MR Diffusion Perfusion Mismatch V1.0

MR Diffusion Perfusion Mismatch V1.0 is an automated software tool designed to assist clinicians by processing MR diffusion and perfusion images to calculate parametric maps, segment volumes, and quantify mismatch regions. The device utilizes DICOM image processing, AI-driven segmentation, and 3D motion correction algorithms, functioning as a docker application on a technical platform to provide quantitative metrics for radiological assessment.

MR Diffusion Perfusion Mismatch V1.0 is an automatic image processing software tool designed for radiology applications. Its core function is to assist trained radiologists and surgeons by processing MR Diffusion-weighted imaging (DWI) and MR Perfusion-weighted imaging (PWI) data to compute parametric maps, extract volumes of interest based on numerical thresholds, and automatically calculate the mismatch between these volumes. The software operates as a docker application that integrates with a separate Medical Image Communications Device (MICD) platform, which manages DICOM image input, output, and visualization, as the tool itself lacks a direct interface. It is intended as an adjunctive tool to be used with other patient information and professional judgment, not for standalone diagnosis, and it does not alter the original medical images. The device employs a 3D rigid motion correction algorithm and utilizes AI algorithms for its Diffusion Brain Extraction Tool (BET) and Perfusion Background Segmentation. Comparative performance testing against its predicate (Olea Sphere V3.0) showed statistical equivalence for several parametric maps (ADC, CBF, CBV, MTT, tMIP), while TTP and Tmax maps, though not statistically equivalent due to acquisition sensitivities, were deemed qualitatively equivalent by a neuroradiologist. Segmented volumes (Volume_1, Volume_2) and various mismatch metrics (Ratio, Volume, Relative Mismatch) demonstrated acceptable biases and differences, falling within expected inter-software variability and considered visually or clinically acceptable by a neuroradiologist.

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 do the specific AI algorithms for Brain Extraction Tool and Perfusion Background Segmentation in MR Diffusion Perfusion Mismatch V1.0 handle variations in image quality or atypical anatomies compared to the predicate's methods?; Given the sensitivity of TTP and Tmax maps to acquisition parameters, what specific MRI acquisition protocols or preprocessing steps are recommended to optimize the reliability of these temporal maps generated by the device?; What is the quantifiable impact of the 3D rigid motion correction algorithm on the accuracy of derived parametric maps and mismatch volumes in clinical scenarios with varying degrees of patient motion?; Beyond the statistical and qualitative equivalence reported, are there specific clinical scenarios or patient subgroups where the differences in TTP, Tmax, or Volume_2 calculations between V1.0 and its predicate might lead to differing interpretations?; What are the validation data supporting the performance of the AI-based segmentation across different MR scanner vendors and field strengths, and how does the device ensure robustness in diverse clinical settings?