Liu, Wei

Abstract Background Accurate and efficient dose calculation is essential for online adaptive planning in proton therapy. Deep learning (DL) has shown promising dose prediction results for pencil beam scanning proton therapy (PBSPT) in recent years, but existing DL‐based dose prediction methods still suffer from limited generalizability and an inability to effectively handle outlier clinical cases. This may lead to inaccurate dose delivery to targets or excessive irradiation to organs at risk (OARs), thereby compromising the safety and efficacy of online adaptive proton therapy. Purpose To design a physics‐aware and generalizable AI‐based PBSPT dose prediction method that incorporates underlying physics to enhance generalizability, particularly in handling outlier clinical cases. Methods This study analyzed PBSPT plans of 103 prostate (93 for training and 10 for testing) and 78 lung cancer patients (68 for training and 10 for testing) from our institution, with each case comprising CT images and structure sets. Using the doses generated by our Monte Carlo‐based dose engine as the reference standard, we compared three methods: the region of interest (ROI)‐based method, the beam mask and sliding window method, and the proposed noisy probing dose method, which rapidly generates a low‐statistics dose via uniformly weighted spots on an expanded spot‐placement target volume without optimization. To evaluate the generalizability of these methods to rare treatment planning scenarios, 12 cases with uncommon beam angles or prescription doses were used to assess their performance, which was evaluated using dose‐volume histogram (DVH) indices, 3D Gamma passing rates (3%/2 mm/10%), and Dice coefficients for dose agreement, while prediction times were measured to gauge model efficiency. Results The proposed noisy probing dose method consistently outperformed the ROI‐based and beam mask baselines across all evaluated metrics, with more accurate dose agreement and superior generalizability. For DVH indices, the noisy probing dose method achieved the smallest deviation in clinical target volume (CTV) dose coverage: in prostate cancer, CTV D98 deviation was reduced by 45% (from 0.53 ± 0.22 Gy [RBE] for ROI‐based) and 29% (from 0.41 ± 0.28 Gy [RBE] for beam mask) to 0.29 ± 0.06 Gy [RBE]; in lung cancer, similar improvements were observed, with CTV D98 deviations reduced to 0.34 ± 0.12 Gy [RBE]. The 3D Gamma passing rates improved to 99.65% ± 1.15% for prostate targets and 97.04% ± 1.17% for lung targets. The dice coefficients of the 90% iso‐dose lines were also the highest with the noisy probing dose method (prostate: 0.983 ± 0.005; lung: 0.967 ± 0.01). For the 12 outlier cases, the noisy probing dose method maintained superior generalizability, yielding higher 3D Gamma passing rates (prostate targets: 96.79% ± 0.83%, OARs: 94.29% ± 1.01%; lung targets: 93.38% ± 1.34%, OARs: 93.95% ± 1.32%), demonstrating robust generalizability to rare clinical scenarios. The dose predictions for all testing cases were completed within 0.3 seconds. Conclusions A novel noisy probing dose method was proposed for PBSPT dose prediction in prostate and lung cancer patients. By embedding more proton‐specific physics, this method demonstrated an improvement in the generalizability of dose prediction.

Abstract Background In the current clinical workflow of radiation oncology, therapists manually summarize the physician‐issued Computed Tomography (CT) simulation orders to prepare patients' CT simulations. This process increases the workload, introduces variability in documentation quality, and is prone to human errors. Purpose This study aims to address these challenges using a large language model (LLM) to automate the generation of summaries from the CT simulation orders and evaluate its performance. Methods A total of 607 CT simulation orders were collected from the Aria database at our institution. A locally hosted LLaMa 3.1 405B model, accessed via the Application Programming Interface (API) service, was used to extract keywords from the CT simulation orders and generate summaries. The downloaded CT simulation orders were categorized into seven groups based on treatment modalities and disease sites. For each group, a customized instruction prompt was developed collaboratively with therapists to guide the LLaMa 3.1 405B model in generating summaries. The ground truth for the corresponding summaries was manually derived by carefully reviewing each CT simulation order and subsequently verified by therapists. The accuracy of the LLM‐generated summaries was evaluated by therapists using the ground truth. Results Over 98% of the LLM‐generated summaries aligned with the ground truth in terms of accuracy. Our evaluations showed an improved consistency in format and enhanced readability from the LLM‐generated summaries compared to the corresponding therapist‐generated summaries. This automated approach demonstrated consistent performance across all groups, regardless of treatment modality or disease site. Conclusions This study demonstrated the high precision and consistency of the LLaMa 3.1 405B model in extracting keywords and summarizing CT simulation orders, suggesting that LLMs have great potential to assist in this task, reduce the workload of CT simulation therapists and improve radiation oncology workflow efficiency.

