ORIGINAL ARTICLE Machine learning approach to predict center of pressure trajectories in a complete gait cycle: a feedforward neural network vs. LSTM network Ahnryul Choi 1,2 & Hyunwoo Jung 2 & Ki Young Lee 1 & Sangsik Lee 1 & Joung Hwan Mun 2 Received: 22 April 2019 /Accepted: 3 October 2019 # International Federation for Medical and Biological Engineering 2019 Abstract Center of pressure (COP) trajectories of human can maintain regulation of forward progression and stability of lateral sway during walking. The insole pressure system can only detect COP trajectories of each foot during single stance. In this study, we developed artificial neural network models that could present COP trajectories in an integrated coordinate system during a complete gait cycle using pressure information of the insole system. A feed forward artificial neural network (FFANN) and a long short-term memory (LSTM) model were developed. For FFANN, among 198 pressure sensors from Pedar-X insoles, proper input variables were selected using sequential forward selection to reduce input dimension. The LSTM model used all 198 signals as inputs because of its self-learning characteristic. As results of cross-validation, the FFANN model showed correlation coef- ficients of 0.98–0.99 and 0.93–0.95 in anterior/posterior and medial/lateral directions, respectively. For the LSTM model, correlation coefficients were similar to those of FFANN. However, the relative root mean square error (12.5%) of the FFANN model was higher than that (9.8%) of the LSTM model in medial/lateral direction ( p = 0.03). This study can be used for quantitative evaluation of clinical diagnosis and rehabilitation status for patient with various diseases through further training using varied databases. Keywords Center of pressure . Gait . Neural network . LSTM . Insole system 1 Introduction Center of pressure (COP) of human body is defined as the centroid of external forces acting on the flat side of the foot and the origin of ground reaction force vector [1, 2]. COP trajectory during gait consists of consecutive COP coordinates passing from hindfoot to forefoot [3]. COP trajectory can maintain the regulation of forward progression and the stabil- ity of lateral sway during a single stance. It can also balance weight distribution of the lower extremities in a double stance [ 4]. In addition, COP trajectories have been used as quantitative parameters for gait stability by considering the center of mass (COM) together [5, 6]. The greater the distance between the COM and the COP during gait, the larger the moment arm for the ground reaction force. Therefore, more control energy is required to counterbalance the increased mo- mentum in order to restore balance. Koldenhoven et al. [7] have reported that chronic ankle in- stability patients’ supporting position of the foot changes at the beginning of the stance phase, leading to an increased variation of COP trajectories during loading response of the gait. According to their study, a high COP variation linked to gait instability can increase the risk of ankle sprain [7]. Hallemans et al. [8] have analyzed the relationship between COM-COP distance and various gait parameters in a typical developmental child and found that COM-COP distance has a significant neg- ative correlation with the duration of stance phase of gait. They have concluded gait imbalance of a developmental child is compensated by rapid step exchange between the two feet [8]. COP parameters have also been used to evaluate various diseases during gait and determine the effectiveness of rehabil- itation devices and treatment methods [9–12]. * Joung Hwan Mun jmun@skku.edu 1 Department of Biomedical Engineering, College of Medical Convergence, Catholic Kwandong University, 24, Beomilro 579beongil, Gangneung, Gangwon 25601, Republic of Korea 2 Department of Bio-Mechatronic Engineering, College of Biotechnology and Bioengineering, Sungkyunkwan University, 2066 Seoburo, Jangan, Suwon, Gyeonggi 16419, Republic of Korea Medical & Biological Engineering & Computing https://doi.org/10.1007/s11517-019-02056-0