Angioscopic Features for Aortic Dissection: Machine Learning Analysis of 500 Patients

A 2026 study by Satoru Takahashi presents data from 500 patients, including 56 with aortic dissection and 444 controls with coronary artery disease, undergoing non-obstructive general angioscopy. The dataset likely contains classifications of 13 types of spontaneously ruptured aortic plaques and injuries analyzed using Random Forest and LASSO regression. The study is a single-center, cross-sectional observational pilot study.

Use Cases

• Training classification models to predict aortic dissection based on angioscopic features like intramural blood and puff sign.
• Interpreting feature importance for aortic injury using SHAP and Permutation Importance methods described in the study.
• Conducting network analysis to explore structural relationships between different types of vascular injuries.
• Validating machine learning approaches for medical image-derived feature analysis in a clinical pilot setting.

Strengths

• Includes data from 500 patients, providing a basis for exploratory machine learning analysis.
• Features were analyzed using established methods including Random Forest, LASSO regression, SHAP, and Permutation Importance.
• The dataset is associated with a detailed methodological description of a clinical angioscopy study.

Limitations

• Row count and column-level documentation are unknown, which limits suitability assessment.
• Data is from a single-center pilot study; the authors note a need for validation in larger multicenter studies.
• The dataset file is a 16.2 KB DOCX document, suggesting it may contain summary results or a manuscript rather than raw structured data.

Provenance

Source

figshare, author Satoru Takahashi.

Collection Method

Single-center, cross-sectional observational pilot study using non-obstructive general angioscopy.

Freshness

Last updated 2026-05-07 05:44:25.