Exploratory machine learning analysis to characterize angioscopic features associated with atherosclerosis-related...

Area of Science:

• Cardiovascular research
• Medical imaging
• Pathophysiology

Background:

• Aortic dissection (AD) is a critical vascular emergency.
• Non-obstructive general angioscopy (NOGA) allows direct visualization of aortic injuries.
• Spontaneously ruptured aortic plaques and injuries (SRAPIs) are potential biomarkers for AD.

Purpose of the Study:

• To evaluate the diagnostic potential of SRAPIs in AD.
• To explore the pathophysiological relevance of SRAPIs to AD.
• To develop a machine learning model for AD classification based on SRAPIs.

Main Methods:

• A single-center, observational pilot study involving 56 AD patients and 444 controls undergoing NOGA.
• Classification of 13 SRAPI types and analysis of their occurrence using machine learning.
• Development of a Random Forest classification model with LASSO regression for feature selection and SHAP/Permutation Importance for interpretation.

Main Results:

• Intramural blood (IB) emerged as the most consistent SRAPI associated with AD.
• Puff sign (P) and salmon-pink appearance (SP) also demonstrated significant importance.
• Fissure bleeding (FB) was frequent but had variable importance; network analysis showed distinct patterns between AD and control groups.

Conclusions:

• AD involves subintimal injury (IB, SP) and vascular wall adherence (P).
• FB may indicate a critical, non-linear pathological state.
• Findings require validation in larger, multicenter studies.