TAIMedImg DeepMets

TAIMedImg DeepMets is an AI-powered software utilizing deep learning neural networks to provide automated Gross Tumor Volume (GTV) contours of brain metastases (>=10mm) on T1-weighted contrast-enhanced brain MR images. Designed as an adjunctive tool, its primary purpose is to accelerate radiation therapy treatment planning workflows by assisting trained medical professionals, who ultimately review and finalize the generated contours.

On May 28, 2025, a determination was issued to Taiwan Medical Imaging Co., Ltd. for their device, TAIMedImg DeepMets, classifying it as a Class II medical image management and processing system. The device was found to be substantially equivalent to legally marketed predicate devices, thereby allowing its marketing subject to compliance with general controls provisions, including annual registration, device listing, good manufacturing practices, labeling, and prohibitions against misbranding and adulteration, as well as adherence to Quality System regulation and Unique Device Identification System rules. TAIMedImg DeepMets is a software application system utilizing artificial intelligence, specifically deep learning neural networks, to provide initial Gross Tumor Volume (GTV) contours of brain metastases on axial T1-weighted contrast-enhanced brain magnetic resonance images. Designed for lesions 10 mm or greater in diameter, its primary purpose is to accelerate radiation therapy treatment planning workflows. Functioning as an adjunctive tool, the software requires trained medical professionals to review and finalize the generated contours. Performance testing, including software verification and validation, confirmed that TAIMedImg DeepMets meets its safety and effectiveness requirements and performs as expected, demonstrating substantial equivalence to other AI-based tumor contouring software like VBrain.

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 are the specific performance metrics (e.g., Dice similarity coefficient, Hausdorff distance) achieved by TAIMedImg DeepMets compared to manual contouring or predicate devices?; How does the accuracy of TAIMedImg DeepMets vary for lesions with diameters close to the 10 mm threshold, and what are the implications for smaller metastases?; What specific datasets were used for training and validation of the deep learning model, and how diverse were these datasets in terms of patient demographics, MR scanner types, and tumor characteristics?; Are there any specific types of brain metastases (e.g., highly irregular shapes, perilesional edema) where the AI's performance is notably different or requires more significant manual adjustment?; What is the observed workflow efficiency improvement (e.g., time saved per case) in a clinical setting when using TAIMedImg DeepMets compared to entirely manual GTV contouring for radiation therapy planning?