Applying patient characteristics, stent-graft selection, and pre-operative computed tomographic angiography data to a...

Area of Science:

• Medical Imaging
• Machine Learning in Healthcare
• Vascular Surgery

Background:

• Endoleak is a primary concern after endovascular aneurysm repair (EVAR).
• Accurate prediction of endoleak is crucial for patient outcomes.
• Current methods may not fully integrate all relevant pre-operative data.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting endoleak post-EVAR.
• To integrate patient characteristics, stent-graft configuration, and pre-operative CTA vessel morphology into the prediction model.

Main Methods:

• Utilized pre-operative computed tomography angiography (CTA) data from 111 patients with abdominal aortic aneurysms.
• Extracted patient characteristics, stent-graft configuration, and vessel morphology (length, diameter, angle).
• Trained an extreme gradient boosting (XGBoost) ML model and evaluated its diagnostic performance using receiver operating characteristic analysis.

Main Results:

• The ML model achieved an area under the curve (AUC) of 0.88.
• High diagnostic performance was observed with an accuracy of 0.88, sensitivity of 0.85, and specificity of 0.91.
• The model demonstrated strong predictive capability for endoleak presence.

Conclusions:

• Machine learning integration of pre-EVAR CTA data, patient characteristics, and stent-graft configuration shows significant potential for endoleak prediction.
• This ML approach can aid in the interpretation of abdominal CTA scans for EVAR planning.
• The findings suggest a valuable application of ML in improving EVAR outcomes.