Foot2hip: A Deep Neural Network Model for Predicting Lower Limb Kinematics From Foot Measurements

Objective: This study aims to develop a neural network (foot2hip) for long-term recording of gait kinematics with improved user comfort. Methods: Foot2hip predicts ankle, knee, and hip joint angle profiles in the sagittal plane using foot kinematics and kinetics during walking. Foot2hip consists of three convolutions, two max-pooling, two LSTM, and three dense layers. An indigenously developed insole and an outsole were used to measure the kinetics and kinematics of the foot, respectively. Seven healthy participants were recruited to follow an experimental protocol consisting of six walking conditions: slow, medium, fast walking speed, rearfoot, flatfoot, and forefoot landing pattern. Results: When tested for leave-one-out and nested cross-validation, foot2hip obtained
$3.04^{\circ }\pm 0.20$
RMSE and 0.97
$\ \pm$
0.01 correlation coefficient for knee joint,
$1.7^{\circ }\pm$
0.09 RMSE and 0.95
$\ \pm$
0.01 correlation coefficient for hip joint, and
$1.32^{\circ }\pm$
0.08 RMSE and 0.91
$\ \pm$
0.02 correlation coefficient for ankle joint (averaged across all folds). Conclusion: The prediction performance of foot2hip is encouraging and shows its applicability in accurately predicting lower limb kinematics with minimal wearables. Significance: The hardware used along with foot2hip is low cost ($268 +
N
× $35,
N
is the number of different foot sizes), comfortable, and easy to use. Therefore, the system is suitable for most clinical and personal applications.