Tarika Devi Vijayaraj
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Tarika Devi Vijayaraj has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Durgashini Annamalai
Department of Microbiology, Sri Ramachandra Medical College & Research Institute (SRMC&RI), Sri Ramachandra Institute of Higher Education and Research, Chennai 600116, India
1 shared publication - Rakshitha Karpagarajan
Department of Microbiology, Sri Ramachandra Medical College & Research Institute (SRMC&RI), Sri Ramachandra Institute of Higher Education and Research, Chennai 600116, India
1 shared publication - Sairam B. Subramaniam
Department of Computer Science and Medical Engineering, Sri Ramachandra Faculty of Engineering and Technology (SRET), Sri Ramachandra Institute of Higher Education and Research, Chennai 600116, India
1 shared publication - Swati Kumari
Department of Microbiology, Sri Ramachandra Institute of Higher Education and Research, India
1 shared publication
Latest Publications
- Open Access
- 25 September 2026
- Published by MDPI AG in Biosensors
- Vol. 16 (10) , 537
Advancing Biosensing Through Artificial Intelligence: Overcoming Conventional Limitations Towards Intelligent Diagnostic Systems
Artificial intelligence (AI) is increasingly integrated with biosensors to improve signal processing, feature extraction, calibration, and clinical interpretation. However, analytical sensitivity and high internal validation accuracy do not necessarily translate into clinically useful or generalisable diagnostic performance. This review therefore reframes AI-enabled biosensing around clinical actionability rather than technological performance alone. We critically examine how AI can address signal noise, drift, matrix effects, and complex or high-dimensional biosensor outputs while evaluating how model selection should reflect data modality, dataset size, computational constraints, and intended deployment. Importantly, we distinguish the physical limit of detection (LoD) of the sensing system from algorithm-assisted effective detectability: AI cannot intrinsically lower the physical detection limit of a fixed transducer, but it can improve extraction and interpretation of weak signals near the noise floor. A central contribution of this review is a clinically actionable threshold framework that links representative biomarkers to decision-relevant concentration ranges, clarifying when further LoD reduction is clinically meaningful and when robustness, selectivity, dynamic range, calibration stability, and reproducibility matter more. We also critically appraise representative AI-enabled biosensor studies by sample size, validation strategy, reference standard, generalisability, interpretability, computational requirements, and evidence maturity, rather than headline accuracy alone. Across electrochemical, optical and spectroscopic, wearable, and point-of-care platforms, major translational barriers include biofouling, matrix interference, signal drift, dataset shift, insufficient external validation, privacy, and regulatory requirements. We conclude that progress in intelligent biosensing should be judged by reliable performance across clinically relevant ranges and demonstrated utility in real-world diagnostic workflows, rather than by progressively lower LoDs or isolated accuracy values alone.
- 25 September 2026
- Published by MDPI AG in Biosensors
- Vol. 16 (10) , 537
Open AccessAdvancing Biosensing Through Artificial Intelligence: Overcoming Conventional Limitations Towards Intelligent Diagnostic Systems
Artificial intelligence (AI) is increasingly integrated with biosensors to improve signal processing, feature extraction, calibration, and clinical interpretation. However, analytical sensitivity and high internal validation accuracy do not necessarily translate into clinically useful or generalisable diagnostic performance. This review therefore reframes AI-enabled biosensing around clinical actionability rather than technological performance alone. We critically examine how AI can address signal noise, drift, matrix effects, and complex or high-dimensional biosensor outputs while evaluating how model selection should reflect data modality, dataset size, computational constraints, and intended deployment. Importantly, we distinguish the physical limit of detection (LoD) of the sensing system from algorithm-assisted effective detectability: AI cannot intrinsically lower the physical detection limit of a fixed transducer, but it can improve extraction and interpretation of weak signals near the noise floor. A central contribution of this review is a clinically actionable threshold framework that links representative biomarkers to decision-relevant concentration ranges, clarifying when further LoD reduction is clinically meaningful and when robustness, selectivity, dynamic range, calibration stability, and reproducibility matter more. We also critically appraise representative AI-enabled biosensor studies by sample size, validation strategy, reference standard, generalisability, interpretability, computational requirements, and evidence maturity, rather than headline accuracy alone. Across electrochemical, optical and spectroscopic, wearable, and point-of-care platforms, major translational barriers include biofouling, matrix interference, signal drift, dataset shift, insufficient external validation, privacy, and regulatory requirements. We conclude that progress in intelligent biosensing should be judged by reliable performance across clinically relevant ranges and demonstrated utility in real-world diagnostic workflows, rather than by progressively lower LoDs or isolated accuracy values alone.