Full Confinement Loss Prediction in a Novel X-Shaped PCF-SPR Cancer Biosensor

Photonic crystal fibers (PCFs) play a key role in modern optical technologies owing to their exceptional capabilities in sensing, photonic devices, and communications. With the rapid progression of the Internet of Things (IoT), the demand for highly sensitive optical fiber sensors has become increasingly critical. In this work, a new X-shaped PCF biosensor based on surface plasmon resonance (SPR) is suggested for the ultra-sensitive detection of cancer cells in blood samples. The sensor’s optical and structural properties were analyzed using the finite element method (FEM), while the geometric parameters were systematically optimized using the Taguchi design of experiments (DOE) approach. The proposed design achieved a peak wavelength sensitivity of 2857.14 nm/RIU and an amplitude sensitivity of 69.46 RIU^-1 for breast cancer cells, while maintaining robust performance across other cancer types. The novelty of this work lies in the development of a multilayer perceptron (MLP) model capable of predicting the full confinement loss curve, rather than only the peak value, providing deeper insight into modal behavior and sensor performance. The MLP model achieved excellent accuracy with a correlation coefficient r = 0.9984 and a coefficient of determination R² = 0.9954. This study demonstrates the effective integration of FEM simulations, DOE-based optimization, and machine learning, paving the way for real-time, highly sensitive, and IoT-compatible biomedical diagnostics.