Automated Cerebral Hemorrhage Detection Using RAPID

Area of Science:

• Neuroradiology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Intracranial hemorrhage (ICH) diagnosis relies on head computed tomography (CT) scans.
• Increased CT utilization can delay timely ICH identification.
• Automated tools are needed to expedite ICH diagnosis.

Purpose of the Study:

• To evaluate the accuracy of RAPID ICH, an AI application, for detecting ICH on head non-contrast CT (NCCT).
• To assess RAPID ICH's ability to quantify ICH volumes.
• To determine if RAPID ICH can facilitate expedited ICH diagnosis.

Main Methods:

• RAPID ICH analyzed NCCT scans for ICH detection.
• Expert neuroradiologists provided consensus for criterion standard comparison.
• RAPID ICH volumetric quantification was compared against manual segmentation by an expert.

Main Results:

• RAPID ICH demonstrated high sensitivity (0.956) and specificity (0.953) for ICH detection.
• The AI tool showed strong correlation (r=0.983) in volumetric quantification of hemorrhages.
• Median absolute error for volume quantification was 3 mL.

Conclusions:

• RAPID ICH is a highly accurate AI tool for detecting intracranial hemorrhage on NCCT.
• The application provides precise volumetric quantification of intraparenchymal and intraventricular hemorrhages.
• RAPID ICH shows potential for expediting ICH diagnosis and management.