We developed intelligent non-invasive LVS risk prediction techniques based on electrocardiograms (ECGs) to automatically predict whether patients suffer from left ventricular hypertrophy (LVH). Our techniques outperform the traditional invasive method (e.g., CT/MRI) with high accuracya low cost. We cooperated with Taipei Veterans General Hospitaltwo Japanese hospitals to validate our techniques.

We have developed an intelligent non-invasive LVH risk prediction technique with two-stage model architecture based on ECGs. The ECG representation module extracts ECG features,a multimodality module combines patients＇ dataECG features for final prediction. This technique can reduce the cost of medical resourcesthe injury of patients from invasive detection. Under the evaluation criteria, this prediction model is far superior to the traditional ECG image diagnosis method. According to in-depth analysis, when the model gives positive for LVH, even if no LVH is observed at the moment, the risk of developing LVH in the future is 8.4 times the risk of being evaluated to be normal by the prediction model.

Left ventricular hypertrophy (LVH) increases the risk of heart failure, arrhythmias, cardiovascular death,sudden cardiac death. It is easy to ignore the diagnosis of LVH in young adults because they are relatively young, resulting in a family tragedy, socioeconomic medical burden,significant loss to the country. An electrocardiogram (ECG) is a non-invasive, inexpensive,widely used clinical test for the diagnosis of heart disease. We used the artificial intelligence (AI) ECG module to identify patients with concealed LVHcardiovascular death. The AI ECG module robustly improved the diagnostic performance of LVH (AUC: 0.89) compared to traditional ECG criteria (AUC: 0.64). The results were similar in the Japanese group