Use of artificial intelligence in body exercise recommendation in scoliosis rehabilition

The aim of this study is to offer a solution that will provide convenience to the health sector and patients. In the study, the scoliosis disease, which has been increasing in the number of patients in recent years, and its treatment process were examined. An artificial intelligence model was created in order to provide convenience to scoliosis patients and physiotherapists during the treatment process of the disease. In the treatment process of scoliosis patients who do not anticipate surgery, the patient is first examined by an orthopedic specialist and a radiological image is requested. The doctor refers patients for whom surgery is not planned to physical therapy and rehabilitation physicians. After the examination by the physical therapy and rehabilitation physician, the patient goes to the physiotherapist with their prescription. Depending on the workload of the physiotherapist, they may see the movements one-on-one or may have to listen to the exercise program and then return. Most patients cannot keep the exercise program in mind and may cause the disease to progress due to irregular exercise. An artificial intelligence model was developed as a solution to this situation. The developed model distinguishes 4 different types of the disease according to the uploaded radiological image and sends the video link of the exercise program suitable for this type to the patient's mobile number. 1297 radiological images were used in the model. The developed model was established with Resnet-50 architecture. The model achieved 89.30% accuracy, 89.45% precision, 89.18% sensitivity, and 89.44% F-1 score. The loss value of the model was reduced to 0.0555. The study aims to facilitate the work of patients and to prevent congestion in the hospital.