FEA-Assisted nanostructured piezoelectric sensor and deep learning framework for noninvasive tumor detection and tissue stiffness analysis

Abstract

Biosensors have long been used in environmental monitoring and clinical diagnostics for detecting harmful bacteria, viruses, and biomolecules, evolving from simple biological sensing systems to advanced nanofabrication-based devices. With help of nanotechnology, nano-biosensors having high sensitivity and accuracy can be created to sense and study the biological process. In the current study, the authors propose a noninvasive method for determining tumors using machine learning based on finite element analysis (FEA). The sensor designed here by integrating vibration absorber technology is designed to sense the different in stiffness of tumor tissue. This sensor has been tested with soft tissues models, whose Young modulus is between 9 kPa up to 185 kPa to simulate the malignant condition. Simulated analysis using ANSYS is modal analysis, harmonic analysis and indentation analysis with respect to the different sizes of the tumor ranging from 5 to 25 mm in each layer of depth. One recurrent neural network (RNN) is trained to predict stiffness of the tissue and another is trained to localize the tumor and anticipate the extent of the tumor. The proposed method perfectly reduces the error of estimated tumor size to be minimum of 0.04 mm and the error of detection in stiffness is minimum of 0.0319 kPa.