Design and Optimization of Rectangle shaped Biosensor for Hemoglobin Detection with Behavior Prediction in THz Regime

The level of Hemoglobin must be accurately and detected early to diagnose and monitor anemia and cardiovascular problems. Due to their molecular interaction with biological tissues, terahertz (THz) biosensors are ideal for such diagnostics. We designed a new sensor setup using a metasurface-encompassing rectangle shape for accurate hemoglobin levels. The rectangle shaped biosensor, meticulously constructed with specific dimensions, optimized by novel neural network framework the Dual Deep Residual Aggregation Multiscale Planet Convolutional Transformer Network which leverages Deep Residual Multiscale Convolutional Neural Network (DRMCNN) and Dual-Aggregation Transformer Network (DATN) structures significantly increases the surface area exposed to target compounds, thereby enhancing sensor sensitivity. Finally, an optimization approach optimizes the network parameters to achieve high levels of accuracy in the response while avoiding the debilitating effect of false positives and in synchronizing the exact resonant shifts that correspond to each change in biomarker values. The proposed THz biosensor achieves remarkable sensitivity, delivering early detection accuracy of approximately 99.8% through simulation using COMSOL Multiphysics. This system demonstrates a highly scalable approach for clinical applications offering a promising tool for timely and accurate monitoring of hemoglobin with a higher level of sensitivity, with a maximum value of 280 GHz/RIU, FOM of 3.10 and Q factor of 12.