Pediatric cancer refractive-index sensing with a deep learning-enhanced far-infrared highly sensitive spoof surface plasmon polariton metasurface absorber

Abstract

Pediatric cancers remain a primary cause of childhood mortality globally, yet early detection is frequently hindered by overlapping clinical symptoms and the limitations of conventional diagnostic methods; we propose a highly sensitive spoof surface plasmon polaritons (SSPP)-based refractive index (RI) sensor in the far-infrared regime that consists of novel tungsten-quartz-tungsten configuration. It comprises a cross-shaped resonator to attain polarization-insensitive performance. A thorough analysis of the sensor’s geometry is conducted using CST Microwave Studio to enhance its efficacy. The findings reveal nearly perfect absorption at 183.25 μ m , with absorption above 90 % from 183.14–183.36 μ m . This narrowband characteristic is primarily attributed to the excitation of high magnetic polariton resonances and strong magnetic field confinement within the structure. The underlying mechanism is thoroughly analyzed using electric field, surface current, impedance matching theory and an equivalent circuit model (ECM). Furthermore, by introducing analytes with varying RI to mimic healthy and various cancer cells, the proposed structure leverages the dependence of its resonance frequencies on RI to effectively distinguish among different cancer cell types, including leukemia, glioma, blastoma, Wilms tumor, and lymphoma. The full-wave numerical simulations exhibit a maximum sensitivity of 53,962 nm / RIU , a quality factor (Q-factor) of 374.98 , and a figure-of-merit (FoM) of 84,800 RIU − 1 for the proposed design. To further improve diagnostic precision, a deep-learning (DL) analysis framework is utilized to accurately distinguish cancerous tissues. The optimized algorithm demonstrates superior performance, achieving a training accuracy of 97.8 % and a validation accuracy of 93.5 % . Owing to high sensitivity and improved detection precision of DL-driven spectral analysis, this work provides a revolutionary foundation for real-world pediatric cancer early-stage screening.