From COMSOL Simulation to Machine Learning: A MEMS Cantilever Biosensor Framework for Tuberculosis Detection

This study presents a simulation-to-AI framework for tuberculosis (TB) detection using a cantilever-based biosensor modeled in COMSOL Multiphysics. Two biosensor configurations with different material compositions (PDMS, silicone, and polyimide) are designed and analyzed to investigate their mechanical responses under antigen–antibody interactions. The biosensor behavior is evaluated in terms of static displacement and eigenfrequency variations across different levels of surface coverage. Finite element simulations are conducted to extract physically meaningful response parameters, including resonant frequency shift and cantilever deflection. These outputs are organized as feature vectors representing each biosensor state under varying biomolecular loading conditions. A data-driven decision framework is introduced based on the extracted simulation features. The problem is formulated as a binary classification task to distinguish between healthy and tuberculosis-infected conditions. Although the study primarily relies on numerical simulations, the proposed framework establishes a structured pathway from physics-based modeling to intelligent diagnostic interpretation. The results demonstrate that cantilever material properties significantly influence sensitivity and dynamic response. Furthermore, the extracted feature space provides a consistent basis for future implementation of supervised machine learning algorithms for automated tuberculosis detection.

کلیدواژه ها:

COMSOL ، MATLAB ، SCR ، PDMS ، design and displacement