Application of Artificial Intelligence to an Impact-Based Analysis Method to Estimate the Soft Tissue Phantom Stiffness: A Proof-of-Concept.

Assessing the biomechanical properties of soft tissues can be useful because they are related to their pathophysiological state. This study explores the application of artificial intelligence (AI) to an Impact-Based Analysis Method (IBAM) to predict the mechanical properties of soft tissues. 40 agar-based soft tissue phantoms with different stiffness were prepared. For each phantom, Young's modulus was estimated using dynamic mechanical analysis and IBAM measurements were performed. Various AI-based models were applied to the results obtained with the IBAM approach to predict Young's modulus. Principal component analysis shows that three parameters can explain 85% of the variation in IBAM data. The times of the different maxima of the force signal peaks are significantly correlated with Young's modulus (R2 = 0.99). Most AI-based models allow a decrease in the prediction error compared to the standard IBAM approach (from 5.4% down to 1.2%). Decision trees and ensemble stacking, as well as convolutional neural networks, show a decrease in the prediction error of 74% and 50%, respectively. Applying AI approaches within the IBAM framework is a powerful approach to identify Young's modulus of soft tissues. This study paves the way for using AI-based methods to characterize superficial soft tissue biomechanical properties.