Automated detection and classification of shoulder arthroplasty models using deep learning

Yi, Paul H.; Kim, Tae Kyung; Wei, Jinchi; Li, Xinning; Hager, Gregory D.; Sair, Haris I.; Fritz, Jan

doi:10.1007/s00256-020-03463-3

Cited by 41 publications

(29 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Deep learning has been met with enthusiasm and excitement by radiologists, as convolutional neural networks (DCNNs) have demonstrated the ability to perform radiologic tasks approaching or exceeding the levels of performance of expert radiologists in a variety of tasks [1][2][3][4]. Specifically, in musculoskeletal radiology, DCNNs have demonstrated wide-ranging utility for interpretation of radiographs for orthopedic trauma and implants [5][6][7], magnetic resonance imaging (MRI) for evaluation of internal derangement of the knee [8][9][10], and computed tomography (CT) for segmentation of pelvic muscles, fat, and bone [11].…”

Section: Introductionmentioning

confidence: 99%

Can AI distinguish a bone radiograph from photos of flowers or cars? Evaluation of bone age deep learning model on inappropriate data inputs

Arun

Hafezi-Nejad

et al. 2021

Skeletal Radiol

Self Cite

View full text Add to dashboard Cite

Objective To evaluate the behavior of a publicly available deep convolutional neural network (DCNN) bone age algorithm when presented with inappropriate data inputs in both radiological and non-radiological domains. Methods We evaluated a publicly available DCNN-based bone age application. The DCNN was trained on 12,612 pediatric hand radiographs and won the 2017 RSNA Pediatric Bone Age Challenge (concordance of 0.991 with radiologist groundtruth). We used the application to analyze 50 left-hand radiographs (appropriate data inputs) and seven classes of inappropriate data inputs in radiological (i.e., chest radiographs) and non-radiological (i.e., image of street numbers) domains. For each image, we noted if (1) the application distinguished between appropriate and inappropriate data inputs and (2) inference time per image. Mean inference times were compared using ANOVA. Results The 16Bit Bone Age application calculated bone age for all pediatric hand radiographs with mean inference time of 1.1 s. The application did not distinguish between pediatric hand radiographs and inappropriate image types, including radiological and non-radiological domains. The application inappropriately calculated bone age for all inappropriate image types, with mean inference time of 1.1 s for all categories (p = 1). Conclusion A publicly available DCNN-based bone age application failed to distinguish between appropriate and inappropriate data inputs and calculated bone age for inappropriate images. The awareness of inappropriate outputs based on inappropriate DCNN input is important if tasks such as bone age determination are automated, emphasizing the need for appropriate oversight at the data input and verification stage to avoid unrecognized erroneous results.

show abstract

Section: Introductionmentioning

confidence: 99%

Can AI distinguish a bone radiograph from photos of flowers or cars? Evaluation of bone age deep learning model on inappropriate data inputs

Arun

Hafezi-Nejad

et al. 2021

Skeletal Radiol

Self Cite

View full text Add to dashboard Cite

show abstract

“…We also examined our model for an open-world configuration and achieved the best results compared to the other deep models, which demonstrates the generalizability of our approach. As reported in previous research [11,45,46], the usage of computer-based algorithms can do better to identify shoulder arthroplasty implants compared to medical experts, which can reduce the risk of delayed operations, perioperative morbidity, and overuse of resources due to lack of correct identification of shoulder arthroplasty implants. Based on these motivations, previous research [11,45,46] has also studied the computerbased algorithms for the identification of shoulder implants.…”

Section: Discussionmentioning

confidence: 94%

“…They used the transfer-learning method and did not involve an open-world setting to address real-world problems. In [46], DL was used for the binary classification of shoulder implant models. They used a transfer learning approach and fine-tuned ResNet-18 for binary classification of the existence of arthroplasty implants.…”

Section: Related Workmentioning

confidence: 99%

“…These results confirm that it is visually difficult to discriminate the data of C1 from Figure 1a, and we can tell that there exist the differences among those intra models. The performances of various state-of-the-art methods [38,41,45,46,69,70] were compared with those of our approach. Table 11 shows the performance comparisons by the state-of-the-art methods and the proposed method without data augmentation, and ResNet-50 [38] outperformed the other methods.…”

Section: Comparison Of Proposed Dre-net With the Subjective Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

