DeepKneeExplainer: Explainable Knee Osteoarthritis Diagnosis From Radiographs and Magnetic Resonance Imaging

Abstract

Osteoarthritis (OA) is a degenerative joint disease, which significantly affects middleaged and elderly people. Although primarily identified via hyaline cartilage change based on medical images, technical bottlenecks like noise, artifacts, and modality impose an enormous challenge on high-precision, objective, and efficient early quantification of OA. Owing to recent advancements, approaches based on neural networks (DNNs) have shown outstanding success in this application domain. However, due to nested non-linear and complex structures, DNNs are mostly opaque and perceived as black-box methods, which raises numerous legal and ethical concerns. Moreover, these approaches do not have the ability to provide the reasoning behind diagnosis decisions in the way humans would do, which poses an additional risk in the clinical setting. In this paper, we propose a novel explainable method for knee OA diagnosis based on radiographs and magnetic resonance imaging (MRI), which we called DeepKneeExplainer. First, we comprehensively preprocess MRIs and radiographs through the deep-stacked transformation technique against possible noises and artifacts that could contain unseen images for domain generalization. Then, we extract the region of interests (ROIs) by employing U-Net architecture with ResNet backbone. To classify the cohorts, we train DenseNet and VGG architectures on the extracted ROIs. Finally, we highlight classdiscriminating regions using gradient-guided class activation maps (Grad-CAM++) and layerwise relevance propagation (LRP), followed by providing human-interpretable explanations of the predictions. Comprehensive experiments based on the multicenter osteoarthritis study (MOST) cohorts, our approach yields up to 91% classification accuracy, outperforming comparable stateof-the-art approaches. We hope that our results will encourage medical researchers and developers to adopt explainable methods and DNN-based analytic pipelines towards an increasing acceptance and adoption of AI-assisted applications in the clinical practice for improved knee OA diagnoses.

Development of a computer-aided tool for detection of COVID-19 pneumonia from CXR images using machine learning algorithm

“…Table 1 shows that previous studies [ [70] , [71] , [72] , [73] , [74] , 82 ] have achieved the minimum accuracy of 91% for multicenter osteoarthritis studies using chest x-ray images, and the maximum accuracy of 98.3% for COVID-RENet and COV-VGGNet models. Some other authors [ 75 , 76 ] reported a maximum accuracy of 94% for CT images using the DRENet model and lower accuracy of 76.47% for the JRIP model.…”

“… SqueezeNet Chest X ray Image 98 [ 71 ] ResNet18 98 [ 71 ] ResNet50 98 [ 71 ] DenseNet-121 98 [ 71 ] 3. Multicenter Osteoarthritis Study Chest X-ray Image 91 [ 72 ] 4. SDD300 model Chest X-ray Image 94.92 [ 73 ] 5.…”

A ResNet‐based approach for accurate radiographic diagnosis of knee osteoarthritis

“…Even though the KL scoring system has been widely used in clinics, some scholars still question the rationality of KL grading [26,27]. The method proposed by Karim et al, based on MRI (magnetic resonance imaging) and X-ray, can get 91.45% accuracy [28]. We will try to use the CT or MRI dataset to help improve the model accuracy during the training to get more depth information.…”

Deep learning-assisted knee osteoarthritis automatic grading on plain radiographs: the value of multiview X-ray images and prior knowledge

“…The development of deep learning (DL) has made it possible to establish accurate and rapid diagnostic methods (8). DL can allow the learning of features directly from the data, and it has recently revolutionized the field of medical image analysis by surpassing the conventional computer vision techniques that require the manual engineering of data representation methods (9). DL can undertake tasks such as image alignment, image recognition, image detection, image segmentation, and image classification and prediction, among others (10)(11)(12)(13)(14)(15)(16).…”

“…DL can undertake tasks such as image alignment, image recognition, image detection, image segmentation, and image classification and prediction, among others (10)(11)(12)(13)(14)(15)(16). These DL models include the well-known U-Net network and Residual Net (ResNet) network (9,17). In terms of medical image processing, the DL model can automatically extract the features of X-ray images through the learning of neural networks and classify the images to assist clinicians in diagnosis and treatment.…”