Automatic Hip Detection in Anteroposterior Pelvic Radiographs—A Labelless Practical Framework

Abstract

Automated detection of the region of interest (ROI) is a critical step in the two-step classification system in several medical image applications. However, key information such as model parameter selection, image annotation rules, and ROI confidence score are essential but usually not reported. In this study, we proposed a practical framework of ROI detection by analyzing hip joints seen on 7399 anteroposterior pelvic radiographs (PXR) from three diverse sources. We presented a deep learning-based ROI detection framework utilizing a single-shot multi-box detector with a customized head structure based on the characteristics of the obtained datasets. Our method achieved average intersection over union (IoU) = 0.8115, average confidence = 0.9812, and average precision with threshold IoU = 0.5 (AP50) = 0.9901 in the independent testing set, suggesting that the detected hip regions appropriately covered the main features of the hip joints. The proposed approach featured flexible loose-fitting labeling, customized model design, and heterogeneous data testing. We demonstrated the feasibility of training a robust hip region detector for PXRs. This practical framework has a promising potential for a wide range of medical image applications.

The Feasibility and Performance of Total Hip Replacement Prediction Deep Learning Algorithm with Real World Data

“…We designed a sequential two-stage hip replacement prediction framework to identify the possibility of THR in three months of hip regions of interest (ROI): HipRD [ 47 ] and SurgHipNet [ 48 ]. The first step was a localization model to identify hip ROI in the provided PXR and cropped ROI from the surrounding background to simplify further processing, avoid noise from other organs, and reduce computational power.…”

“…For hip localization training, an automatic practical framework was utilized to detect the ROI of hip joints in each PXRs. The detailed framework development method was described in a previous study [ 47 ]. We placed a bounding box at the center of the femoral head to detect the ROIs.…”

Automatic Detection of Medial and Lateral Compartments from Histological Sections of Mouse Knee Joints Using the Single-Shot Multibox Detector Algorithm

“…IoU > 0.5 (the cutoff value used in a previous paper). 19 2. The predicted bounding box contained both the articular cartilage and subchondral plate of the loading surfaces of both the femur and tibia.…”

“…When adapting deep learning to medical images where only a specific region of the image is needed for evaluation, a 2-step approach, which first detects the region of interest, and then performs the classification on the detected region, is increasingly being used for accurate evaluation. 18 - 20 We thought that cartilage degeneration scoring for the image of a histological section of the knee joint would benefit from this 2-step approach because the evaluation is mostly limited to the medial and lateral compartments, except when there is a special need to detect lesions outside of these compartments. Automating the first detection step ensures that the examiner does not arbitrarily exclude areas with strong or weak degeneration when setting the regions to be used for scoring.…”

Taking the Next Step in Personalised Orthopaedic Implantation

“…Artificial intelligence-based recognition of different types of implants and detection of the region of interest in standard x-rays will help to improve treatment by gathering big data [ 19 , 20 ]. With these large amounts of data, it will be easier to optimize standard situations for the individual patient and his or her specific anatomical requirements [ 21 , 22 ].…”