Utility of a novel integrated deep convolutional neural network for the segmentation of hip joint from computed tomography images in the preoperative planning of total hip arthroplasty

Abstract

Purpose Preoperative three-dimensional planning is important for total hip arthroplasty. To simulate the placement of joint implants on computed tomography (CT), pelvis and femur must be segmented. Accurate and rapid segmentation of the hip joint is challenging. This study aimed to develop a novel deep learning network, named Changmugu Net (CMG Net), which could achieve accurate segmentation of the femur and pelvis. Methods The overall deep neural network architecture of CMG Net employed three interrelated modules. CMG Net included the 2D U-net to separate the bony and soft tissues. The modular hierarchy method was used for the main femur segmentation to achieve better performance. A layer classifier was adopted to localise femur layers among a series of CT scan images. The first module was a modified 2D U-net, which separated bony and soft tissues; it provided intermediate supervision for the main femur segmentation. The second module was the main femur segmentation, which was used to distinguish the femur from the acetabulum. The third module was the layer classifier, which served as a post-processor for the second module. Results There was a much greater overlap in accuracy results with the “gold standard” segmentation than with competing networks. The dice overlap coefficient was 93.55% ± 5.57%; the mean surface distance was 1.34 ± 0.24 mm, and the Hausdorff distance was 4.19 ± 1.04 mm in the normal and diseased hips, which indicated greater accuracy than the other four competing networks. Moreover, the mean segmentation time of CMG Net was 25.87 ± 2.73 s, which was shorter than the times of the other four networks. Conclusions The prominent segmentation accuracy and run-time of CMG Net suggest that it is a reliable method for clinicians to observe anatomical structures of the hip joints, even in severely diseased cases.

Optimization of Revision Hip Arthroplasty Workflow by Means of Detailed Pre-Surgical Planning Using Computed Tomography Data, Open-Source Software and Three-Dimensional-Printed Models

“…Such solutions have been shown to be effective used in trauma patients and may be helpful in pre-surgical planning of RHA [44] Another method is the use of neural network (NN) and artificial intelligence (AI) in the process of bone tissue segmentation [45]. Wu et al [46] showed that the use of new neural networks (e.g., CNN-CMG Net) increases the accuracy of bone tissue segmentation while reducing the time required to create 3D models of such anatomical structures, which clearly indicates a potential research direction to optimize the presurgical planning process. However, it should be emphasized that these promising results focused on healthy bone tissue.…”

Extra‐articular Blocking Technique to Resolve Severe Acetabular Bone Defect in Developmental Dysplasia of the Hip

“…In addition, quantitative analysis of CT data, such as artificial intelligence–assisted preoperative planning (AI‐hip, Longwood Valley MedTech, Beijing, China), 7 is an efficient way to visualize the bone defect located at the superolateral aspect of the acetabulum, and to predict the precise data of uncoverage when the acetabular cup is placed at 40° of inclination and 15° of anteversion at the level of true acetabulum. The intelligent image segmentation system of AI‐assisted preoperative planning (AI‐HIP software, Longwood Valley MedTech, Beijing, China) is based on convolutional neural networks and deep learning algorithms, and was constructed on the basis of over 1.2 million CT images from 3000 patients.…”

“…The comprehensive workflow is integrated into the AI‐HIP software. The accuracy of the AI‐HIP has been validated through several clinical studies 7 …”

“…The accuracy of the AI-HIP has been validated through several clinical studies. 7 The close relevance of coverage difference and the failure rate of the acetabular cup has been fully established, and most studies have revealed that the cutoff value of the uncoverage percentage to acetabular cup loosening is 30%-40%. 8 Therefore, the second indication for extraarticular blocking technique is that the bone defect is verified at the superolateral aspect of the acetabulum, and that the uncoverage percentage is between 20% to 40% of the aera of the acetabular cup by a CT data-based measurement.…”

A novel image augmentation based on statistical shape and intensity models: application to the segmentation of hip bones from CT images

“…In this work, we chose the task of segmenting bones from computed tomography (CT) images for total hip arthroplasty (THA) planning. Most works that relied on deep learning to segment (some) bones of the hip from CT in the context of THA [ 15 , 16 ] used basic geometric augmentations such as rotation, translation, scaling, cropping, and left–right flipping. This was also observed with other studies that did not focus on THA but also segmented bones of the hip joint from CT [ 17 ] and sometimes included simple intensity augmentations such as intensity scaling [ 18 ].…”