Brainomix 360 e-ASPECTS

Brainomix 360 e-ASPECTS is computer-aided diagnosis software that analyzes non-contrast CT brain images to automatically calculate an ASPECTS (Alberta Stroke Program Early CT) score, assisting clinicians in evaluating ischemic brain tissue injury within 24 hours of last known well. The software registers images, segments ASPECTS regions using an atlas, extracts voxel-level imaging features, and synthesizes them into a single automated ASPECTS score that clinicians can review and edit.

The proposed device uses machine learning/AI (Random Forest) algorithms to process NCCT images on standard off-the-shelf servers with DICOM compliance. Key differences from the predicate include: (1) addition of a voxel volume calculation showing the volume of voxels contributing to the ASPECTS score displayed as a heatmap; (2) removal of the gating condition requiring confirmation of MCA/ICA occlusion prior to scoring; and (3) extension of the clinical time window from 6 hours to 24 hours since last known well.

ISO 14971:2019 (risk management), FDA Cybersecurity Guidance, and IEC 81001-5-1 (cybersecurity). The device was validated using digital phantom testing (110 synthetic datasets) and standalone performance testing on 137 clinical non-contrast CT scans from three U.S. institutions with ground truth from three board-certified neuroradiologists.

The proposed device has substantially similar intended use and technological characteristics to the predicate (same software class, ASPECTS scoring methodology, CT image processing, and ML/AI approach). The three changes—voxel volume calculation, removal of gating condition, and time window extension—do not raise new safety or effectiveness questions because: (1) the heatmap is a visualization aid with no clinical decision impact, validated through digital phantom testing; (2) standalone performance testing on 137 cases with and without LVO confirmed the device improves ASPECTS accuracy across suspected stroke populations regardless of occlusion status; and (3) the clinical study included late time-window cases up to 24 hours, demonstrating generalized performance across the extended timeframe.