Machine Learning Approach for SPR based Photonic Crystal Fiber Sensor for Breast Cancer Cells Detection

Abstract

This article presents a highly sensitive gold/TiO2 coated photonic crystal fiber (PCF) biosensor for sensing the breast cancer cells. The surface plasmon resonance (SPR) mechanism is achieved by coating the hybrid gold/TiO2 layer over the PCF. The simulation and numerical analysis are done by the finite element method. The normal and malignant cells are detected by measuring the resonance wavelength shift which helps to estimate the wavelength sensitivity and resolution. In results, the maximum wavelength sensitivity in x-polarization mode is achieved as 11,034 nm/RIU for breast cancer MCF-7 cells and 9,285 nm/RIU in y-polarization mode with very low resolution in the range of $10^{-6}$RIU. In addition, the impact of the presence of metallic thin film, air holes pitch and RI of the analytes is observed and optimized through supervised machine learning approach with low mean square error (mse). With enhanced sensing performance of the proposed sensor, it can be used as fast, efficient and low-cost cancer and other blood related disease detection device.