Performance Prediction of a Highly Sensitive Optimized PCF-SPR Biosensor for Cancer Cell Detection Using MLP-Based ANN Model

Recent developments in surface plasmon resonance (SPR) sensors based on photonic crystal fiber (PCF) technology have greatly enhanced sensitivity. Nevertheless, existing configurations often face challenges related to design complexity and real-time application optimization. In this study, we present a straightforward and highly sensitive circular PCF-SPR biosensor tailored for the detection of early-stage cancer cells, with efficient performance across both visible and near-infrared spectrums. Using the Taguchi ${L} _{{27}}$ (35) orthogonal array approach, we optimized five critical structural parameters: air hole diameters, pitch, and the thickness of the gold (Au) and titanium dioxide (TiO2) layers. The resulting sensor achieved a remarkable spectral sensitivity of 18 000 nm/RIU and an amplitude sensitivity of 681.655 RIU-1. Designed to function across a wide refractive index (RI) range (1.29–1.40), the biosensor is particularly effective for identifying cervical, blood, and skin cancer cells, reaching a peak sensitivity of 7500 nm/RIU for cervical cancer detection. In addition, performance prediction utilized machine learning techniques, specifically multiple linear regression (MLR) and multi-layer perceptron artificial neural network (MLP-ANN) models. The MLP-ANN model demonstrated superior predictive accuracy, highlighting the potential of artificial intelligence (AI) in optimizing biosensor configurations for advanced cancer diagnostics.