Background/Objectives: Proton therapy delivers highly conformal doses to the target area without producing an exit dose, minimizing cumulative doses to healthy liver tissue. This study aims to evaluate current practices, challenges, and variations in the implementation of proton stereotactic body radiation therapy (SBRT) and hypofractionated therapy for liver malignancies, with the goal of providing a technical assessment to promote broader adoption and support future clinical trials. Methods and Materials: An extensive survey was conducted by NRG Oncology across North American proton treatment centers to assess the current practices of proton liver SBRT and hypofractionated therapy. The survey focused on key aspects, including patient selection, prescription and normal tissue constraints, simulation and motion management, treatment planning, quality assurance (QA), treatment delivery, and the use of image-guided radiation therapy (IGRT). Results: This survey captures the current practice patterns and status of proton SBRT and hypofractionated therapy in liver cancer treatment. Proton therapy is increasingly preferred for treating inoperable liver malignancies due to its ability to minimize healthy tissue exposure. However, the precision required for proton therapy presents challenges, particularly in managing uncertainties and target motion during high-dose fractions and short treatment courses. Survey findings revealed significant variability in clinical practices across centers, highlighting differences in motion management, dose fractionation schedules, and QA protocols. Conclusion: Proton SBRT and hypofractionated therapy offer significant potential for treating liver malignancies. A comprehensive approach involving precise patient selection, treatment planning, and QA is essential for ensuring safety and effectiveness. This survey provides valuable insights into current practices and challenges, offering a foundation for technical recommendations to optimize the use of proton therapy and guide future clinical trials.

Abstract Background The radiation oncology clinical practice involves many steps relying on the dynamic interplay of abundant text data. Large language models have displayed remarkable capabilities in processing complex text information. But their direct applications in specific fields like radiation oncology remain underexplored. Purpose This study aims to investigate whether fine‐tuning LLMs with domain knowledge can improve the performance on Task (1) treatment regimen generation, Task (2) treatment modality selection (photon, proton, electron, or brachytherapy), and Task (3) ICD‐10 code prediction in radiation oncology. Methods Data for 15 724 patient cases were extracted. Cases where patients had a single diagnostic record, and a clearly identifiable primary treatment plan were selected for preprocessing and manual annotation to have 7903 cases of the patient diagnosis, treatment plan, treatment modality, and ICD‐10 code. Each case was used to construct a pair consisting of patient diagnostics details and an answer (treatment regimen, treatment modality, or ICD‐10 code, respectively) for the supervised fine‐tuning of these three tasks. Open source LLaMA2‐7B and Mistral‐7B models were utilized for the fine‐tuning with the Low‐Rank Approximations method. Accuracy and ROUGE‐1 score were reported for the fine‐tuned models and original models. Clinical evaluation was performed on Task (1) by radiation oncologists, while precision, recall, and F‐1 score were evaluated for Task (2) and (3). One‐sided Wilcoxon signed‐rank tests were used to statistically analyze the results. Results Fine‐tuned LLMs outperformed original LLMs across all tasks with p value ≤ 0.001. Clinical evaluation demonstrated that over 60% of the fine‐tuned LLMs‐generated treatment regimens were clinically acceptable. Precision, recall, and F1‐score showed improved performance of fine‐tuned LLMs. Conclusion Fine‐tuned LLMs demonstrated statistically significant improvements over original LLMs upon three clinically important tasks in radiation oncology. This study explored the feasibility of applying fine‐tuned LLMs in radiation oncology, inspiring further development of utilizing LLMs to assist with radiation oncology tasks.

