AI Metrics

AI Metrics is a software-based Picture Archiving and Communications System designed to enhance clinical workflows by enabling trained medical professionals to view, analyze, and compare medical images (CT, MR) over time, specifically for assessing disease extent and therapeutic response. Utilizing DICOM data, it offers tools for manual and semi-automatic image annotation, lesion segmentation, standardized measurements (e.g., RECIST 1.1), and automated structured reporting, without providing direct diagnoses.

AI Metrics is a software-only medical device designed as a Picture Archiving and Communication System (PACS) for use by trained medical professionals in environments like radiology and oncology. Its primary purpose is to facilitate the viewing, manipulation, communication, storage, annotation, analysis, and comparison of medical images, such as CT and MR scans, from multiple modalities and/or different time points. The methodology involves deploying the software via internet download onto a Linux platform (native or virtualized), where it functions as an interactive workflow automation tool. It enables users to assess and document the extent of a disease or tumor response to therapy according to user-selected standards, like RECIST 1.1, and track changes in imaging findings over time, thereby aiding in the interpretation and evaluation of examinations. The underlying science involves processing DICOM compliant image data, enabling features such as image navigation, manual and semi-automatic lesion segmentation suggestions, automatic measurement of lesion axes, tabulation and summation of measurements, semi-automatic anatomical lesion labeling suggestions, and the calculation of quantitative and qualitative metrics based on annotation data. AI Metrics automatically generates structured reports that include this data, presented in graphs, tables, key images, and structured text. While it does not directly generate diagnoses, it provides crucial information to support clinicians, who retain ultimate diagnostic responsibility. The software has undergone extensive verification and validation testing according to ISO standards (62304, 14971, 13485), with all identified risks deemed acceptable and the software passing all critical tests, ensuring it meets design specifications and demonstrates acceptable performance. AI Metrics is not intended for use in mammography.

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 does the semi-automatic lesion segmentation algorithm perform across different tumor types and image qualities, and what is the typical level of user adjustment required?; What specific quantitative and qualitative metrics, beyond RECIST 1.1, can be calculated or customized within the AI Metrics platform for advanced oncological research or specific clinical trial protocols?; Considering the workflow automation aspect, what is the measured impact of AI Metrics on radiologist/oncologist reporting time and inter-observer variability when compared to traditional disjointed image comparison methods?; How does AI Metrics handle and integrate images from various scanner manufacturers and DICOM versions, and what are the specific limitations regarding CT and MR dataset compatibility?; What are the security protocols and data management strategies employed by AI Metrics, particularly for deployment via internet software download and handling of sensitive patient DICOM data on Linux platforms?