO-IMPT solution includes simultaneous optimization of sparse range compensation, sparse range modulation, and spot intensity. The daunting computational tasks vital to this endeavor were resolved iteratively with a distributed computing framework to enable Simultaneous Intensity and Energy Modulation and Compensation (SIEMAC). SIEMAC was demonstrated on a human lung cancer patient and a minipig. Results:SIEMAC improves maps of spot intensities and patient-field-specific sparse range compensators and range modulators. For the lung cancer patient, at our max nozzle current of 300 nA, dose rate coverage above 100 Gy/s increased from 57% to 96% in the lung and from 93% to 100% in the heart, and LET coverage above 4 keV/um dropped from 68% to 9% in the lung and from 26% to <1% in the heart. For a simple minipig plan, the FWHM of the dose, dose rate, and LET distributions decreased by 30%, 1.6%, and 57%, respectively, again with similar target dose coverage, thus reducing uncertainty in these quantities for preclinical studies. Conclusion:The inverse solution to IPO-IMPT demonstrated the capability to simultaneously modulate sub-spot proton energy and intensity distributions for clinical and preclinical studies.

Purpose: To develop a Monte Carlo (MC) therapeutic dose calculation toolkit of a recently released ring gantry linac in Geant4 (Version 10.7) for secondary dose validation of radiotherapy plan. Methods: For the Halcyon (Varian Medical Systems), the DSMLC was modeled and radiation transport in DSMLC and patient phantom was simulated using Geant4. Radiation source was sampled from a phase space file for linac head above the DSMLC. The phase space file was obtained using a cloud-based Monte Carlo (MC) simulator, VirtuaLinac (VL) provide by Varian. Dosimetric profiles for different square field widths (2x2, 4x4, 6x6, 8x8, 10x10, 20x20, and 28x28 cm2), i.e., percent depth dose (PDD) curves and lateral profiles are simulated and compared against the experimental profiles. IMRT (intensity modulated radiation therapy) plans in two anatomical sites (prostate and brain) were also calculated using the developed toolkit and compared against the TPS calculated dose (Acuros, Eclipse 15.6). 3D dose difference and 3D gamma analysis were used to evaluate the simulation accuracy compared against the TPS calculated dose. Results: The simulated lateral dose profiles and PDD curves in water phantom match well with the measured ones for all the simulated field sizes with relative difference +-2%. For the prostate and brain IMRT plans, the simulated dose showed a good agreement with the TPS calculated dose. The 3D gamma pass rate (3%/3mm) are 98.08% and 95.4% for the two prostate and brain plans, respectively. Conclusion: The developed full MC dose calculation toolkit for Halcyon performs well in dose calculations in water phantom and patient CT phantom. The developed toolkit shows promising possibility for future secondary dose calculation for IMRT and serve as clinical quality assurance (QA) tool for Halcyon.

Ultra-high dose rate (FLASH) proton planning with only transmission beams (TBs) has limitations in normal tissue sparing. The single-energy spread-out Bragg peaks (SESOBPs) of FLASH dose rate have been demonstrated feasible for proton FLASH planning. A hybrid inverse optimization method was developed to combine the TBs and SESOBPs (TB-SESOBP) for FLASH planning. The SESOBPs were generated from spreading out the BPs by pre-designed general bar ridge filters and placed at the central target by range shifters to obtain a uniform dose within the target. The SESOBPs and TBs were fully sampled field-by-field allowing automatic spot selection and weighting in the optimization process. The TB-SESOBP plans were validated in comparison with the TB only (TB-only) plans and the plans with the combination of TBs and BPs (TB-BP) regarding 3D dose and dose rate distributions for five lung cases. Comparing to the TB-only plans, the mean spinal cord D1.2cc drastically reduced 41%, the mean lung V7Gy and V7.4Gy moderately reduced by up to 17% and the target dose homogeneity slightly increased in the TB-SESOBP plans. Comparable dose homogeneity was achieved in both TB-SESOBP and TB-BP plans. Besides, prominent improvements were achieved in lung sparing for the cases of relatively large targets by the TB-SESOBP plans comparing to the TB-BP plans. The targets were fully covered with the FLASH dose rate in all the three plans. For the OARs, V40Gy/s = 100% was achieved by the TB-only plans while V40Gy/s > 85% was obtained by the other two plans. We have demonstrated that the hybrid TB-SESOBP planning was feasible to achieve FLASH dose rate for proton therapy. The hybrid TB-SESOBP planning has great potential in improving OAR sparing while maintaining high target dose homogeneity, and can be potentially implemented for adaptive radiotherapy.

