A machine learning-based approach to identify peripheral artery disease using texture features from contrast-enhanced...

Area of Science:

• Medical imaging
• Machine learning
• Cardiovascular diagnostics

Background:

• Peripheral artery disease (PAD) diagnosis relies on assessing impaired blood circulation, particularly in calf muscles, leading to intermittent claudication.
• Altered microvascular perfusion and connective tissue changes in PAD patients may manifest as detectable texture variations in skeletal muscles.

Purpose of the Study:

• To develop an automated pipeline for extracting textural features from CE-MRI scans.
• To utilize these features to train machine learning models for PAD detection and risk stratification.

Main Methods:

• An automated pipeline was created for textural feature extraction from contrast-enhanced magnetic resonance imaging (CE-MRI).
• Machine learning models were trained using features from 36 PAD patients and 20 controls, with data split 7:3 and cross-validation.
• Feature selection methods were applied to optimize model performance.

Main Results:

• A 2-class classification (controls vs. PAD) achieved a peak accuracy of 94.11% and a mean testing accuracy of 84.85%.
• A 3-class classification identified PAD subgroups (controls vs. PAD without diabetes vs. PAD with diabetes) with 83.23% average accuracy.
• Classification of exercise outcomes (controls, PAD completers, non-completers) yielded 78.60% average accuracy.

Conclusions:

• Machine learning applied to CE-MRI texture features offers a promising non-invasive method for PAD diagnosis.
• This approach may aid in identifying high-risk PAD subgroups and assessing disease progression or treatment response.
• Imaging-based texture analysis holds potential for studying lower extremity ischemia.