176 Views

609 - Adam Optimization Algorithm in Machine Learning for Telemedicine Rehabilitation Models in Brain Injury Care

Thursday, February 27, 2025

5:00 PM - 6:30 PM MST

Presenting Author(s)

OH

Oles Hospodarskyy, MBA

MBA
Lviv Polytechnic National University
Ternopil, Ternopil's'ka Oblast', Ukraine

Co-Author(s)

AH

Andriy Hospodarskyy, MD

As. Professor, General Surgery Department
Ternopil National Medical University
Ternopil, Ternopil's'ka Oblast', Ukraine
AT

Andriy Tsvyakh, PhD

Professor
Ternopil National Medical University
Ternopil, Ternopil's'ka Oblast', Ukraine

Objectives: Machine learning and artificial intelligence are significant areas of interest in both contemporary science and medical care. The algorithm's speed depends on the size of the dataset, the number of model parameters, and the number of iterations. The algorithm selects a smaller learning rate for frequently updated parameters and a larger one for parameters corresponding to rare features. Analysis of the experiment results conducted on various datasets, indicates a significant advantage of the Adam optimization algorithm.

Design: The study analyzed the medical records of patients in the retrospective group (186 cases of upper limb injuries as a component of brain injuries) and 62 patients in the main group who additionally underwent rehabilitation using the AI telemedicine model. To measure the effectiveness and universality of the AI model, we compared it with the traditional rehabilitation group.

Results: The retrospective group consisted of 135 men (72.58%) and 51 women (27.42%), while the main group comprised 44 men (70.97%) and 18 women (29.03%). Rehabilitation in the retrospective group included early functional mobilization, edema treatment, and physiotherapy for three weeks. The main group additionally used an AI-based telemedicine model with a portable device for real-time monitoring and exercise management. Results showed statistically significant improvements in the daily duration of physical exercises and a reduction in visits to medical facilities in the main group, with higher scores for functional recovery. The AI-based telemedicine model, allowing real-time monitoring, led to better functional recovery and reduced use of medical services compared to traditional methods.

Conclusions:

The use of AI-based telemedicine technologies can help a larger number of patients by providing opportunities for telerehabilitation in remote areas, improving the quality of life for those affected. Experiments on various datasets showed that the AI telemedicine algorithm is capable of achieving good results across a wide range of machine learning tasks in healthcare.