عنوان (فارسي) بررسي عوامل موثر بر شكست مهره مجاور مهره تقويت‌شده با سيمان به كمك المان محدود و ارائه پارامترهاي بهينه براي پيشگيري از شكست به كمك تكنيك يادگيري‌ماشين عنوان (انگليسي)eva‎luating of mechanism of fractures of adjacent vertebrae to augmented vertebra and predicting optimal parameters with Machine learning

2/15/2024 12:00:00 AM

Clinical outcomes of Vertebroplasty due to simple surgery technique, low complications, and immediate relieving pain are encouraging. However, 25% of patients experience following fractures after treatment. Moreover, 50-67% of these fractures occur in adjacent augmented vertebras. Therefore, the potential risk for later compression fractures after vertebroplasty declines the benefits of vertebroplasty. However, the relationship between vertebroplasty and new fractures has not been approved and there is not any consensus on the main reason for this happening. But, if vertebroplasty varies the mechanism of the load transition in the adjacent vertebra, it can be considered as an agent of adjacent vertebra fracture. The result of this study shows that Vertebroplasty is a beneficial approach and the negative distinctive outcomes are due to the different conditions of surgeries and patients. Thus, some factors determine the result of this surgery. By reviewing different studies, factors that are extracted from them are as follows: volume of cement, distribution of cement, stiffness of cement, degree of osteoporosis, the field of cement, method of cement injection, degree of intervertebral disc degeneration, BMD, the morphology of vertebra, area of vertebra and properties of the disc. Developing models by utilizing Machine Learning techniques and Classification algorithms based on collecting data according to the aforementioned factors leads to the development of models that are able to predict fracture at the adjacent level of the augmented vertebra, and by that, the determination of parameters which do not lead to adjacent fracture, preventing adjacent fractures after vertebroplasty is possible. In this study, models are developed based on K-nearest neighbors, Decision tree, Support vector machine, and Logistic regression methods that the Logistic regression model and Decision tree model reached 94% of precision.

ورتبروپلاستي , شكست فشاري , يادگيري ماشين , شكست مهره مجاور , الگوريتم دسته‌بندي

Vertebroplasty , Compression fracture , Machine learning , Adjacent vertebra fracture , Classification algorithm

Maede Hasanpour Emam moghaddam

Mohammad Haghpanahi