AUTHOR=Zhu Wanbo , Zhang Xianzuo , Fang Shiyuan , Wang Bing , Zhu Chen 

TITLE=Deep Learning Improves Osteonecrosis Prediction of Femoral Head After Internal Fixation Using Hybrid Patient and Radiograph Variables

JOURNAL=Frontiers in Medicine

VOLUME=Volume 7 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2020.573522

DOI=10.3389/fmed.2020.573522

ISSN=2296-858X

ABSTRACT=Femoral neck fractures (FNFs) are a significant recognized public health problem with high incidence of death and disability. Osteonecrosis of femoral head (ONFH) after internal fixation of FNF is a major cause of disability and reoperation. Early intervention is able to prevent ONFH in preliminary stage. However, unavoidable missed diagnosis limits the early detection. The primary objective of this study was to develop a predictive model for postoperative ONFH using hybrid patient and radiograph variables based on deep leaning (DL). A two-center retrospective study of patients undergo closed reduction and cannulated screw fixation was performed. We trained a CNN model on postoperative pelvic radiographs and output regressive radiograph variables. Nomograms incorporating hybrid patient and radiograph variables was developed to analysis predictive performance. A total of 238 patients from two hospitals with 95 ONFH and 143 non-ONFH were included. A convolutional neural network (CNN) model was trained based on postoperative radiograph and output radiograph variables. Accuracy in validation set was 0.873 for the CNN model, and the algorithm achieved an AUC of 0.912 for prediction. Addition of DL-based radiograph variables to the clinical nomogram showed improved predictive performance, with an AUC of 0.948 (95% CI, 0.920 to 0.976) and better calibration. Decision curve analysis demonstrated increased clinical usefulness of the DL facilitated nomogram. In conclusion, we present a DL facilitated nomogram that incorporates the hybrid radiograph and patient variables, which improves preoperative osteonecrosis prediction of femoral head after internal fixation.