AI-Driven Control of Flexible Bevel-Tip Needles in Minimally Invasive Surgery Utilizing Adaptive Multi-Layer Perceptron Architecture

In Minimally Invasive Surgical (MIS) procedures, needles are extensively employed for percutaneous interventions, significantly enhancing the accuracy of target-reaching. The primary goal of percutaneous interventional method is to prioritize the patient safety while achieving precise targeting of the intended position. Therefore, the objective of this research work is to devise and deploy a needle steering controller within the tissue environment, utilizing adaptive Multi-layer Perceptron (MLP) architecture. Initially, the kinematic model is formulated to describe the motion characteristics of the steerable needles within tissue, adhering to a unicycle model while incorporating non-holonomic constraints. Subsequently, an AI-driven closed loop needle steering strategy is devised to achieve a precise needle-placement on a desired plane. The efficacy of the proposed controller is validated through the simulation studies along with comparative analysis involving existing controllers.