Semi-automatic carotid intraplaque hemorrhage detection and quantification on Magnetization-Prepared Rapid...

Area of Science:

• Cardiovascular Imaging
• Medical Physics
• Radiology

Background:

• Intraplaque hemorrhage (IPH) is a key marker in atherosclerosis progression and predicts cardiovascular events.
• Accurate detection and quantification of IPH are crucial for risk stratification.

Purpose of the Study:

• To develop and validate a semi-automatic method for carotid IPH detection and quantification.
• To optimize intensity thresholds for IPH identification on MP-RAGE cardiovascular magnetic resonance (CMR) images.

Main Methods:

• Utilized 3D MP-RAGE CMR in 14 patients undergoing carotid endarterectomy.
• Compared semi-automatic IPH detection and area quantification against histology using optimized intensity thresholds normalized to SCM muscle, adjacent muscle, or local median values.
• Evaluated performance using sensitivity, specificity, and Pearson's correlation coefficients via leave-one-out cross-validation.

Main Results:

• Optimized thresholds varied based on normalization reference (SCM, adjacent muscle, local median).
• The semi-automatic method achieved high sensitivity (up to 80%) and specificity (up to 85%) for IPH detection.
• Strong agreement was found for IPH area quantification (Pearson's r up to 0.93) when using adjacent muscle or local median normalization.

Conclusions:

• A semi-automatic method with optimized thresholds demonstrates good performance for carotid IPH detection and quantification on MP-RAGE CMR.
• Normalization to adjacent muscle or local median values is recommended for reliable results.
• SCM normalization requires coil sensitivity correction, limiting its general applicability.