Contour ProtegeAI

Contour ProtégéAI is an advanced medical imaging software accessory that employs machine-learning algorithms, specifically neural networks, to automatically generate anatomical contours on CT and MR images. This technology assists medical professionals by streamlining quantitative analysis, supporting adaptive therapy, and facilitating radiation therapy treatment planning through efficient segmentation of normal structures, requiring subsequent clinical review and validation.

Contour ProtégéAI is a medical imaging software accessory designed to assist trained medical professionals by automatically processing digital medical images, specifically CT and MR modalities compliant with ACR/NEMA DICOM 3.0. Its core function is the automated creation of anatomical contours using machine-learning algorithms, intended for applications such as quantitative analysis, aiding adaptive therapy, transferring contours to radiation therapy treatment planning systems, and archiving for patient follow-up. The software segments normal structures across various CT anatomical locations and specific structures like the prostate, seminal vesicles, and urethra in T2-weighted MR images. It is not intended for detecting or contouring lesions, and all outputs require review and potential editing via appropriate image visualization software. The system utilizes neural network models trained for CT and MR modalities, covering regions such as Head and Neck, Prostate, Thorax, Abdomen, Liver, and Lungs. Performance is evaluated using Dice coefficients, with studies showing Contour ProtégéAI achieving equivalent or better results compared to the MIM Maestro atlas segmentation reference device, based on statistical analysis of these coefficients. This version of Contour ProtégéAI includes modifications to its intended use, new and updated CT and MR neural network models with additional contour capabilities, and the option for local installation on Windows, Mac, or Linux systems, in addition to cloud-based deployment via MIMcloud service.

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. What specific anatomical structures, beyond the explicitly named prostate, seminal vesicles, and urethra for MR, are covered by the 'variety of CT anatomical locations' and what is the typical Dice coefficient range achieved for these structures?; How robust are the machine-learning algorithms to variations in image acquisition protocols, scanner manufacturers, and diverse patient populations not extensively represented in the initial training data?; What is the average time saved in the contouring workflow when using Contour ProtégéAI compared to fully manual contouring or traditional atlas-based segmentation, and what is the typical extent of manual editing required post-automation?; Can the neural network models be fine-tuned or retrained by local institutions using their own curated datasets to potentially improve performance for specific local patient demographics or imaging protocols?; How does the performance of the updated neural network models (e.g., new Prostate MR model, new/modified CT models including Abdomen, Lungs and Liver CT) compare quantitatively to their predecessors or the MIM Maestro reference device for the newly added or modified contours?