Convolutional Neural Networks to Study Contrast-Enhanced Magnetic Resonance Imaging-Based Skeletal Calf Muscle...

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Cardiovascular research

Background:

• Peripheral artery disease (PAD) impairs lower extremity blood flow and causes skeletal muscle microvascular changes.
• Accurate differentiation of PAD patients from controls is crucial for effective management.

Purpose of the Study:

• To investigate the efficacy of deep learning (DL) models, specifically CNNs, in distinguishing PAD patients from controls.
• To utilize CE-MRI perfusion patterns in skeletal calf muscles for PAD classification.

Main Methods:

• Acquired CE-MRI perfusion data from 56 individuals (36 PAD patients, 20 controls) after reactive hyperemia.
• Segmented skeletal calf muscles and processed 3D CE-MRI perfusion scans for DL analysis.
• Trained and evaluated multiple CNN models (resNet, divNet) for PAD classification.

Main Results:

• The best performing CNN models achieved a peak accuracy of 75% in differentiating PAD patients.
• Specificity for resNet and divNet was 80% and 94%, respectively.
• DL models demonstrated capability in classifying PAD based on calf muscle perfusion.

Conclusions:

• Deep learning models, utilizing CNNs and CE-MRI calf muscle perfusion, can successfully discriminate between PAD patients and controls.
• DL methods offer a potential new avenue for the investigation and diagnosis of PAD.