Purpose Automating quality assurance (QA) for contours generated by automatic algorithms is critical in radiotherapy treatment planning. Manual QA is tedious, time-consuming, and prone to subjective experiences. Automatic segmentation reduces physician workload and improves consistency. However, an effective QA process for these automatic contours remains an unmet need in clinical practice. Materials and Methods The patient data used in this study was derived from the AAPM Thoracic Auto-Segmentation Challenge dataset, including left and right lungs, heart, esophagus, and spinal cord. Two groups of organ-at-risk (OAR) were generated. A ResNet-152 network was used as a feature extractor, and a one-class support vector machine (OC-SVM) was employed to classify contours as ‘high’ or ‘low’ quality. To evaluate the generalizability, we generated low-quality contours using translation and resizing techniques and assessed correlations between detection limits and metrics such as volume, Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), and mean surface distance (MSD). Results The proposed OC-SVM model outperformed binary classifiers n metrics such as balanced accuracy and area under the receiver operating characteristic curve (AUC) . It demonstrated superior performance in detecting various types of contour errors while maintaining high interpretability. Strong correlations were observed between detection limits and contour metrics. Conclusion Our proposed model integrates an attention mechanism with a one-class classification framework to automate QA for OAR delineations. This approach effectively detects diverse types of contour errors with high accuracy, significantly reducing the burden on physicians during radiotherapy planning.

Background/Objectives: Stereotactic body radiation therapy (SBRT) for skull base reirradiation is particularly challenging, as patients have already received substantial radiation doses to the region, and nearby normal organs may have approached their tolerance limit from prior treatments. In this study, we reviewed the characteristics and capabilities of four advanced external beam radiation delivery systems and four modern treatment planning systems and evaluated the treatment plan quality of each technique using skull base reirradiation patient cases. Methods: SBRT plans were generated for sixteen skull base reirradiation patients using four modalities: the GK plan for the Elekta Leksell Gamma Knife Perfexion/ICON, the CyberKnife (CK) plan for the Accuray CyberKnife, the intensity-modulated proton therapy (IMPT) plan for the Hitachi ProBeat-FR proton therapy machine, and the volumetric-modulated arc therapy (VMAT) plan for the Varian TrueBeam STx. These plans were evaluated and compared using two novel gradient indices in addition to traditional dosimetry metrics for targets and organs at risk (OARs). The steepest border gradient quantified the percent prescription dose fall-off per millimeter at the boundary between the target and adjacent critical structures. This gradient index highlighted the system’s ability to spare nearby critical OARs. The volume gradient assessed the extent of dose spread outside the target toward the patient’s body. Results: All plans achieved comparable target coverage and conformity, while IMPT and VMAT demonstrated significantly better uniformity. The GK plans exhibited the highest border gradient, up to 20.9%/mm, followed by small-spot-size IMPT plans and CK plans. Additionally, IMPT plans showed the benefit of reduced dose spread in low-dose regions and the lowest maximum and mean doses to the brainstem and carotid artery. Conclusions: The advanced external beam radiotherapy modalities evaluated in this study are well-suited for SBRT in skull base reirradiation, which demands precise targeting of tumors with highly conformal doses and steep dose gradients to protect nearby normal structures.

Purpose: This study compares the performance of small, fine-tuned LLMs with the large original models on two radiation oncology clinical tasks: treatment modality selection and ICD-10 code prediction. Methods: We used data of 403 treatment modality selection cases and 377 ICD-10 code prediction cases. The models used were fine-tuned LLaMA 2 7B and Mistral 7B, as well as the larger LLaMA 3.1 70B and 405B models for comparison. We evaluated the accuracy of all models on these two tasks. Results: Both fine-tuned LLaMA 2 7B and Mistral 7B outperformed LLaMA 3.1 70B and 405B in treatment modality selection task. The fine-tuned LLaMA 2 7B and Mistral 7B models achieved accuracy comparable to the larger models in ICD-10 code prediction. Conclusions: Small models, fine-tuned with radiation oncology data, can achieve performance comparable to-or even surpass-that of larger models.