Robust prediction of tool-tissue interaction force using ISSA-optimized BP neural networks in robotic surgery

A persistent problem with robot-assisted minimally invasive surgery is soft tissue damage caused by the exertion of excessive force due to the surgeon's lack of direct access to the surgical site. A solution to predict clamp force accurately is needed to enhance surgical safety and efficiency. The current proposal concerns a deep learning-based solution utilizing a backpropagation neural network (BPNN) optimized by improved sparrow search algorithm (ISSA) to predict clamp force on soft tissue. This method optimizes the BPNN using ISSA and combines dynamic parameters and geometric characteristics, such as contact area of the clamp blade, loading speed, displacement and time, during clamping to predict clamping force on soft tissue. Circular chaotic mapping, golden sine and crisscross strategies were introduced to increase sparrow search algorithm performance, enabling ISSA-optimized BP to achieve substantial improvements in precision and prediction speed for estimating soft tissue clamping force. The ISSA-BP clamping force prediction model outperforms the BP, ALO-BP, GA-BP, GWO-BP, WOA-BP and SSA-BP models for model evaluation indicators such as RMSE, MSE, MAE, SSE and R². The R² of ISSA-BPNN is 99.24%. The enhanced ISSA-BPNN model demonstrates superior performance in predicting clamp force on soft tissues during robot-assisted surgeries. The novel method has the potential to increase surgical safety, accuracy and efficiency, representing an advance in the field of surgical robotics.