AI-Rad Companion Organs RT

AI-Rad Companion Organs RT is a deep-learning driven medical imaging software designed to automate the contouring of pre-defined structures from CT and MR images for external beam radiation therapy treatment planning. It utilizes advanced deep-learning algorithms for post-processing and has undergone extensive non-clinical validation to demonstrate its safety and effectiveness as a tool requiring professional oversight for clinical use.

Siemens Healthcare GmbH has obtained marketing authorization for its medical imaging software, AI-Rad Companion Organs RT, after demonstrating substantial equivalence to an existing device. This post-processing software leverages deep-learning algorithms to automatically delineate pre-defined anatomical structures from DICOM CT and MR images. Its primary role is to serve as an input for various clinical workflows, most notably external beam radiation therapy treatment planning, with the critical caveat that all generated contours necessitate review, editing, and final acceptance by trained medical professionals. The software is explicitly not designed for the automated detection or contouring of pathological lesions. The latest iteration of AI-Rad Companion Organs RT incorporates an improved CT contouring algorithm, which now supports an additional 37 organs and structures, along with multi-guideline support and user interface enhancements. Comprehensive non-clinical evaluations, including rigorous software validation and bench testing, were performed to ascertain the device's performance and confirm its equivalence to its predecessor. These tests conclusively established that the modifications do not alter the product's intended use and that the device maintains comparable safety and effectiveness, while also adhering to relevant industry benchmarks for quality, safety, and cybersecurity.

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. Given the reliance on professional review, what specific training and validation steps are recommended for clinicians using this software in a real-world setting?; How does the enhanced CT contouring algorithm's performance compare to manual contouring or previous versions across the newly supported 37 organs/structures, particularly for inter-observer variability?; What are the specific deep-learning architectures and training data sets used to ensure the accuracy and robustness of the automated contouring, especially across diverse patient anatomies and pathologies?; Beyond external beam radiation therapy, what other clinical applications, if any, could potentially benefit from or be impacted by the automated contouring capabilities of this software?; What post-market surveillance or real-world data collection mechanisms are in place or planned to continuously assess the safety and efficacy of the software, particularly concerning the accuracy of contours and potential for clinical errors?