Scholarlink

<p>This paper presents a sensor featuring a hybrid bismuth–graphene architecture optimized for simultaneous hemoglobin and chemical detection. Electromagnetic simulation using COMSOL Multiphysics demonstrates the robustness of the proposed sensor’s sensing capability across the 0.1–0.7&#xa0;THz frequency bandwidth. It also shows that this sensor has high refractive-index peak sensitivity with 200&#xa0;GHz/RIU for detecting hemoglobin at refractive indices from 1.34 to 1.43 and 500&#xa0;GHz/RIU for detecting chemicals (analytes) at refractive indices from 1.33 to 1.366. In order to increase both the quality of the data interpretation and the predictability of results, multiple machine learning algorithms (Neural Networks, Gradient Boosting, Random Forest, Support Vector Regression) have been applied to analyze sensor data. These data-driven approaches significantly enhance analytical performance, yielding prediction accuracies above 98%, with coefficients of determination of <i>R</i><sup>2</sup> = 0.987 for hemoglobin sensing and <i>R</i><sup>2</sup> = 0.983 for chemical detection. Feature importance analysis reveals that resonance-frequency shift and sensitivity contribute over 60% to the model’s decision-making process. The sensor demonstrates excellent angular stability (0°–80° incidence), robust noise resilience (maintaining &gt; 90% accuracy at 10&#xa0;dB SNR), and consistent cross-validation performance (standard deviation &lt; 0.02). The compact planar architecture, rapid resonance-based spectral interrogation, and machine-learning-assisted prediction enable near real-time analyte estimation while maintaining compatibility with integrated terahertz source–detector and microfluidic platforms. These characteristics position the proposed sensor as a promising design for portable biomedical diagnostics and environmental monitoring applications, combining high sensitivity with intelligent data-driven signal interpretation.</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

High-sensitivity dual-mode hemoglobin and chemical detection using a machine-learning-assisted bismuth–graphene THz sensor

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • S. Vaishnavi,
  • K. Maharajan,
  • C. Sheeja Herobin Rani,
  • D. Jessintha,
  • Manjunathan Alagarsamy

摘要

This paper presents a sensor featuring a hybrid bismuth–graphene architecture optimized for simultaneous hemoglobin and chemical detection. Electromagnetic simulation using COMSOL Multiphysics demonstrates the robustness of the proposed sensor’s sensing capability across the 0.1–0.7 THz frequency bandwidth. It also shows that this sensor has high refractive-index peak sensitivity with 200 GHz/RIU for detecting hemoglobin at refractive indices from 1.34 to 1.43 and 500 GHz/RIU for detecting chemicals (analytes) at refractive indices from 1.33 to 1.366. In order to increase both the quality of the data interpretation and the predictability of results, multiple machine learning algorithms (Neural Networks, Gradient Boosting, Random Forest, Support Vector Regression) have been applied to analyze sensor data. These data-driven approaches significantly enhance analytical performance, yielding prediction accuracies above 98%, with coefficients of determination of R2 = 0.987 for hemoglobin sensing and R2 = 0.983 for chemical detection. Feature importance analysis reveals that resonance-frequency shift and sensitivity contribute over 60% to the model’s decision-making process. The sensor demonstrates excellent angular stability (0°–80° incidence), robust noise resilience (maintaining > 90% accuracy at 10 dB SNR), and consistent cross-validation performance (standard deviation < 0.02). The compact planar architecture, rapid resonance-based spectral interrogation, and machine-learning-assisted prediction enable near real-time analyte estimation while maintaining compatibility with integrated terahertz source–detector and microfluidic platforms. These characteristics position the proposed sensor as a promising design for portable biomedical diagnostics and environmental monitoring applications, combining high sensitivity with intelligent data-driven signal interpretation.