Rapid ASPECTS (v3)

Rapid ASPECTS (v3) is an artificial intelligence-driven, computer-aided diagnostic software that assists clinicians by automatically generating an ASPECTS score from Non-Contrast CT images of adult stroke patients to characterize early ischemic brain injury. The device employs a machine learning algorithm (random forest) to perform image registration, segment Regions of Interest, analyze morphological characteristics, and quantify the extent of disease, thereby offering a standardized adjunct to clinical image interpretation.

Rapid ASPECTS (v3) is a computer-aided diagnosis (CADx) software designed to assist clinicians in the assessment and characterization of brain tissue abnormalities using CT image data, specifically for adult patients presenting with symptoms of ischemic stroke, typically within 24 hours since last known well. The software automatically processes Non-Contrast Computed Tomography (NCCT) brain images to generate an Alberta Stroke Program Early CT (ASPECT) Score. It achieves this by registering images, segmenting and analyzing predefined ASPECTS Regions of Interest (ROIs), extracting image data based on morphological characteristics, and synthesizing these features using an artificial intelligence algorithm. The system presents the original and annotated images, along with the calculated score and highlighted ROIs based on an ASPECTS atlas, to aid clinicians in evaluating the extent of early ischemic brain tissue injury as an adjunct tool, not replacing clinical judgment. The core technology of Rapid ASPECTS (v3) is a machine learning implementation, specifically utilizing a random forest technique to analyze regional imaging features from NCCT data. Its processing pipeline includes orientation and spatial normalization of input images, delineation of ROIs, computation of numerical values characterizing underlying voxels, identification and highlighting of previous/old stroke areas alongside areas of early ischemic change, and labeling these regions to provide a summary ASPECT score. Performance data indicated that the software has a standalone agreement of 82.8% with expert reference and improves reader agreement with the reference standard when used as an assistive tool, increasing from 82% to 84% overall, and from 80.4% to 83.3% for non-expert readers. The device aims to provide a more consistent and timely standardized assessment of ASPECT scores, particularly benefiting non-expert readers.

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 'random forest machine learning technique' specifically contribute to the accuracy and robustness of the Rapid ASPECTS (v3) score compared to other potential AI models?; What is the quantifiable impact on scoring accuracy when CT scans exhibit patient motion, hemorrhagic transformation, or very thin/no ventricles, and are there internal quality checks for these conditions?; Beyond the reported increase in reader agreement (e.g., from 80.4% to 83.3% for non-experts), what is the demonstrated clinical significance of using Rapid ASPECTS (v3) in terms of changes in patient management or improved patient outcomes?; How does the Rapid ASPECTS (v3) algorithm technically differentiate between 'previous/old stroke areas' and 'areas of early ischemic change' within the NCCT data?; Given the expansion of the analysis window from <6 hours in the predicate device to typically <24 hours in Rapid ASPECTS (v3), what specific validation studies were conducted to ensure sustained diagnostic performance and sensitivity across this broader timeframe, particularly for detecting subtle ischemic changes at later stages?