DrAid™ for Liver Segmentation

DrAid™ for Liver Segmentation is a web-based, AI-powered software designed to semi-automatically segment the liver and measure lesions from multiphase CT images, enabling quantitative imaging for medical professionals. This Class II medical device utilizes an editable AI algorithm and manual tools to process DICOM data, demonstrating high accuracy and robustness across diverse patient and scanner conditions, supporting efficient visualization and reporting without direct diagnosis.

DrAid™ for Liver Segmentation, developed by VinBrain Joint Stock Company, has been deemed substantially equivalent to existing market devices, allowing its commercialization as a Class II medical image management and processing system. This regulatory clearance permits its intended use by trained professionals, such as physicians and technicians, for the visualization, evaluation, and reporting of the liver and physician-identified lesions. The software operates as a web-based platform, processing multiphase CT DICOM images from patients aged 22 and older. Its core functionality includes semi-automated quantitative imaging driven by an AI algorithm for liver segmentation, which remains editable by the physician to ensure accuracy. Additionally, it offers manual tools for the precise segmentation and measurement of liver and lesion volumes and diameters, supporting robust analysis without generating direct diagnoses. Extensive performance testing has validated the software's accuracy and robustness across a wide range of patient conditions and various CT scanner manufacturers.

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 metrics (e.g., Dice coefficient, mean surface distance) were used to evaluate the accuracy and robustness of the AI algorithm, and what were the reported performance ranges?; How does the physician's ability to edit the AI's segmentation output impact overall workflow efficiency and diagnostic confidence compared to purely manual methods?; Are there any known limitations or specific patient conditions (e.g., severe liver disease, presence of artifacts) where the DrAid4 algorithm's performance might be suboptimal?; What is the training data size and diversity (e.g., number of patients, disease states, CT scanner models) used to develop and validate the AI model?; Beyond volume and diameter, does the system offer capabilities for other quantitative analyses, such as liver fat quantification or lesion characterization based on enhancement patterns?