Background: Atrial fibrillation (AF) is a common yet dangerous cardiac arrhythmia that can lead to severe cardiovascular events such as strokes. Despite the capability of current electrocardiogram (ECG) technologies to detect AF, their accuracy is affected by signal interference and human error. In order to enhance the accuracy of AF detection and improve the efficiency of diagnosis by physicians, we introduce artificial intelligence (AI) into the analysis of electrode patch locations, intending to achieve more precision in the predictions. Research Objective: The goal of this research is to explore the optimal torso surface bipolar electrode locations for AF detection using neural network technology. Research Method: ECG patch data was collected from 52 patients, with each segment lasting 30 seconds, and simple noise reduction processing was conducted. Despite the limited data, it was randomly divided into training and validation sets in a 4:1 ratio and performed five splits to ensure reliable results. ResNet neural network was employed for data analysis, with a batch size of 32 and training epochs of 2. Finally, we analysed the results of electrode patch data collection from different locations and evaluated their impact on the accuracy of AF detection. Main Results: The third position has the highest accuracy and the lowest loss. The 227th position has the lowest accuracy and the highest loss rate. The accuracy of the front chest is significantly higher than that of the back position and the loss rate is significantly lower than that of the back position.However, the accuracy of the model needs to be further improved Interpretation and Analysis: Physicians should pay special attention to the location of electrode patch data collection when diagnosing AF and choose appropriate analysis strategies based on this information to enhance accuracy
Keywords: Atrial fibrillation, artificial intelligence, torso surface bipolar electrode patches, electrocardiogram, neural network