ART-Plan

ART-Plan is an AI-driven software designed to enhance radiation therapy treatment planning by providing tools for multi-modal image registration, automated segmentation, synthetic CT generation from MR and CBCT, dose computation, and adaptive therapy decision support. Its methodology relies on deep learning models trained on diverse clinical data and validated against rigorous performance criteria to ensure accuracy and clinical utility in complex radiotherapy workflows.

ART-Plan, developed by TheraPanacea, is an advanced software system designed for radiation therapy treatment planning for adult cancer patients. It assists trained clinicians, including radiotherapists, oncologists, physicists, and dosimetrists, by providing a suite of tools to enhance the radiotherapy workflow, such as multi-modal image visualization (CT, MR, PET-CT, 4D-CT, CBCT), rigid and deformable image registration, display of fused/non-fused images for delineation, manual/semi-automatic/AI-driven automatic contouring of organs and lymph nodes on CT/MR images, generation of synthetic CT (sCT) from MR/CBCT, dose calculations for external beam photon therapy, and assisted decision-making for CBCT-based off-line treatment adaptation. The system operates via SmartFuse, Annotate, and AdaptBox modules, utilizing deep learning algorithms trained on extensive, pseudo-anonymized, retrospective clinical data, carefully managed to ensure robustness and prevent overfitting. Training data demographics are representative of cancer patient populations. Performance validation for its modules involved rigorous testing against established criteria (e.g., Dice Similarity Coefficients, gamma passing rates, expert evaluations), demonstrating equivalent performance on US/non-US datasets and addressing potential biases, with specific limitations on age, sex, and image inputs disclosed to users.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of ART-Plan's AI-based auto-segmentation vary across different imaging vendors and acquisition protocols not extensively represented in the training data?; What is the clinical impact on treatment planning efficiency and plan quality when using synthetic CT generated by ART-Plan from MR versus CBCT, particularly for anatomies with significant inter-fractional motion?; For the off-line adaptation decision-making module (AdaptBox), what specific quantitative metrics are presented to the user, and how have these metrics been validated to correlate with improved patient outcomes or avoidance of unnecessary re-planning?; Given the limitations of some models (e.g., sex-specific, MR sequence dependent), what is the development roadmap for expanding the applicability of these advanced features to a broader patient population and range of imaging inputs?; How does ART-Plan's deformable image registration algorithm handle large anatomical changes or areas with poor image contrast, and what are the quality assurance measures recommended to verify registration accuracy in challenging cases before clinical use?