We aim to investigate the feasibility of online monitoring of irradiation time (IRT) and scan time for FLASH radiotherapy using a pixelated semiconductor detector. Measurements of the time structure of FLASH irradiations were performed using fast, pixelated spectral detectors, AdvaPIX-TPX3 and Minipix-TPX3. The latter has a fraction of its sensor coated with a neutron sensitive material. With little or no dead time and an ability to resolve events that are closely spaced in time (tens of ns), both detectors can accurately determine IRTs as long as pile-ups are avoided. To avoid pile-ups, we placed the detectors beyond the Bragg peak or at a large scattering angle. We acquired prompt gamma rays and secondary neutrons and calculated IRTs based on timestamps of the first (beam-on) and the last (beam-off) charged species. We also measured scan times in x, y, and diagonal directions. We performed these measurements for a single spot, a small animal field, a patient field, and a ridge filter optimized field to demonstrate in vivo online monitoring of IRT. All measurements were compared to vendor log files. Differences between measurements and log files for a single spot, a small animal field, and a patient field were within 1%, 0.3% and 1%, respectively. In vivo monitoring of IRTs was accurate within 0.1% for AdvaPIX-TPX3 and within 6.1% for Minipix-TPX3. The scan times in x, y, and diagonal directions were 4.0, 3.4, and 4.0 ms, respectively. Overall, the AdvaPIX-TPX3 can measure FLASH IRTs within 1% accuracy, indicating that prompt gamma rays are a good surrogate for primary protons. The Minipix-TPX3 showed a higher discrepancy, suggesting a need for further investigation. The scan times (3.4 \pm 0.05 ms) in the 60-mm distance of y-direction were less than (4.0 \pm 0.06 ms) in the 24-mm distance of x-direction, confirming the much faster scanning speed of the Y magnets than that of X., Comment: 11 pages, 5 figures

Purpose: Patient-specific ridge filters can modulate proton energy to obtain a conformal dose. We describe a new framework for optimization of filter design and spot maps to meet the unique demands of FLASH radiotherapy. We demonstrate an Integrated Biological Optimization IMPT (IBO-IMPT) approach for optimization of dose, dose-averaged dose rate (DADR), and dose-averaged LET (LETd). Methods: We developed inverse planning software to design patient-specific ridge filters that spread the Bragg peak from a fixed-energy, 250 MeV beam to a proximal beam-specific planning target volume. The software optimizes patient-specific ridge filter and uses a Monte Carlo calculation engine, based on Geant4, to provide dose and LET influence matrices. Plan optimization, using matRAD, accommodates the IBO-IMPT objective function considering dose, dose rate, and LET simultaneously with minimum MU constraints. The framework enables design of both regularly spaced and sparse-optimized ridge filters, which allow faster delivery and selective LET optimization. Volume distributions and histograms for dose, DADR, and LETd are compared using evaluation structures specific to the heart and lung. Results: We used IBO-IMPT to design ridge filters for a central lung tumor patient. The IBO-IMPT framework selectively spared heart and lung by reducing LET and increasing dose rate, relative to IMPT planning. Sparse-optimized ridge filters were superior to regularly spaced ridge filters in dose rate. Together, these innovations substantially increased the DADR in the heart and lung while maintaining good target coverage. The volume that received a FLASH dose rate of higher 40 Gy/second increased by 31% for heart and 50% for lung. Conclusion: This proof-of-concept study demonstrates the feasibility of using an IBO-IMPT framework to accomplish proton FLASH SBPT, accounting for dose, DADR, and LETd simultaneously.

As an effective way to integrate the information contained in multiple medical images under different modalities, medical image synthesis and fusion have emerged in various clinical applications such as disease diagnosis and treatment planning. In this paper, an invertible and variable augmented network (iVAN) is proposed for medical image synthesis and fusion. In iVAN, the channel number of the network input and output is the same through variable augmentation technology, and data relevance is enhanced, which is conducive to the generation of characterization information. Meanwhile, the invertible network is used to achieve the bidirectional inference processes. Due to the invertible and variable augmentation schemes, iVAN can not only be applied to the mappings of multi-input to one-output and multi-input to multi-output, but also be applied to one-input to multi-output. Experimental results demonstrated that the proposed method can obtain competitive or superior performance in comparison to representative medical image synthesis and fusion methods., Comment: Page 10. arXiv admin note: text overlap with arXiv:2002.05000, arXiv:2103.15061 by other authors