A novel image-based machine learning model with superior accuracy and predictability for knee arthroplasty loosening detection and clinical decision making

Enabling Personalized Medicine in Orthopaedic Surgery Through Artificial Intelligence: A Critical Analysis Review

» The application of artificial intelligence (AI) in the field of orthopaedic surgery holds potential for revolutionizing health care delivery across 3 crucial domains: (I) personalized prediction of clinical outcomes and adverse events, which may optimize patient selection, surgical planning, and enhance patient safety and outcomes; (II) diagnostic automated and semiautomated imaging analyses, which may reduce time burden and facilitate precise and timely diagnoses; and (III) forecasting of resource utilization, which may reduce health care costs and increase value for patients and institutions.» Computer vision is one of the most highly studied areas of AI within orthopaedics, with applications pertaining to fracture classification, identification of the manufacturer and model of prosthetic implants, and surveillance of prosthesis loosening and failure.» Prognostic applications of AI within orthopaedics include identifying patients who will likely benefit from a specified treatment, predicting prosthetic implant size, postoperative length of stay, discharge disposition, and surgical complications. Not only may these applications be beneficial to patients but also to institutions and payors because they may inform potential cost expenditure, improve overall hospital efficiency, and help anticipate resource utilization.» AI infrastructure development requires institutional financial commitment and a team of clinicians and data scientists with expertise in AI that can complement skill sets and knowledge. Once a team is established and a goal is determined, teams (1) obtain, curate, and label data; (2) establish a reference standard; (3) develop an AI model; (4) evaluate the performance of the AI model; (5) externally validate the model, and (6) reinforce, improve, and evaluate the model's performance until clinical implementation is possible.» Understanding the implications of AI in orthopaedics may eventually lead to wide-ranging improvements in patient care. However, AI, while holding tremendous promise, is not without methodological and ethical limitations that are essential to address. First, it is important to ensure external validity of programs before their use in a clinical setting. Investigators should maintain high quality data records and registry surveillance, exercise caution when evaluating others' reported AI applications, and increase transparency of the methodological conduct of current models to improve external validity and avoid propagating bias. By addressing these challenges and responsibly embracing the potential of AI, the medical field may eventually be able to harness its power to improve patient care and outcomes.

Reliability of artificial intelligence in predicting total knee arthroplasty component sizes: a systematic review

PURPOSE

This systematic review aimed to investigate the reliability of AI predictive models of intraoperative implant sizing in total knee arthroplasty (TKA).

METHODS

Four databases were searched from inception till July 2023 for original studies that studied the reliability of AI prediction in TKA. The primary outcome was the accuracy ± 1 size. This review was conducted per PRISMA guidelines, and the risk of bias was assessed using the MINORS criteria.

RESULTS

A total of four observational studies comprised of at least 34,547 patients were included in this review. A mean MINORS score of 11 out of 16 was assigned to the review. All included studies were published between 2021 and 2022, with a total of nine different AI algorithms reported. Among these AI models, the accuracy of TKA femoral component sizing prediction ranged from 88.3 to 99.7% within a deviation of one size, while tibial component sizing exhibited an accuracy ranging from 90 to 99.9% ± 1 size.

CONCLUSION

This study demonstrated the potential of AI as a valuable complement for planning TKA, exhibiting a satisfactory level of reliability in predicting TKA implant sizes. This predictive accuracy is comparable to that of the manual and digital templating techniques currently documented in the literature. However, future research is imperative to assess the impact of AI on patient care and cost-effectiveness.

LEVEL OF EVIDENCE III

PROSPERO registration number: CRD42023446868.

Automatic detection of Cryptosporidium in optical microscopy images using YOLOv5x: a comparative study

Here, a machine learning tool (YOLOv5) enables the detection of Cryptosporidium microorganisms using optical and phase contrast microscope images. The two databases were processed using 520 images (optical microscopy) and 1200 images (phase contrast microscopy). It used Python libraries to label, standardize the size, and crop the images to generate the input tensors to the YOLOv5 network (s, m, and l). It implemented two experiments using randomly initialized weights in optical and phase contrast microscope images. The other two experiments used the parameters for the best training time obtained before and after retraining the models. Metrics used to assess model accuracy were mean average accuracy, confusion matrix, and the F1 scores. All three metrics confirmed that the optimal model used the best epoch of optical imaging training and retraining with phase contrast imaging. Experiments with randomly initialized weights with optical imaging showed the lowest precision for Cryptosporidium detection. The most stable model was YOLOv5m, with the best results in all categories. However, the differences between all models are lower than 2%, and YOLOv5s is the best option for Cryptosporidium detection considering the differences in computational costs of the models.

A multi-branch network to detect post-operative complications following hip arthroplasty on X-ray images

Background: Postoperative complications following total hip arthroplasty (THA) often require revision surgery. X-rays are usually used to detect such complications, but manually identifying the location of the problem and making an accurate assessment can be subjective and time-consuming. Therefore, in this study, we propose a multi-branch network to automatically detect postoperative complications on X-ray images. Methods: We developed a multi-branch network using ResNet as the backbone and two additional branches with a global feature stream and a channel feature stream for extracting features of interest. Additionally, inspired by our domain knowledge, we designed a multi-coefficient class-specific residual attention block to learn the correlations between different complications to improve the performance of the system. Results: Our proposed method achieved state-of-the-art (SOTA) performance in detecting multiple complications, with mean average precision (mAP) and F1 scores of 0.346 and 0.429, respectively. The network also showed excellent performance at identifying aseptic loosening, with recall and precision rates of 0.929 and 0.897, respectively. Ablation experiments were conducted on detecting multiple complications and single complications, as well as internal and external datasets, demonstrating the effectiveness of our proposed modules. Conclusion: Our deep learning method provides an accurate end-to-end solution for detecting postoperative complications following THA.

