Scholarlink

<p>This paper describes the development of a Quantum-Tunneling-Assisted Plasmonic Nanostructure (QEPNS) incorporated into a Quantum-Tunneling-Assisted Neural Plasmonic Hybrid Architecture (QENPHA). The configuration enables sensitive and easy cancer biomarker detection that is ultra-sensitive. It was demonstrated by Finite-Difference Time-Domain (FDTD) simulations that charge delocalization in gold-silver bimetallic nanoparticles, which were 60&#xa0;nm in diameter and 3&#xa0;nm in suspension space, gave a 12.6x near-field enhancement. The optimized structure exhibited an optimally tuned localized surface plasmon resonance of (480–780) nm with a sensitivity of 810&#xa0;nm/RIU and a full width at half maximum (FWHM) of 13.2 ± 0.5&#xa0;nm and a Figure of Merit (FOM) of 61 leading to the estimated limit of detection (i.e. of the order of 10<sup>− 15</sup> M) at the femtomolar scale given by reference to experimentally determined resolution of the refractive index. Multispectral Dynamic Feature Resonance Model (MDFRM) recorded the dipole (520&#xa0;nm), quadrupole (630&#xa0;nm), and Fano (780&#xa0;nm) modes and retained 96.3% of the variance using PCA and VAE compression. The Hybrid Neural Architecture (CNN, BiLSTM, Attention) achieved a test accuracy of 98.3% and ROC AUC of 0.990 when using four cancer biomarkers. Explainability PlasmaSHAP and Grad-CAM have given a combined explainability of 0.89: the model’s attention was correlated to physical plasmonic hotspots. The experimental results were able to verify the theoretical predictions with an R<sup>2</sup> value of 0.996 and the coefficient of variation of less than 1.8 per cent, indicating great reproducibility and usability in real-time and label-free cancer diagnostics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging XAI for Interpretable Quantum Neuro-Plasmonic Resonance Analysis

  • 期刊论文

  • 发表日期：

    2026年7月28日

查看全文
  • Aravind Govindaram,
  • Jayaraman Indumathi

摘要

This paper describes the development of a Quantum-Tunneling-Assisted Plasmonic Nanostructure (QEPNS) incorporated into a Quantum-Tunneling-Assisted Neural Plasmonic Hybrid Architecture (QENPHA). The configuration enables sensitive and easy cancer biomarker detection that is ultra-sensitive. It was demonstrated by Finite-Difference Time-Domain (FDTD) simulations that charge delocalization in gold-silver bimetallic nanoparticles, which were 60 nm in diameter and 3 nm in suspension space, gave a 12.6x near-field enhancement. The optimized structure exhibited an optimally tuned localized surface plasmon resonance of (480–780) nm with a sensitivity of 810 nm/RIU and a full width at half maximum (FWHM) of 13.2 ± 0.5 nm and a Figure of Merit (FOM) of 61 leading to the estimated limit of detection (i.e. of the order of 10− 15 M) at the femtomolar scale given by reference to experimentally determined resolution of the refractive index. Multispectral Dynamic Feature Resonance Model (MDFRM) recorded the dipole (520 nm), quadrupole (630 nm), and Fano (780 nm) modes and retained 96.3% of the variance using PCA and VAE compression. The Hybrid Neural Architecture (CNN, BiLSTM, Attention) achieved a test accuracy of 98.3% and ROC AUC of 0.990 when using four cancer biomarkers. Explainability PlasmaSHAP and Grad-CAM have given a combined explainability of 0.89: the model’s attention was correlated to physical plasmonic hotspots. The experimental results were able to verify the theoretical predictions with an R2 value of 0.996 and the coefficient of variation of less than 1.8 per cent, indicating great reproducibility and usability in real-time and label-free cancer diagnostics.