Orthopedic Implant Fixation Type Detection in Hip Arthroplasty Using Two-Staged Low-Shot Network

Hip Arthroplasty is a surgical procedure in which the hip joint is replaced by a prosthetic implant. These implants fail over time and need revision surgery. The burden of revision of a failed arthroplasty is gradually increasing as the number of primary arthroplasty have undergone an exponential increase. It is very important to identify the type of manufacturer and fixation from x-ray images to plan the surgeries better as well in getting the right set of instruments for the surgery. There are 3 types of fixation in hip arthroplasty namely cemented, cementless, and hybrid. In spite of its significance, to the best of our knowledge, we could not find any automated work on fixation type detection. In this work, we have annotated the open source dataset [2] with fixation type and proposed a two-stage low-shot learning approach for detecting the different types of fixation. The key challenge was inter-class confusion and class imbalance. Our proposed work provides 93% classification accuracy and 86% accuracy when blind-tested from another source of clinical data.