Regression Prediction of Lesion for HPSD Radiofrequency Ablation Based on Backpropagation Algorithm

Atrial fibrillation is a dangerous arrhythmia and radiofrequency ablation technology is a newly emerging surgical treatment. Compared to the standard radiofrequency catheter ablation mode with low power and long duration, high-power and short duration radiofrequency catheter ablation significantly improves surgical efficiency and has good efficiency and safety. This study will conduct in vitro experiments on high power short-term radiofrequency catheter ablation of pig hearts. The experiments will be divided into 5 groups according to power/time and 4 pressure groups will be divided under each power/time group. The damage size of each ablation point will be measured manually, and 111 sets of experimental data will be obtained. Finally, the algorithm of back-propagation BP neural network is used for machine learning regression prediction of data and evaluation indicators are used to evaluate the advantages and disadvantages of this regression model, such as determination coefficient (R²), Mean Average Error (MAE), Mean Square Error (MSE), and Mean Average Percentage Error (MAPE). According to the results, the four damage sizes of Surface Width (SW), Maximum Damage Width (MW), damage Depth (D) and calculated damage Volume (V) are correlated with eight parameters: power, ablation time, contact force, initial impedance, impedance drop, initial temperature, maximum temperature and temperature rise.