Application of machine learning and deep learning in diagnosis and treatment of anterior cruciate ligament and meniscus tears. New techniques and outcomes.Review

Background and aims: Innovative technology such as artificial intelligence (AI) is an essentialcomponent of orthopedic surgery. Deep learning is one of the branches of artificial intelligencethat can be used in the interpretation of magnetic resonance imaging (MRI) images and helpsurgeons in the diagnosis of injuries to the anterior cruciate ligament (ACL) and meniscus tears,and has a great impact on improving the treatment process. In this research, the use of artificialintelligence, especially deep learning and machine learning, in the diagnosis of anterior cruciateligament injury and meniscus tear is discussed.Method: For this purpose, articles registered in the last ۵ years in the PubMed database wereanalyzed. Articles related to the use of artificial intelligence in the diagnosis of ACL injuries andmeniscal tears were carefully reviewed with related terms. The terms included: ‘artificial intelligence’,‘deep learning’, ‘machine learning’, ‘knee injury’, ‘ACL’, ‘meniscus tear’. The paperswere further classified and reviewed according to the type of artificial intelligence algorithm used.Algorithms used in the desired topics were reviewed and the results were presented based on thefactors of accuracy and correctness of the used models.Overall, ۵۵ papers were reviewed: ۴۰ on deep learning and ۱۲ on machine learning and ۳ for both.Three papers also focused on the use of the two artificial intelligence algorithms.Results: Reviewing the studies showed that ۷۸% of the studies of the last ۵ years in the field ofartificial intelligence application to diagnosis of injuries in anterior cruciate ligament and meniscustears were using deep learning algorithms and only ۱۲% of the studies were using machinelearning algorithms and the accuracy of artificial intelligence algorithms in knee injuries predictionis in high range.Deep learning algorithms for ACL and meniscus tear detection have AUC (area under curve) andpredictive accuracy better than machine learning ones and are closer to specialists.Conclusion: The use of artificial intelligence in the diagnosis of knee injuries based on deeplearning algorithms is expanding and more than machine learning one, due to the growing numberof articles presented in this field. The review of the articles submitted shows that the use of AIalgorithms (deep learning and machine learning algorithms) gives acceptable results, but beforeusing on a large scale, it is necessary to mention some of its limitations.The limitations of deep learning algorithms are: the imbalance and homogenization of input data,the inability to generalize the models according to the technological level of imaging devices,and the lack of damage classification studies in this field. With the ability to expand artificialintelligence algorithms and the low cost of developing algorithms, it can be expected that thesealgorithms will widely help doctors in diagnosis in the near future.

کلیدواژه ها:

artificial intelligence ، deep learning ، machine learning ، knee injury ، ACL ، meniscus tear