A Quantitative Method Using Head Noncontrast CT Scans to Detect Hyperacute Nonvisible Ischemic Changes in Patients...

Area of Science:

• Neuroradiology
• Medical Imaging Analysis
• Computational Pathology

Background:

• Noncontrast computed tomography (CT) has limited sensitivity for early acute ischemic stroke detection.
• Computer-aided diagnosis (CAD) tools can enhance the interpretation of medical images.
• Previous methods focused on visible infarcts, necessitating improved techniques for early, nonvisible ischemia.

Purpose of the Study:

• To evaluate and extend the Stroke Imaging Marker (SIM) algorithm for localizing nonvisible hyperacute ischemia.
• To assess the performance of a modified SIM using Hounsfield Unit (HU) distribution ratios and voxel counts.
• To improve early detection of ischemic stroke undetectable by conventional CT evaluation.

Main Methods:

• The study utilized 140 early and follow-up scans from 70 patients, with initial scans showing no visible infarcts.
• The modified SIM algorithm incorporated percentile differences in HU subranges (P-ratio) and voxel counts.
• Infarct localization was assessed by comparing detected regions with a gold standard from follow-up scans, measuring hemisphere detection rate (HDR) and localization sensitivity.

Main Results:

• The optimal algorithm performance was achieved using a P-ratio with seven percentile subranges (35th-75th percentile).
• The modified SIM demonstrated a 76% ischemic hemisphere detection rate (HDR).
• The spatial localization sensitivity for hyperacute ischemia was 54% (68% for correctly detected infarct sides).

Conclusions:

• The improved SIM is a specialized tool for detecting hyperacute, nonvisible brain infarcts on CT scans.
• This CAD approach has the potential to reduce "image-to-needle" time for patients requiring revascularization therapy.
• The enhanced SIM may significantly improve the diagnostic capabilities for early ischemic stroke detection.