Brainomix 360 e-ASPECTS

Brainomix 360 e-ASPECTS is an artificial intelligence-powered software that assists clinicians by automatically analyzing non-contrast CT scans to generate an ASPECTS score for patients with acute ischemic stroke due to MCA or ICA occlusion. Its methodology involves atlas-based segmentation, voxel-level analysis using machine learning, and heatmap visualization to quantify ischemic injury within the critical 6-hour window, aiming to improve diagnostic accuracy and consistency in stroke assessment.

Brainomix 360 e-ASPECTS is a computer-aided diagnosis (CADx) software designed to assist clinicians in evaluating brain tissue abnormalities from Non-Contrast CT (NCCT) scans in patients with known Middle Cerebral Artery (MCA) or Internal Carotid Artery (ICA) occlusions, specifically within 6 hours from the time last known well. The software employs an artificial intelligence algorithm, utilizing a random forest machine learning technique, to automatically register images, segment ASPECTS (Alberta Stroke Program Early CT) regions using an atlas, and analyze individual voxels. It then synthesizes these imaging features into a single, editable ASPECTS score, intended to quantify the extent of ischemic brain tissue injury. A key feature is the visualization of voxels contributing to or excluded from the score via a heatmap, providing transparency to the AI's assessment and differentiating acute ischemic changes from old infarcts. Performance evaluations demonstrated the software's utility and accuracy. Stand-alone testing on 256 NCCT scans yielded an area under the curve (AUC) of 83%, with 68% sensitivity and 97% specificity. A multiple reader multiple case (MRMC) study involving 10 clinicians and 54 scans showed that using Brainomix 360 e-ASPECTS resulted in a statistically significant improvement in reader AUC by 0.02 (from 0.81 to 0.83), primarily driven by increased sensitivity (from 66% to 70%). The software particularly benefited "lower performing" readers and reduced inter-reader variability in ASPECTS scoring. Digital phantom testing confirmed a strong correlation (r≥0.95) for its heatmap feature. The system is designed to improve consistency and timeliness in stroke workup, aiding in patient selection for therapeutic interventions.

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 algorithm differentiate between acute ischemic changes and pre-existing chronic infarcts, and what is the accuracy of this differentiation?; What is the performance variability of the Brainomix 360 e-ASPECTS software across different CT scanner manufacturers and models, beyond the validated Siemens Somatom Definition scanners?; How does the 'heatmap visualization' of voxel contribution specifically influence clinicians' decision-making and their confidence in the automated ASPECTS score, especially in borderline or complex cases?; Given the lower performance in M4, M6, and internal capsule regions observed in stand-alone testing, what specific algorithmic or training data enhancements are planned to improve accuracy in these anatomically challenging areas?; What is the impact of image quality variations (e.g., noise, motion artifacts not severe enough for exclusion) on the software's ability to accurately segment ASPECTS regions and generate a reliable score?