TAIMedImg DeepMets

TAIMedImg DeepMets is a software application that uses deep learning to automatically segment and contour brain metastases on T1-weighted contrast-enhanced MRI images to accelerate radiation therapy treatment planning. It is intended as an adjunctive tool for trained medical professionals to review and modify, and is limited to adult patients with known brain metastases ≥10 mm diameter, excluding patients with prior craniotomy or other brain tumors.

Both DeepMets and the predicate (VBrain) employ deep learning neural networks for tumor contouring on axial T1 contrast-enhanced brain MRI, operate on Linux, output DICOM-RT format, and include no manual editing or data visualization features. The primary difference is that DeepMets is narrowly focused on brain metastases only (≥10 mm diameter), whereas VBrain also addresses meningiomas and acoustic neuromas.

IEC 62304:2006/A1:2016 (Medical device software lifecycle processes) and ISO 14971:2019 (Risk management application to medical devices). Software verification and validation activities followed FDA's June 2023 guidance for Enhanced Documentation Level device software.

DeepMets and VBrain share the same regulatory classification (Class II, Product Code QKB), intended user population (trained medical professionals), anatomical site (brain), imaging modality (T1 contrast-enhanced MRI), intended use (adjunctive contouring for radiation therapy planning), and technological approach (deep learning segmentation with mandatory physician review). Although DeepMets is narrower in scope (brain metastases only versus three tumor types), this narrower scope does not introduce new risks or alter the fundamental function, design, or safety profile compared to the predicate.