Radiomics-driven perfusion prediction in clinical photoacoustic foot imaging.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Accurate tissue perfusion assessment is critical for managing chronic foot ulcers in patients with diabetes and peripheral arterial disease.
• Current perfusion evaluation methods using photoacoustic (PA) imaging have limitations.

Purpose of the Study:

• To develop and validate a fully automated radiomics-based framework for predicting perfusion conditions using single-wavelength clinical PA foot imaging.
• To enhance the clinical utility of PA imaging for perfusion assessment.

Main Methods:

• Extracted radiomics features from raw radiofrequency (RF) signals and maximum amplitude projection (MAP) images.
• Applied minimum redundancy maximum relevance (mRMR) and ReliefF for feature selection.
• Utilized a k-nearest neighbors ensemble model for perfusion prediction.

Main Results:

• Achieved an area under the curve (AUC) of 0.90 for training and 0.94 for testing.
• Identified eight key radiomics features correlating with physiological indicators like vessel density and tissue structure.
• Demonstrated a reliable and interpretable method for perfusion assessment.

Conclusions:

• The proposed radiomics framework offers a reliable and interpretable method for assessing tissue perfusion in PA imaging.
• This approach shows strong potential for improving the management of chronic foot ulcers in diabetic patients with peripheral arterial disease.