[4L1-GS-10] AI application: Medicine / Healthcare

[4L1-GS-10-03] Attempt to detect signs using ECG inverse problem analysis with cumulative Gaussian functions

Shingo Tsukada², 〇Akihiro Shiozawa¹, Yuki Iwasaki³, Yayoi Tsukada³, Takuji Oba¹ (1. NTT DATA Mathematical Systems, 2. Nippon Telegraph and Telephone, 3. Nippon Medical School)

Keywords:ECG, TCG, inverse problem, repolarization

In recent years, many analyzes of electrocardiogram (ECG) data have been performed, and most of them utilize neural networks(NN). However, although NN shows high perfomances at classifying and detecting abnormalities, it is difficult to detect medical causes. In this study, we employ a method called ``TENSOR CARDIOGRAPHY (TCG)'' which expresses the ECG by superimposing cumulative Gaussian functions. This makes it possible to express temporal changes in the ECG as fiiting parameters, and as a result, it becomes possible to quantify and visualize slight changes in the ECG, especially changes in the repolarization phase. We actually applied this method to ECG and performed initial verification to see if it can detect signs of abnormalities. As a result, oscillatory changes in fiiting parameters were observed before the occurrence of abnormalities, suggesting that the abnormalities could be predicted.