Accuracy of New Deep Learning Model-Based Segmentation and Key-Point Multi-Detection Method for Ultrasonographic Developmental Dysplasia of the Hip (DDH) Screening

Abstract

Hip joint ultrasonographic (US) imaging is the golden standard for developmental dysplasia of the hip (DDH) screening. However, the effectiveness of this technique is subject to interoperator and intraobserver variability. Thus, a multi-detection deep learning artificial intelligence (AI)-based computer-aided diagnosis (CAD) system was developed and evaluated. The deep learning model used a two-stage training process to segment the four key anatomical structures and extract their respective key points. In addition, the check angle of the ilium body balancing level was set to evaluate the system’s cognitive ability. Hence, only images with visible key anatomical points and a check angle within ±5° were used in the analysis. Of the original 921 images, 320 (34.7%) were deemed appropriate for screening by both the system and human observer. Moderate agreement (80.9%) was seen in the check angles of the appropriate group (Cohen’s κ = 0.525). Similarly, there was excellent agreement in the intraclass correlation coefficient (ICC) value between the measurers of the alpha angle (ICC = 0.764) and a good agreement in beta angle (ICC = 0.743). The developed system performed similarly to experienced medical experts; thus, it could further aid the effectiveness and speed of DDH diagnosis.

Development of a Fully Automated Graf Standard Plane and Angle Evaluation Method for Infant Hip Ultrasound Scans

“…It can be seen from Table 2 that the mean α value of the test set is 60.94°, which means that measurement error may lead to misclassification [ 28 ]. While in SM mode, both the classification agreement rate and Cohen’s kappa show high classification performance compared with that of experienced sonographers (classification agreement rate: 82%~91%, Cohen’s kappa: 0.60~0.86) reported by [ 38 ] and better than for other CAD methods [ 28 , 29 , 31 ]. Nevertheless, we still suggest sonographers pay more attention to the potential misclassification situation of hips with α around 60° (58°~62°).…”

“…The above results demonstrate that, in terms of measurement accuracy, the performance of SM mode is better than that of human observers. The MAEs of the α, β angles demonstrate that SM mode also performs better than other deep learning-based methods [ 28 , 29 ].…”

“…In summary, the dynamic measurement mode of the proposed framework shows good measurement performance and classification performance, similar to experienced sonographers. In addition, compared with the methods based on two-stage instance segmentation networks [ 28 , 29 ], our YOLOv3-tiny-based method has more advantages in terms of the number of parameters, computational cost, and inference speed, which makes it possible to implement our framework in ultrasound devices. Considering that it is a fully automated diagnostic pipeline with high running speed and does not require any manual intervention, the application of the dynamic measurement mode may effectively improve the work quality and efficiency of sonographers in infant DDH screening.…”

“…To deal with these problems, a large number of computer-aided diagnostic (CAD) methods have been proposed to reduce manual intervention and improve the quality and efficiency of the DDH diagnostic process. As shown in Table 1 , most of the CAD methods focused on one of the following steps in the Graf diagnosis process, such as the evaluation of the standard plane [ 17 , 18 , 19 , 20 , 21 , 22 , 23 ], the measurement of α and β [ 15 , 24 , 25 , 26 , 27 , 28 , 29 ], or the classification of hip dysplasia severity [ 30 , 31 ]. Due to their powerful ability to hierarchically mine and utilize multi-level features, deep learning-based methods can generally bring more accurate and robust diagnostic results than conventional traditional image processing-based methods.…”

Comment on Lee et al. Accuracy of New Deep Learning Model-Based Segmentation and Key-Point Multi-Detection Method for Ultrasonographic Developmental Dysplasia of the Hip (DDH) Screening. Diagnostics 2021, 11, 1174

“…We have read the article titled “Accuracy of New Deep Learning Model-Based Segmentation and Key-Point Multi-Detection Method for Ultrasonographic Developmental Dysplasia of the Hip (DDH) Screening” by Lee et al with great interest [ 1 ]. This paper focused on the usability of an artificial intelligence (AI)–computer aided detection (CAD) system for screening and diagnosis of DDH, for which the authors used Graf’s method to evaluate the ultrasonographic results.…”