Clot Analog Attenuation in Non-contrast CT Predicts Histology: an Experimental Study Using Machine Learning

Area of Science:

• Radiology and Imaging Science
• Biomedical Engineering
• Pathology and Laboratory Medicine

Background:

• The precise histological composition of thrombi is not well understood.
• Variability in non-contrast CT (NCCT) imaging of thrombi complicates interpretation.
• Identifying imaging biomarkers for clot composition is crucial for clinical applications.

Purpose of the Study:

• To determine optimal imaging variables from NCCT for predicting clot composition.
• To analyze the variability in thrombus imaging based on histological differences.

Main Methods:

• Utilized a CT-phantom with 80 clot analogs of diverse histologies.
• Analyzed X-ray attenuation (HU values) at 24 and 48 hours post-production.
• Employed supervised machine learning (decision trees) to correlate imaging with histological and chemical analyses (water, iron content).

Main Results:

• Red blood cells (RBCs) were the primary determinant of CT attenuation (p < 0.001).
• Maximum axial CT attenuation accurately discriminated up to four histological groups (p < 0.001).
• Similar RBC-rich thrombi showed imaging variability linked to iron and white blood cell content.

Conclusions:

• NCCT provides valuable insights into clot histological and chemical composition.
• RBCs and iron content are independent factors influencing clot attenuation on NCCT.
• Advanced analysis of NCCT can improve understanding of thrombus characteristics.