Artificial Intelligence-Based Recognition of Different Types of Shoulder Implants in X-ray Scans Based on Dense Residual Ensemble-Network for Personalized Medicine

Sultan

Owais

Park

et al. 2021

JPM

View full text Add to dashboard Cite

Re-operations and revisions are often performed in patients who have undergone total shoulder arthroplasty (TSA) and reverse total shoulder arthroplasty (RTSA). This necessitates an accurate recognition of the implant model and manufacturer to set the correct apparatus and procedure according to the patient’s anatomy as personalized medicine. Owing to unavailability and ambiguity in the medical data of a patient, expert surgeons identify the implants through a visual comparison of X-ray images. False steps cause heedlessness, morbidity, extra monetary weight, and a waste of time. Despite significant advancements in pattern recognition and deep learning in the medical field, extremely limited research has been conducted on classifying shoulder implants. To overcome these problems, we propose a robust deep learning-based framework comprised of an ensemble of convolutional neural networks (CNNs) to classify shoulder implants in X-ray images of different patients. Through our rotational invariant augmentation, the size of the training dataset is increased 36-fold. The modified ResNet and DenseNet are then combined deeply to form a dense residual ensemble-network (DRE-Net). To evaluate DRE-Net, experiments were executed on a 10-fold cross-validation on the openly available shoulder implant X-ray dataset. The experimental results showed that DRE-Net achieved an accuracy, F1-score, precision, and recall of 85.92%, 84.69%, 85.33%, and 84.11%, respectively, which were higher than those of the state-of-the-art methods. Moreover, we confirmed the generalization capability of our network by testing it in an open-world configuration, and the effectiveness of rotational invariant augmentation.

show abstract

“…Recently, two separate studies have demonstrated the efficiency of CNNs in detecting and even differentiating shoulder implants by the manufacturer [21,22]. However, these studies relied on frontal view radiographs, whereas our data included a broad variety of views to allow detection of surgical implants independent of image settings used for acquisition.…”

Section: Discussionmentioning

confidence: 99%

Deep learning for accurately recognizing common causes of shoulder pain on radiographs

et al. 2021

View full text Add to dashboard Cite

Objective Training a convolutional neural network (CNN) to detect the most common causes of shoulder pain on plain radiographs and to assess its potential value in serving as an assistive device to physicians. Materials and methods We used a CNN of the ResNet-50 architecture which was trained on 2700 shoulder radiographs from clinical practice of multiple institutions. All radiographs were reviewed and labeled for six findings: proximal humeral fractures, joint dislocation, periarticular calcification, osteoarthritis, osteosynthesis, and joint endoprosthesis. The trained model was then evaluated on a separate test dataset, which was previously annotated by three independent expert radiologists. Both the training and the test datasets included radiographs of highly variable image quality to reflect the clinical situation and to foster robustness of the CNN. Performance of the model was evaluated using receiver operating characteristic (ROC) curves, the thereof derived AUC as well as sensitivity and specificity. Results The developed CNN demonstrated a high accuracy with an area under the curve (AUC) of 0.871 for detecting fractures, 0.896 for joint dislocation, 0.945 for osteoarthritis, and 0.800 for periarticular calcifications. It also detected osteosynthesis and endoprosthesis with near perfect accuracy (AUC 0.998 and 1.0, respectively). Sensitivity and specificity were 0.75 and 0.86 for fractures, 0.95 and 0.65 for joint dislocation, 0.90 and 0.86 for osteoarthrosis, and 0.60 and 0.89 for calcification. Conclusion CNNs have the potential to serve as an assistive device by providing clinicians a means to prioritize worklists or providing additional safety in situations of increased workload.

show abstract

Automated detection and classification of shoulder arthroplasty models using deep learning

Cited by 41 publications

References 31 publications

Can AI distinguish a bone radiograph from photos of flowers or cars? Evaluation of bone age deep learning model on inappropriate data inputs

Can AI distinguish a bone radiograph from photos of flowers or cars? Evaluation of bone age deep learning model on inappropriate data inputs

Artificial Intelligence-Based Recognition of Different Types of Shoulder Implants in X-ray Scans Based on Dense Residual Ensemble-Network for Personalized Medicine

Deep learning for accurately recognizing common causes of shoulder pain on radiographs

Contact Info

Product

Resources

About