Related Experiment Video
Updated: Aug 8, 2026
06:26
Novel and Innovative Hybrid Technique for Type A Aortic Dissection
Published on: March 28, 2025
An explainable machine learning model for type B aortic dissection identification: development and internal
Donglin Li1, Dilinuerkezi Abulimiti1, Zaiying Yeerbao1
1Department of Vascular Surgery, The First Affiliated Hospital of Xinjiang Medical University, Urumqi, China.
Frontiers in Public Health
|August 7, 2026
Summary
This study developed an explainable machine learning model to identify Type B aortic dissection (TBAD) using clinical history and lab data. The model, particularly a neural network, showed good accuracy, highlighting key predictors like lymphocyte percentage and hypertension.
Area of Science:
- Cardiovascular Medicine
- Artificial Intelligence
- Medical Informatics
Background:
- Type B aortic dissection (TBAD) is a critical cardiovascular emergency requiring prompt diagnosis.
- Accurate and timely identification of TBAD is essential for effective patient management and improved outcomes.
Purpose of the Study:
- To develop and internally evaluate an explainable machine learning (ML) model for identifying existing TBAD.
- To utilize routinely available clinical history and admission laboratory data for TBAD prediction.
- To enhance the interpretability of ML models in diagnosing cardiovascular emergencies.
Main Methods:
- A single-center retrospective case-control study involving 1,640 participants (854 TBAD, 786 controls).
- Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify key predictors from 38 candidate features.
- Development and comparison of eight ML algorithms, including neural networks, SVM, and gradient boosting methods.
- Model performance evaluated using Area Under the Receiver Operating Characteristic Curve (AUC) in an internal test set.
- Exploratory interpretation using SHapley Additive exPlanations (SHAP) for the selected model.
Main Results:
- LASSO regression identified five key predictors: hypertension, white blood cell count, lymphocyte percentage, basophil percentage, and monocyte count.
- The neural network model demonstrated strong discrimination, achieving an AUC of 0.852 in the internal test set.
- SHAP analysis revealed lymphocyte percentage, hypertension, and monocyte count as major contributors to TBAD prediction.
- Lymphocyte percentage exhibited the highest mean absolute SHAP value, indicating its significant role in the model's predictions.
Conclusions:
- An explainable ML model was successfully developed and internally evaluated for identifying TBAD using clinical and laboratory data.
- The neural network model showed promising discriminatory performance, with lymphocyte percentage and hypertension being significant predictors.
- The study is considered exploratory due to the lack of external validation; further validation in acute symptomatic populations is recommended before clinical implementation.