Prognosis of Tissue Stiffness Through Multilayer Perceptron Technique With Adaptive Learning Rate in Minimal Invasive Surgical Procedures

Abstract

Flexible needles are navigated through anatomical pathways to reach deep seated tissues for minimally invasive surgical procedures. During such risky navigation, anatomical obstacles and the target malignant tissue regions could be dislodged due to excessive stress upon needle-tissue interaction. Hence, knowledge about the interactive forces is essential to execute a safe needle steering procedure during percutaneous cancerous treatments. This paper proposes an adaptive learning rate based multilayer perceptron technique for determining Young’s modulus of tissue at each stage of navigation and then utilizing this value to predict the deflection of flexible needle in tissue environment. To validate the accuracy of predictions, an energy-based model is incorporated into the analysis. Simulation results demonstrate that the proposed model can efficiently predict Young’s modulus in just 0.59 secs. To further validate the efficacy of this novel methodology, extensive experimental studies are conducted, including rigorous statistical analysis using ANOVA with a 5% accuracy level. The effectiveness of neural networks is underscored through a two-sample t-test across 5 different trials, revealing consistently low mean absolute errors, typically below 1.5 kPa, except in trial 3. These findings highlight the reliability of the proposed novel technique in predicting Young’s modulus and ensuring accurate needle deflection predictions.