| Abstract: | Walking has been shown to benefit individuals include Diabetes Mellitus (DM) patients and peripheral artery disease. However, brisk walking and continuous walking could produce repetitive loads and stresses on the plantar foot resulting in increased plantar tissue stiffness and peak plantar pressure (PPP), leading to a high risk of foot ulcer formation and tissue injury. Therefore, quantifying the walking intensity is essential for rehabilitation interventions to indicate suitable walking exercise.
This study is divided into three objectives. First, this study aims to identify differences in walking speeds to the plantar pressure response using deep learning methods, including Resnet50, InceptionV3, and MobileNets. Second, this study proposed a machine learning model to classify the walking speed and duration using plantar region pressure images. Third, prediction of plantar tissue stiffness based on plantar stress pattern using vision transformer. The F-scan system (Tekscan, South Boston, MA, USA) was used to measure plantar pressures during walking. An elastographic ultrasound (Aloka Pro Sound Alpha 7, Hitachi Healthcare Americas, Twinsburg, OH, USA) with a linear array transducer (UST-5412, 5–13 MHz, Hitachi Healthcare Americas) was used to measure plantar tissue mechanical property. In the first study, the deep learning models were used to classify the plantar pressure images of healthy people walking on a treadmill. The design consisted of three walking speeds (0.8 m/s, 1.6 m/s, and 2.4 m/s). The second study, an Artificial Neural Network (ANN), was adopted to develop a model for walking intensity classification using different plantar region pressure images, including the first toe (T1), the first metatarsal head (M1), the second metatarsal head (M2), and the heel (HL). The classification consisted of three walking speeds (i.e., slow at 0.8 m/s, moderate at 1.6 m/s, and fast at 2.4 m/s) and two walking durations (i.e., 10 min and 20 min). The third study used vision transformers to predict the relationship between plantar tissue stiffness with a plantar pressure pattern image.
The experimental results show that artificial intelligence technology could predict walking intensity and analyze the relationship between plantar tissue stiffness and plantar pressure pattern image. The first study indicated that Resnet50 had the highest accuracy compared to InceptioanV3 and MobileNets on analyzing plantar pressure distribution images. Furthermore, the experimental results of estimation of walking speed and duration based on four regions of plantar pressure (i.e., T1, M1, M2, and HL) with an ANN showed that the T1 region was more easily recognized by the ANN model, as evidenced by the highest F1-score value than other regions. Meanwhile, detection of the relationship between plantar tissue stiffness with plantar pressure pattern images showed that vision transformers could map the relationship between plantar tissue stiffness and plantar stress pattern images. |
