Plantar Fasciitis Detection Based on Deep Learning Architecture

Background: Plantar fasciitis is one of the most common foot pain problems in adults. The current diagnosis mainly relies on the inquiry of medical history and a physical examination of the body. In the objective laboratory examination, the blood test has not yet provided an effective diagnostic reference. In this study, we combine a deep learning algorithm architecture with thermal imaging to develop a plantar fasciitis medical decision system that predicts whether the patient has the condition. Methods: This study collected patient image-related data, including 360-degree thermal video and RGB images of the affected area (foot), and patient clinical data. In data preprocessing, we first adjust the thermal image data, based on the different detection environments. After data processing, we employed the Convolutional Neural Networks (CNN) deep learning architecture to develop a prediction model. Results: In total, 1,000 frames were used as the training dataset in this study---300 cases that had the condition and 700 cases that did not. The results showed that the CNN model can effectively predict plantar fasciitis. The inflammatory response is often accompanied by redness and swelling. This study used thermal imaging to detect the temperature of the affected area, which it combined with a deep learning algorithm to successfully detect the inflammatory condition. In the future, this technique can be used to detect other inflammatory reactions such as wound healing and hemorrhoids.