Metabolic Palm Scanner: A SERS–Thermal Hybrid Analyzer for Non-Invasive Metabolic Risk Screening
Metabolic Palm Scanner: A SERS–Thermal Hybrid Analyzer for Non-Invasive Metabolic Risk Screening
Babak Sharifi Taskuh1
1) Department of Biomedical Engineering, Faculty of Technical Engineering, Islamic Azad University, Sari Branch, Sari, Iran
IOI Code
:
ABEB-ABHGEF
Publication :
4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
The escalating global burden of metabolic disorders — including type 2 diabetes mellitus (T2DM), metabolic syndrome (MetS), and mitochondrial dysfunction — demands rapid, non-invasive, multi-parameter diagnostic tools capable of capturing heterogeneous metabolic signatures at the point of care. This paper proposes a handheld Metabolic Palm Scanner that integrates Surface-Enhanced Raman Spectroscopy (SERS) with thermal micro-pattern imaging to non-invasively estimate a person’s metabolic health state from the palm, and establishes its technical feasibility through a fully computational (in-silico) validation pipeline. The dual-modality architecture targets three complementary biomarker dimensions: glucose and insulin-resistance markers via SERS spectral analysis, oxidative-stress indicators (uric acid, reduced glutathione), and microvascular thermal patterns reflecting metabolic heat dysregulation and altered perfusion. A multi-stage MATLAB pipeline combining principal component analysis (PCA), support vector machine (SVM) classification, and Pennes bioheat-equation-based thermal modeling was used to generate and classify synthetic data representing three metabolic states — Normal, Pre-MetS, and Full MetS/T2DM. Synthetic SERS spectra were produced by Lorentzian peak superposition with Poisson and Gaussian noise, and palm thermal maps were generated by finite-element bioheat-transfer simulation, both parameterized from published physiological ranges. Under 10-fold cross-validation on this synthetic dataset, the fusion classifier achieved an overall accuracy of 89.3% and macro-AUC of 0.94, outperforming single-modality SERS-only (79.4%, AUC 0.87) and thermal-only (72.1%, AUC 0.79) baselines by 9.9 and 17.2 percentage points, respectively. These results should be interpreted as evidence of algorithmic and theoretical feasibility rather than clinical performance, as no human tissue, sweat, or in vivo measurements were used. The reported metrics describe how well the fusion algorithm separates the assumed statistical models of the three classes, rather than diagnostic accuracy in real patients. The study nonetheless provides a concrete, reproducible computational framework — including the governing equations, noise models, and fusion architecture — to guide subsequent hardware prototyping, gold-nanostar substrate characterization, and clinical validation.
Keywords : sers thermal imaging metabolic syndrome insulin resistance oxidative stress non invasive diagnostics raman spectroscopy bioheat transfer multimodal fusion svm classification