Abstract
Here, we introduce an AI-based system specifically developed for efficient rheumatoid arthritis (RA) diagnostics and patient management. Early RA diagnosis is important, as its symptomatological mimicry with different arthritis, and treatment of the disease requires an effective control. To address this limitation, we propose a system leveraging the ATG-LSMSA model to learn patient text data from online sources and generate clinically relevant responses. Finally, a feature consisting of symptoms is fed to an Adaptable Nested Dilated Temporal Convolution Network (AND-TCN) which classifies RA severity as low, medium or high. Tailored recommendations — the system based on its information will give you tips for dietary interventions and exercise suggestions. A Revised Arbitrary Variable WO (RAV-WO) algorithm is then proposed to optimize the parameters of several models, decrease prediction errors and enhance the classification performance. Using a chatbot-assisted framework would allow for ongoing surveillance of RA activity, quality of life, and level of functional impairment status with appropriate recommendations made as indicated. It is evaluated and also compared with the traditional methods to show that there are better results in terms of almost all the metrics. This single center, integrated AI strategy provides a pragmatic support for patient self-management, and could serve as supporting communication and earliest signal of intervention, adding value in extending the clinical decision-making process on chronic rheumatologic conditions.
Keywords: Disease Classification, Deep Learning, Personalized Recommendations. Rheumatoid Arthritis, Response Prediction.