Detection of hemoglobin concentration to determine anaemia using AI/ML-based SPR fiber-optic biosensor

This paper proposes a Surface Plasmon Resonance (SPR) biosensor designed to detect hemoglobin (Hb) levels in blood, facilitating the diagnosis of anaemia. The sensor uses a silver (Ag) and cadmium selenide (CdSe) coating, leveraging wavelength interrogation to achieve high sensitivity and resolution. A thin layer of Ag (45nm) and a CdSe film (15nm) serve as sensing layers. Using AI/ML techniques, we identify optimized parameters and calculate Mean Square Error (MSE) and R-squared (R²). Using Bayesian Regularization Artificial Neural Network (BRANN) we minimized the loss function ‘F’ by decaying weight of ANN and solved the problem of overfitting due to fewer number of samples. Also, we obtained best values of MSE and R² compared to others models i.e., 0.9647 and 0.9992 respectively. The maximum sensitivity is achieved in RI range of 1.32919–1.34919, indicating precise detection capability with value of around 6000 nm/RIU. It reaches a highest resolution of 1.6×10⁻⁵ RIU, enabling the detection of alterations in hemoglobin concentration as low as 6.1025 g/L. We have also calculated concentration sensitivity 0.9832 nm/g/L. This study marks a crucial advancement in creating sensitive, label-free optical biosensors for use in medicine.