Automating Abnormality Detection in Musculoskeletal Radiographs through Deep Learning

Mehr, Goodarz

doi:10.48550/arxiv.2010.12030

Cited by 3 publications

(11 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A single CNN model is used in the one-stage method in ref. 9 , and our framework has better accuracy, F1 score and Cohen's Kappa.…”

Section: More Discussionmentioning

confidence: 81%

“…In addition to the above work, Mehr 9 proposed a method to automatically detect abnormalities in musculoskeletal X-rays by using MuRAD (musculoskeletal X-ray abnormality detection tool). MuRAD used a convolutional neural network (CNN) to accurately predict whether skeletal X-rays are abnormal and used a category activation map (CAM) to locate abnormalities in images.…”

Section: Methods Based On Deep Learningmentioning

confidence: 99%

“…Based on the above difficulties, this paper proposes a two-stage bone classification and bone abnormality detection framework for seven types of bones instead of only one as in the literature. 3,4,7,9,[13][14][15][16][17][18] In the first stage, the framework uses ResNeXt50 to classify based on seven kinds of bone X-rays, combined with (1) augmentation; (2) lr-attenuation; (3) pre-trained model; (4) focal loss, which can accurately distinguish the bone types of upper limb areas (shoulder, humerus, forearm, elbow, wrist, hand, and finger), can effectively eliminate the impact of the difference between bone types on the accuracy of final abnormality detection. In the second stage, abnormality detection is carried out based on effectively distinguishing seven bone types.…”

Section: Introductionmentioning

confidence: 99%

“…The computer vision method of deep learning can extract the features that are difficult to be detected by clinicians from the X-ray as a whole and can distinguish bone abnormalities with huge shape differences or even variant types.This ability can grow effectively with the data. 3,4,7,9,[13][14][15][16][17][18] Therefore, the computer vision method based on deep learning has successfully assisted clinicians in image medical diagnosis. 3,6,7,9,[18][19][20][21] However, the computer vision method using deep learning still has some limitations: (1) due to the influence of the size and shape of the dataset, it is difficult for ordinary medical institutions to collect enough bone X-rays with different shape types;…”

Section: Introductionmentioning

confidence: 99%

“…3,4,7,9,[13][14][15][16][17][18] Therefore, the computer vision method based on deep learning has successfully assisted clinicians in image medical diagnosis. 3,6,7,9,[18][19][20][21] However, the computer vision method using deep learning still has some limitations: (1) due to the influence of the size and shape of the dataset, it is difficult for ordinary medical institutions to collect enough bone X-rays with different shape types;…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Enhanced deep residual network for bone classification and abnormality detection

Guo

2022

Medical Physics

View full text Add to dashboard Cite

Purpose A two‐stage deep learning framework for bone classification and abnormality detection is proposed based on X‐rays. The primary focus is on improving the speed of orthopedic disease diagnosis and helping physicians reduce the probability of false diagnoses. Methods The method is based on two stages. In the first stage, one classifier with ResNeXt50 as the backbone is used to classify bones to eliminate the effect of bone type differences on abnormality detection. In the second stage, seven anomaly detectors are trained based on each type of training data. The seven detectors tested the seven results of the first stage, respectively. Pretrained models, data augmentation, focal loss, label smoothing loss, LR‐attenuation and early stopping are used to improve performance and reduce the risk of overfitting. Results Experiments are based on the largest dataset for bone abnormality detection, MURA. In the first stage for bone classification, we got an accuracy of 96.69%, a sensitivity of 96.69%, a specificity of 99.46%, and an F1 score of 96.42%. In the second stage for abnormality detection, we got an accuracy of 84.15%, a sensitivity of 84.15%, a specificity of 87.50%, an F1 score of 84.10%, a Cohen's Kappa of 0.72, and an area under the ROC curve (AUC) score of 0.90. Conclusions Compared with other excellent convolutional neural network models, the framework's effectiveness was verified with better accuracy, sensitivity, specificity, F1 score, Cohen's Kappa score, and AUC score for bone classification and abnormality detection.

show abstract

“…A single CNN model is used in the one-stage method in ref. 9 , and our framework has better accuracy, F1 score and Cohen's Kappa.…”

Section: More Discussionmentioning

confidence: 81%

Section: Methods Based On Deep Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Enhanced deep residual network for bone classification and abnormality detection

Guo

2022

Medical Physics

View full text Add to dashboard Cite

show abstract

LGAG-Net: Lesion-Guided Adaptive Graph Network for Bone Abnormality Detection From Musculoskeletal Radiograph

Liao,

Li,

Peng

2024

IEEE Access

View full text Add to dashboard Cite

Musculoskeletal abnormality is routinely presented in tissues and organs of the human locomotor system across the life course, and it is essential to detect musculoskeletal abnormality in X-rays. However, it is difficult to diagnose musculoskeletal abnormality from Radiographs due to the following issues: 1) There are other interfering organ tissues in the complicated backgrounds; 2)The MURA dataset contains seven different musculoskeletal radiographs, which makes general convolution neural networks unable to model the weird relationship between them. To address such problems, a Lesion-Guided Adaptive Graph Network (LGAG-Net) is proposed for bone abnormality detection from musculoskeletal radiograph, where the Lesion-guided Recurrent Feature Sampling (LRFS) module is first designed to localize the corresponding musculoskeletal abnormality regions, and then the Adaptive Graphsage Attention (AGA) module is developed to perform bone abnormality detection on the located musculoskeletal abnormality regions. Experiments on MURA dataset show that the proposed LGAG-Net can achieve an accuracy of 87.81% and Cohen Kappa statistic of 0.868, which outperforms the state-of-the-art methods, assisting the radiologists to rapidly estimate the physical development in patients.

show abstract

A Comparative Study of Deep Convolutional Neural Networks for Musculoskeletal X-Ray Images

2023

Proceedings of 2023 the 13th International Workshop on Computer Science and Engineering

View full text Add to dashboard Cite

Technological achievements in science and technology, especially deep learning models have brought about a promising overall performance in many medical image evaluation tasks. Evaluating abnormal conditions from Body part X-rays images is such a need for radiological examination. To fulfill the need of automated detecting of Musculoskeletal conditions, this paper experiments and explores the state-ofthe-art deep neural networks which are most commonly applied for image processing. Some of the typical Convolutional neural networks are experimented and evaluated each of the model's classification performance using bone X-ray images. In this study, a typical CNN architecture, CNN architecture with Adabound optimizer, VGG16, ResNet50 and DenseNet architetures are experimented and evaluated. The effectiveness of these Deep CNN models are accessed and compared on the Cohen's kappa statistic and classification accuracy. DenseNet169 Model outperformed the other pre-trained models tested in this study, with the greatest Training Accuracy of 97.98%. However, the standard CNN model with Adam optimizer has a little higher kappa score than the DenseNet169 model. The DenseNet169 Model obtained roughly 20% in terms of Train and Test Loss, which is less than the training loss compared to the other pre-trained CNN models.

show abstract

Automating Abnormality Detection in Musculoskeletal Radiographs through Deep Learning

Cited by 3 publications

References 25 publications

Enhanced deep residual network for bone classification and abnormality detection

Enhanced deep residual network for bone classification and abnormality detection

LGAG-Net: Lesion-Guided Adaptive Graph Network for Bone Abnormality Detection From Musculoskeletal Radiograph

A Comparative Study of Deep Convolutional Neural Networks for Musculoskeletal X-Ray Images

Contact Info

Product

Resources

About