Key result
A Tree machine learning algorithm classifying Doppler ultrasound blood flow signals achieved 92.1% accuracy and an AUC of 0.91 for the early detection of peripheral arterial disease in diabetic patients.
Why the study?
Peripheral arterial disease is a key indicator of diabetic foot that can be identified by ultrasound, motivating early detection through classification of lower extremity arterial blood flow signals.
Does machine learning classification of Doppler ultrasound signals accurately detect peripheral arterial disease in diabetic patients?
Cross-Sectional (n=354)
No
Does machine learning classification of Doppler ultrasound signals accurately detect peripheral arterial disease in diabetic patients?
A Tree-based machine learning model can classify Doppler ultrasound blood flow signals of the posterior tibial artery with 92.1% accuracy, offering a potential automated screening tool for early peripheral arterial disease in diabetic patients.
May support ML-based Doppler screening for diabetic PAD; hypothesis-generating and requires prospective validation before practice change.
Peripheral arterial disease is one of the key indicators of diabetic foot, which can be easily identified by ultrasound diagnostic techniques. The work aims to detect diabetic foot in early stages by classifying the blood flow signals of lower extremity arteries being captured by ultrasound doppler methods. Samples are collected from diabetic patients with and without having probable symptoms of arterial diseases. Doppler examination has been conducted on posterior tibial artery for 354 subjects with a transducer of 8 MHz frequency. The auscultation, method of listening sounds of internal organs, is employed as medical diagnostic tool for identifying pathological conditions. Each artery in the human body has a unique profile of Doppler flow. This fixed profile may be changed with the presence of a particular disease. The received signal has a spectrum of Doppler-shifted signals with respect to the existence of a velocity profile across the vessel lumen. Changes to the shape of this profile is an indicator of the severity of disease. Various features are extracted by using various statistical and signal processing functions. The feature analysis was accomplished with machine learning algorithms. Naïve Bayes, Tree and SVM algorithms are employed with MATLAB Toolboxes. Comparing the performance of these algorithms, the Tree method is found superior than the others. So, the proposed classification methodology can be employed as a key factor for the early stage detection of diabetic foot. As diabetic foot is correlated with many other parameters which effects the pressure and flow velocity of lower extremities, an integrated disease prediction model is proposed by incorporating the ultrasound doppler technique.
No takes yet. Share an insight, caveat, or question.
KS et al. (2021) conducted a cross-sectional in Diabetic foot and Peripheral Arterial Disease (n=354). Tree machine learning algorithm for Doppler ultrasound signal classification vs. Naïve Bayes and SVM algorithms was evaluated on Classification accuracy. A Tree machine learning algorithm classifying Doppler ultrasound blood flow signals achieved 92.1% accuracy and an AUC of 0.91 for the early detection of peripheral arterial disease in diabetic patients.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: