Bone X-Rays Classification and Abnormality Detection using Xception Network

Computer aided diagnosis (CAD) has a vital role and becomes an urgent demand nowadays. Bone fractures cases are considered from the most frequently occured dieases among individuals. Moreover, the incorrect diagnosis of the bone fractures cases may cause disability for the patient. Hence, CAD system for bone fractures has become a must. This paper proposes a two-stage classifcation method for bone type classification and bone abnormality detection. Xception pre-trained model is considered for all experiments. Two different approaches are utilized for the testing phase: 1) Singl-view and 2) Multi-view approachs. The enhanced images are fed into the first stage to be classified into one of the seven classes: shoulder, humerus, forearm, elbow, wrist, hand and finger. Thereafter, the classified bones are fed into the second stage to detect whether the bone is normal or abnormal. MURA dataset has been utilized for all experiments. Moreover, the last layer of the utilized model is replaced by Support Vector Machine (SVM) layer. The results reveal the superiority of the SVM layer.