Application of artificial intelligence and natural language processing in chronic low back pain: a domain review

Background and aims: Chronic low back pain is a symptom that may be caused by severaldiseases and is currently the leading cause of disability worldwide. Usually, when it lasts morethan twelve weeks, it is caused by a large set of diseases such as degenerative disc diseases, discherniation, spondyl arthritis and spondylolisthesis. In recent years, the most groundbreaking technologiesin LBP care have been explored, including artificial intelligence and computer science,natural language processing. In this study, the impact of the most important technologies, artificialintelligence, in chronic back pain has been investigated.Method: This study has been reviewed with a domain review method, studies related to the use ofartificial intelligence in the diagnosis and treatment of LBP. For this purpose, the reliable PubMeddatabase was searched. The search strategy was set as a combination of the following keywords:“artificial intelligence”, “machine learning”, “deep learning”, “natural language processing”,“pain”, “back”. Extracted articles are summarized in terms of content.Results: Most recent methods use deep learning models instead of digital image processing techniques.The best methods for segmenting vertebrae, intervertebral discs, spinal canal, and backmuscles achieve a Sorensen-Dice score of more than ۹۰%, while studies focused on localizationand identification of structures show overall accuracies of more than ۸۰%.There are three main methods of NLP namely classification, annotation and prediction. Both ofthe first approaches concern the identification of a category (class) to which a document belongs,which is different in the case of NLP methods. In the classification approach, the system associatesa label to each test sample (patient records).A classification system may provide informationabout a diagnosis, as a computer-aided diagnosis system, that physicians may use to make decisions,for example, whether or not to operate on a patient. Also, healthcare providers may usesuch a system to improve quality control, while researchers may use it to retrieve a large group ofpatients suffering from a particular disease and then perform some research analysis.Conclusion: Future advances in artificial intelligence are expected to increase the autonomy andreliability of systems, thus providing more effective tools for the diagnosis and treatment of LBP.Further studies on larger data sets are needed to better define the role of NLP in the care of patientswith spinal disorders.

کلیدواژه ها:

artificial intelligence ، natural language processing ، low back pain ، decision supportsystems ، computer vision