ART-PLAN

ART-Plan is a medical image management and processing software that utilizes AI-driven deep learning algorithms for automated contouring of anatomical structures and generation of pseudo-CT images from multi-modal scans (CT, MR, PET). Its purpose is to enhance efficiency and standardization in radiotherapy treatment planning for cancer patients by providing tools for image visualization, registration, segmentation, and data export to treatment planning systems, with performance validated through extensive non-clinical testing.

ART-Plan version 1.10.0 is an advanced software application designed for cancer patients undergoing radiation treatment planning. It assists medical professionals by providing tools for 3D multi-modal medical image visualization, processing, and management, supporting DICOM 3.0 compliant datasets from CT, PET-CT, CBCT, 4D-CT, and MR imaging. Key functionalities include AI-driven automatic contouring of organs at risk (OARs) and healthy lymph nodes on both CT and MR images, as well as semi-automatic and manual segmentation tools. The software facilitates the registration (rigid and deformable) of anatomical and functional images, allowing for fused image displays to aid in comparison and delineation. A significant feature is its capability to generate, visualize, evaluate, and modify pseudo-CT images from MRI data, particularly for brain and male pelvis anatomies. The system comprises two main modules, SmartFuse for image registration and Annotate for contouring. ART-Plan v1.10.0 introduces improvements such as enhanced automatic segmentation across more anatomies (including female pelvis on CT, and various regions on MR), image registration for 4D-CT and CBCT, and cloud-based deployment. Its AI models for segmentation and pseudo-CT generation were trained on extensive, pseudo-anonymized real-world datasets, with performance validated against established clinical acceptance criteria (e.g., Dice Similarity Coefficient, gamma passing rates, qualitative evaluations) and compared favorably to predicate devices. The training data demographics were shown to be representative, and strategies were employed to mitigate biases from imaging hardware or patient anatomical diversity. Non-clinical testing confirmed the device's usability, safety, and effectiveness for its intended use in radiotherapy planning.

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 the AI-based contouring for specific, challenging OARs (e.g., brachial plexus, individual cranial nerves) compare to expert manual delineation in terms of accuracy and time-saving, especially across different image qualities and pathologies?; What is the quantitative impact of using the generated pseudo-CT from MRI on dose calculation accuracy for different radiotherapy treatment modalities (e.g., VMAT, IMRT, SBRT) and across various anatomical sites beyond brain and male pelvis?; How robust is the deformable image registration algorithm when dealing with significant anatomical changes between imaging sessions, such as tumor shrinkage or post-surgical alterations, and what are the typical failure modes?; Given the training data dependencies (e.g., ViewRay 0.35T MR, Philips 1.5T T2 for specific models), what is the performance degradation, if any, when ART-Plan is used with data from MR scanners with different field strengths, vendors, or sequences not heavily represented in the training set?; What are the specific user validation steps recommended or enforced by the ART-Plan workflow before exporting AI-generated contours or pseudo-CT data to a treatment planning system, and how does the system flag uncertainties or areas requiring closer scrutiny?