A preliminary examination of the diagnostic value of deep learning in hip osteoarthritis

Abstract

Hip Osteoarthritis (OA) is a common disease among the middle-aged and elderly people. Conventionally, hip OA is diagnosed by manually assessing X-ray images. This study took the hip joint as the object of observation and explored the diagnostic value of deep learning in hip osteoarthritis. A deep convolutional neural network (CNN) was trained and tested on 420 hip X-ray images to automatically diagnose hip OA. This CNN model achieved a balance of high sensitivity of 95.0% and high specificity of 90.7%, as well as an accuracy of 92.8% compared to the chief physicians. The CNN model performance is comparable to an attending physician with 10 years of experience. The results of this study indicate that deep learning has promising potential in the field of intelligent medical image diagnosis practice.

Stacked sparse autoencoder networks and statistical shape models for automatic staging of distal femur trochlear dysplasia

“…7,39 As Autoencoders and convolutional networks have been recently applied to biomedical signal/image analysis with applications in bone age assessment, 36 cardiac rhythm modelling, 37 cancer cell detection, 38 schizophrenia pattern identification, 39 automatic diagnosis of the prostate cancer, 40 and hip osteoarthritis study. 41 In this work, we confirmed that, by means of the morphing parameters α j (Equation (2)), SSPA network can be used to classify the morphologic variability in 3D femur models with results superior to traditional polynomial clustering with LDA/QDA. We demonstrated that a single autoencoder layer, stacked to a Softmax layer, enabled the classification of the three categories of distal femur shapes discriminating healthy morphologies from mild and severe anomalies of the trochlear region.…”

“…To get a deeper insight into the application of SSM for trochlear staging, here we investigate SSM ability to quantify distal femur abnormalities and the feasibility of using patient‐specific morphing parameters to automate the classification of distal femur shapes affected by dysplastic condition. Stacked sparse autoencoder (SSPA) networks, which were recently adopted in many different clinical applications, were tailored and trained using deep learning techniques to enable the discrimination of mild and severe morphologic anomalies of the trochlear region with respect to healthy morphologies. Particularly, the specific contributions of this study comprise the following: the evaluation of the morphologic differences between healthy and dysplastic femurs, described by single modes of variations in the SSM; the description of the interplay among modes of variations, which map specific anatomic features in the distal femur correlated with trochlear dysplasia; the discussion about the SSM generation of new associations among local morphologic features, which are, in principle, uncorrelated with the dysplastic condition; the use of the patient‐specific morphing parameters to train one SSPA network for the classification of multiple dysplastic severity grades. …”

“…To get a deeper insight into the application of SSM for trochlear staging, here we investigate SSM ability to quantify distal femur abnormalities and the feasibility of using patient-specific morphing parameters to automate the classification of distal femur shapes affected by dysplastic condition. Stacked sparse autoencoder (SSPA) networks, which were recently adopted in many different clinical applications, [36][37][38][39][40][41] were tailored and trained using deep learning techniques to enable the discrimination of mild and severe morphologic anomalies of the trochlear region with respect to healthy morphologies. Particularly, the specific contributions of this study comprise the following:…”

Evaluation of artificial intelligence for detecting periapical pathosis on cone‐beam computed tomography scans

“…This AI method can learn adaptive image characteristics and simultaneously make image classifications (LeCun et al , Shin et al , Tajbakhsh et al , Kim & MacKinnon ). CNNs have been successfully used for automatic assessment of various medical and dental problems, including image‐based automated diagnosis to detect lung and brain lesions (Akkus et al , Song et al , Wang et al , Blanc‐Durand et al ), breast cancer in mammography images (Becker et al ), colorectal polyps and prostate cancer (Wang et al , Byrne et al ), skin cancer (Esteva et al ), diabetic retinopathy in retinal fundus photographs (Gulshan et al ), hip osteoarthritis (Xue et al ) and bone age assessment (Lee et al ). In dentistry, CNNs have been applied to detect carious lesions, periapical lesions, tooth eruption and numbering, vertical root fractures, assess root morphology or periodontal bone loss, dental and jaw pathosis, osteoporosis, and maxillary sinusitis on dental radiographs (Kositbowornchai et al , Miki et al , Ezhov et al , Murata et al , Poedjiastoeti & Suebnukarn , Lee et al ,b, Zakirov et al , Zakirov et al , Chen et al , Ekert et al , Hiraiwa et al , Hwang et al , Krois et al , Tuzoff et al ).…”

Automating Three-dimensional Osteoarthritis Histopathological Grading of Human Osteochondral Tissue using Machine Learning on Contrast-Enhanced Micro-Computed Tomography

“…The development of machine learning techniques has enabled a data-driven approach in pattern recognition and decision making without the need for explicit programming. Machine learning has been applied in clinical OA research in several domains, such as the prediction of OA severity [28][29][30][31] and progression 15,32,33 using X-ray radiographs 28,29,31,32 or MRI analysis 15,30,33 . However, little attention has been paid to machine learning in pre-clinical OA research 26,34,35 .…”