Explainable Prediction of Whether the Acetabular Cup is Placed in the “safe Zone” from X-ray Images

Fuchang Han,Shenghui Liao,Chao Xiong,Haitao Wei,Renzhong Wu,Yingqi Zhang

DOI: https://doi.org/10.1109/bibm52615.2021.9669453

2021-01-01

Abstract:Appropriate acetabular cup placement is an important step for the prevention of postoperative dislocation in total hip arthroplasty. Existing state-of-the-art methods focus on developing statistical methods and image-matching techniques. Such developments employ regression equations or 3D planning software to evaluate the acetabular cup angles, and little attention has been given to automatic prediction with machine/deep learning. In addition, existing approaches ignore the explainability and clinical utility of models, which creates a tension between accuracy and explainability. This paper proposes an innovative hybrid framework for acetabular cup angle prediction. The proposed method is based on deep stacked network (DSN) and SHapley Additive exPlanations (SHAP) with four improvements: (1) an arc-support line segments-based unsupervised segmentation module, obtaining the region of interest (ROI) occupied by the acetabular cup component automatically and efficiently; (2) a multiscale feature extraction and attribute optimization module, obtaining comprehensive radiomic features including the conventional texture features and deep learning features; (3) appending a SHAP module to the framework to improve the explainability of models through three main steps: local explanation, global explanation, and interactive visualization; (4) our method employs DSN method to predict the acetabular cup angles, and it considers the decision curve analysis (DCA) to tackle the problem of clinical utility evaluation. By integrating the above improvements in series, the models’ performances are gradually enhanced. Experimental results show that our model achieves superior results to existing acetabular cup angle prediction approaches.