A systematic review and meta-analysis comparing outcomes following total knee arthroplasty for rheumatoid arthritis versus for osteoarthritis

PURPOSE

Total knee arthroplasty (TKA) in patients with osteoarthritis (OA) are considered to be a successful procedure, but with little being known about outcomes in patients with rheumatoid arthritis (RA). The aim of this study was to compare the outcomes of TKA in patients with RA versus OA.

METHODS

Data were obtained from PubMed, Cochrane Library, EBSCO and Scopus for all available studies comparing the outcomes of THA in RA and OA patients (From January 1, 2000 to October 15, 2022). Outcomes of interest included infection, revision, venous thromboembolism (VTE), mortality, periprosthetic fractures, prosthetic loosening, length of stay, and satisfaction. Two reviewers independently assessed each study for quality and extracted data. The quality of the studies was scored using the Newcastle-Ottawa scale (NOS).

RESULTS

Twenty-four articles with a total 8,033,554 patients were included in this review. The results found strong evidence for increased risk of overall infection (OR = 1.61, 95% CI, 1.24-2.07; P = 0.0003), deep infection (OR = 2.06, 95% CI, 1.37-3.09; P = 0.0005), VTE (OR = 0.76, 95% CI, 0.61-0.93; P = 0.008), pulmonary embolism (PE) (OR = 0.84, 95% CI, 0.78-0.90; P<0.00001), periprosthetic fractures (OR = 1.87, 95% CI, 1.60-2.17; P<0.00001); and reasonable evidence for increased risk of deep venous thrombosis (DVT) (OR = 0.74, 95% CI, 0.54-0.99; P = 0.05), and length of stay (OR = 0.07, 95% CI, 0.01-0.14; P = 0.03) after TKA in patients with RA versus OA. There were no significant differences in superficial site infection (OR = 0.84,95% CI, 0.47-1.52; P = 0.57), revision (OR = 1.33,95% CI, 0.79-2.23; P = 0.28), mortality (OR = 1.16,95% CI, 0.87-1.55; P = 0.32), and prosthetic loosening (OR = 1.75, 95% CI, 0.56-5.48; P = 0.34) between the groups.

CONCLUSION

Our study demonstrated that patients with RA have a higher risk of postoperative infection, VTE, periprosthetic fracture, and lengths of stay, but did not increase revision rate, prosthetic loosening and mortality compared to patients with OA following TKA. In conclusion, despite RA increased incidence of postoperative complications, TKA should continue to be presented as an effective surgical procedure for patients whose conditions are intractable to conservative and medical management of RA.

Development of Hallux Valgus Classification Using Digital Foot Images with Machine Learning

Hallux valgus, a frequently seen foot deformity, requires early detection to prevent it from becoming more severe. It is a medical economic problem, so a means of quickly distinguishing it would be helpful. We designed and investigated the accuracy of an early version of a tool for screening hallux valgus using machine learning. The tool would ascertain whether patients had hallux valgus by analyzing pictures of their feet. In this study, 507 images of feet were used for machine learning. Image preprocessing was conducted using the comparatively simple pattern A (rescaling, angle adjustment, and trimming) and slightly more complicated pattern B (same, plus vertical flip, binary formatting, and edge emphasis). This study used the VGG16 convolutional neural network. Pattern B machine learning was more accurate than pattern A. In our early model, Pattern A achieved 0.62 for accuracy, 0.56 for precision, 0.94 for recall, and 0.71 for F1 score. As for Pattern B, the scores were 0.79, 0.77, 0.96, and 0.86, respectively. Machine learning was sufficiently accurate to distinguish foot images between feet with hallux valgus and normal feet. With further refinement, this tool could be used for the easy screening of hallux valgus.

Detection of Prosthetic Loosening in Hip and Knee Arthroplasty Using Machine Learning: A Systematic Review and Meta-Analysis

Background: prosthetic loosening after hip and knee arthroplasty is one of the most common causes of joint arthroplasty failure and revision surgery. Diagnosis of prosthetic loosening is a difficult problem and, in many cases, loosening is not clearly diagnosed until accurately confirmed during surgery. The purpose of this study is to conduct a systematic review and meta-analysis to demonstrate the analysis and performance of machine learning in diagnosing prosthetic loosening after total hip arthroplasty (THA) and total knee arthroplasty (TKA). Materials and Methods: three comprehensive databases, including MEDLINE, EMBASE, and the Cochrane Library, were searched for studies that evaluated the detection accuracy of loosening around arthroplasty implants using machine learning. Data extraction, risk of bias assessment, and meta-analysis were performed. Results: five studies were included in the meta-analysis. All studies were retrospective studies. In total, data from 2013 patients with 3236 images were assessed; these data involved 2442 cases (75.5%) with THAs and 794 cases (24.5%) with TKAs. The most common and best-performing machine learning algorithm was DenseNet. In one study, a novel stacking approach using a random forest showed similar performance to DenseNet. The pooled sensitivity across studies was 0.92 (95% CI 0.84-0.97), the pooled specificity was 0.95 (95% CI 0.93-0.96), and the pooled diagnostic odds ratio was 194.09 (95% CI 61.60-611.57). The I2 statistics for sensitivity and specificity were 96% and 62%, respectively, showing that there was significant heterogeneity. The summary receiver operating characteristics curve indicated the sensitivity and specificity, as did the prediction regions, with an AUC of 0.9853. Conclusions: the performance of machine learning using plain radiography showed promising results with good accuracy, sensitivity, and specificity in the detection of loosening around THAs and TKAs. Machine learning can be incorporated into prosthetic loosening screening programs.