Diabetes stands as a widespread and critical health concern on a global scale, presenting formidable obstacles to healthcare systems around the world. The increasing prevalence of this condition, coupled with its numerous complications, poses significant challenges for effective medical intervention and resource allocation. It is of paramount importance that diabetes is detected early and accurately to facilitate proper management strategies and prevent severe health consequences that can arise from uncontrolled diabetes. Unfortunately, current diagnostic methods often fall short in terms of precision and sensitivity when it comes to conducting large-scale screenings. This inadequacy is particularly evident in their ability to detect specific biomarkers associated with diabetes at low concentrations, which are crucial for accurate diagnosis.

Moreover, existing techniques tend to not take into consideration the intricate interrelations among various biomarkers, thereby further diminishing their effectiveness for reliable diabetes diagnostics. In response to these pressing limitations, the present study proposes an innovative solution in the form of an octagon-shaped terahertz (THz) photonic crystal fiber (PCF) biosensor specifically engineered for enhanced coordination in diabetes detection. This cutting-edge biosensor has been meticulously optimized through the application of a novel approach known as the Simplicial Causal Graph Dilated Botox Quaternion Convolutional Attention Network (SCG-DBQCAN). This advanced framework seamlessly integrates methodologies such as simplicial causal graph modeling and dilated quaternion convolutions along with attention mechanisms inspired by Botox technology—all aimed at achieving robust feature extraction and highly efficient classification of biomarkers.

The design of this biosensor significantly boosts light–matter interactions within its structure, resulting in exceptional sensitivity towards detecting diabetes-related biomarkers while simultaneously minimizing potential signal loss during measurement processes. Impressively, this state-of-the-art device boasts a remarkable detection accuracy rate of 99.9%, showcasing its high sensitivity even towards those low-concentration biomarkers that are often missed by traditional methods. Furthermore, its adaptability makes it suitable for various diagnostic contexts, presenting a game-changing solution for early detection of diabetes—a critical step toward mitigating long-term health risks associated with this chronic disease.