Scholarlink

<p>The use of on-body biosensors for nutrition tracking and public health surveillance is steadily increasing, enabling continuous monitoring of biochemical, physiological, and behavioral indicators. Innovations in nanomaterials and scalable nanotechnology have broadened the sensing capabilities of a variety of personalized biosensors. However, challenges persist in ensuring the data reliability derived from biosensors in practical settings, thereby hindering widespread implementation and public health initiatives. Sensor drift, batch variability, biofluid-dependent confounding, physiological detection delays, and inter-individual heterogeneity introduce systematic uncertainty that cannot be resolved solely by device improvements. Moreover, Artificial intelligence (AI) models themselves introduce additional failure modes related to label noise, domain shift, bias, and model opacity. This review highlights current research on AI-guided biosensors framework that examines reliability at four levels: sensor analytical performance, algorithm robustness, error propagation through systems, and whether results apply across different populations. We analyze several biosensor types-those measuring biochemical nutrition markers and vitamins, and platforms detecting substance exposure-identifying both shared error sources and those unique to specific technologies. Attention is given to how small sensor biases can be amplified through AI pipelines and to the challenges posed by population-scale deployment in terms of equity, privacy, and governance. Finally, we identify emerging research directions-including uncertainty-aware modeling, multimodal and context-aware sensor fusion, and end-to-end validation-as critical enablers for the trustworthy translation of AI-guided biosensors into precision nutrition and data-driven public health surveillance.</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

Reliability and Accuracy of AI-Guided Biosensors in Nutrition Tracking and Public Health Surveillance

  • 期刊论文

  • 发表日期：

    2026年7月6日

查看全文
  • Sajia Afrin Tima,
  • Sudeshana Paramita Ghose,
  • Mahan Ghosh,
  • Md Luthfar Rahman Liman,
  • Md Reazuddin Repon,
  • Asfaker Rahman

摘要

The use of on-body biosensors for nutrition tracking and public health surveillance is steadily increasing, enabling continuous monitoring of biochemical, physiological, and behavioral indicators. Innovations in nanomaterials and scalable nanotechnology have broadened the sensing capabilities of a variety of personalized biosensors. However, challenges persist in ensuring the data reliability derived from biosensors in practical settings, thereby hindering widespread implementation and public health initiatives. Sensor drift, batch variability, biofluid-dependent confounding, physiological detection delays, and inter-individual heterogeneity introduce systematic uncertainty that cannot be resolved solely by device improvements. Moreover, Artificial intelligence (AI) models themselves introduce additional failure modes related to label noise, domain shift, bias, and model opacity. This review highlights current research on AI-guided biosensors framework that examines reliability at four levels: sensor analytical performance, algorithm robustness, error propagation through systems, and whether results apply across different populations. We analyze several biosensor types-those measuring biochemical nutrition markers and vitamins, and platforms detecting substance exposure-identifying both shared error sources and those unique to specific technologies. Attention is given to how small sensor biases can be amplified through AI pipelines and to the challenges posed by population-scale deployment in terms of equity, privacy, and governance. Finally, we identify emerging research directions-including uncertainty-aware modeling, multimodal and context-aware sensor fusion, and end-to-end validation-as critical enablers for the trustworthy translation of AI-guided biosensors into precision nutrition and data-driven public health surveillance.