Abstract:Objective: To assess the feasibility of textural feature analysis of Cine-cardiac magnetic resonance (CMR) as a tool to evaluate heart failure grades in patients with hypertrophic cardiomyopathy (HCM). Methods: Eighty-two patients with HCM that underwent CMR examination at The Southwest Hospital of Army Medical University from 2013 to 2016 were enrolled in this retrospective study. All of the patients were divided into two groups according to the New York Heart Association (NYHA) classification of heart failure. Patients with NYHAⅠ-Ⅱwere selected as the mild heart failure group, and the other patients with NYHAⅢ-Ⅳwere selected as the severe heart failure group. The textures of left ventricular myocardium at the end of diastasis were extracted manually by MATLAB software twice from Cine sequence, and the Bland-Altman analysis was used to calculate the features consistency of the two times. The convolutional neural networks (CNN) calculated the area under the curve (AUC) of the features after optimization. Chi-square test was used to compare the difference of gender distribution between the two groups. The percentage of late gadolinium enhancement (LGE) in the left ventricle was calculated by Circle Cardiovascular Imaging (CVI, Canada) software. Argus software was used to calculate the left ventricular ejection fraction (EF), cardiac output (CO) and cardiac index (CI). Mann-Whitney U test was performed to compare the difference of the age, body mass index (BMI) and left ventricular function parameters between the two groups, and the receiver operating characteristic (ROC) was used to analyze the AUC of those function parameters. Results: The gender distribution, age and BMI were not significantly different between the two groups (P=0.407, 0.338, 0.120). The EF, LGE, CO and CI cardiac function parameters of left ventricles between the mild and severe heart failure groups were of no significant difference (P=0.860, 0.072, 0.944, 0.568). The AUC calculated by CNN was 91.31%, which was significantly larger than that of EF, LGE, CO and CI (51.22%, 62.53%, 49.51%, 48.91%). Conclusion: Textural feature analysis algorithm can effectively identify the difference of NYHAⅠ-Ⅱand NYHAⅢ-Ⅳgrades in patients with HCM, and with higher sensitivity than that of CMR function parameters.
