No abstract

In recent years, the number of patients with heart failure (HF) has been increasing, and there is an urgent need to elucidate the mechanism and establish treatment methods. Although ejection fraction (EF) is one of the most used indices of cardiac function, some HF patients have preserved EF. Therefore, it is important to identify small changes that do not appear in indices such as EF to elucidate the mechanisms of HF and/or cardiac hypertrophy. In this study, we proposed a semi-automatic method for extracting regions inside the ventricle for analysis. Furthermore, we proposed a data-driven analysis method and applied it to mice with mild transverse aortic constriction (TAC) in which EF did not change much. We created a model that distinguishes the echo images of mice before and after mild TAC using bag-of-features and evaluated the differences in phase and position. After parameter optimization, the best models showed greater than 89% classification performance. In these models, end-systolic phase and proximity to the ventricular boundary were found to be important in discriminating between the two types of mice.