Abstract

Contemporary individuals frequently face spine-related issues, which significantly impact their health and quality of life. However, traditional detection methods such as MRI and CT imaging entail high costs and radiation risks, limiting the screening and treatment of spine-related problems. This study leverages ubiquitous WiFi transceivers to collect WiFi Channel State Information (CSI) datasets representing three distinct spinal statuses. Employing a transformer-based neural network for data processing, we propose an efficient and cost-effective approach to assess spinal statuses, achieving a classification accuracy of 91%.