Visual Abstract

Due to muscle dystrophy, diabetes patients suffer from foot gait movement and need to define whether their muscle contraction-relaxation is normal. However, the way to determine diabetic foot is inconvenient and may give stress to patients during the test by attaching multiple electrodes on their foot. Also, after measuring the EMG, it takes several hours or days to figure out whether the testing subject has a diabetic foot. Therefore, we developed an EMG measured diabetic determination system by machine learning protocol based on LSTM (Long Short-Term Memory). We tested 15 individuals with normal glycemia and 15 diabetic patients who have been suffered more than 10 years with fasting glucose levels over 180mg/dl. Every patient’s EMG wave pulse was measured for one minute with 3-second terms of contraction and relaxation for three times, Subjects’ wave pulse was correct as normal and diabetes to clear machine learning process. Every second 150 data were created and the middle range was normalized. Numbers between 0 and 1 range were used as a y normalization for machine learning test training. After test learning, we input normal and diabetes wave pulse patterns whether this LSTM model could make a clear definition. (Fig 1.) As result, this machine learning model could distinguish 88.21% of diabetes patients’ feet by comparing foot EMG wave pulse, but 11.79% of diabetes foot cannot be determined due to an inefficient data pool.

Disclosure

D. Kim: None. M. Park: None. D. Jee: None. G. Im: None. J. Lee: None. J. Shin: None. J. Kim: None. W. Kim: None.

Funding

Korea Institute of Startup & Entrepreneurship Development

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.