Femoral Fracture Assessment Using Acceleration Signals Combined with Convolutional Neural Network

PurposeThe treatment of fractured bones is crucial for the recovery of injuries during the healing process of femur fractures. Both qualitative and quantitative analyses are critical in the treatment of fractured bones. The healing process of femur fractures can be regarded as the reverse process of its damage degradation. This paper presents a qualitative and quantitative method to assess fracture healing based on the combination of acceleration signals and a convolution neural network (CNN).Materials and methodsThree types of normal bone, osteoporotic bone, and severely osteoporotic bone were investigated. Femurs with different fracture degrees were fabricated to simulate the damage process. Ten cracks with different depths created in the 1/2 and 1/3 locations of artificial femurs were fabricated. Three different scenarios were investigated to confirm the correctness and effectiveness of the proposed method. Accelerometer signals were used to monitor the fracture healing process, and these signals serve as inputs for the CNN.ConclusionThe results show that the averaged accuracy exceeds 99% in prediction of the fracture location under all the scenarios. The fracture assessment method proposed in this study can provide reliable reference values for the quantitative fracture degree analyses.

Automated summary

This study presents a qualitative and quantitative method to assess fracture healing based on acceleration signals and CNN. It investigates different degrees of fractures and confirms the correctness and effectiveness of the proposed method through experiments.