Application of deep learning ultrasound imaging in monitoring bone healing after fracture surgery

Abstract

Ultrasound imaging can be used to monitor bone healing after fracture. It has the advantages of simplicity, rapidity, non-radiation, and real-time dynamic observation of fracture healing. It is increasingly valued and developed in clinical applications. Deep learning is an essential branch of artificial intelligence widely used in the medical field. After fracture surgery, 134 patients were selected for ultrasonic imaging at 2 weeks, 4 weeks, 8 weeks, and 12 weeks after surgery. Among them, 104 patients were selected for ultrasonic imaging to establish a prediction model for ultrasonic deep learning, and then 30 patients were used for imaging to verify and improve the model. Finally, 30 patients were selected for ultrasonic imaging and X-ray imaging at 2 weeks, 4 weeks, 8 weeks, and 12 weeks after fracture surgery for the same patient; test the ultrasonic imaging prediction model, compare the ultrasonic imaging and X-ray imaging manual reading, and evaluate the ultrasonic imaging prediction model by accuracy, loss curve, F1 score value, roc curve, AUC value, etc. The accuracy of the ultrasonic imaging deep learning prediction model in the test set was 0.73. In the test set, the predicted AUC values for 2 weeks, 4 weeks, 8 weeks, and 12 weeks after fracture surgery were 0.82, 0.90, 0.86, and 0.67, respectively. The accuracy of ultrasonic imaging manual film reading was 0.7, and the AUC values for the four periods were 0.76, 0.81,0.81, and 0.86, respectively. X-ray imaging manual film reading accuracy was 0.7, and the AUC values for the four periods were 0.86, 0.76, 0.81, and 0.84, respectively. Based on the artificial intelligence framework, a deep learning prediction model is constructed by preprocessing the ultrasonic imaging after fracture surgery to predict fracture healing, and its application value is discussed. The prediction model of ultrasonic imaging based on deep learning can be used as an effective monitoring method for bone healing after fracture surgery.

Ähnliche Arbeiten