RESUMEN
Objective·To analyze the differences and classify hypertrophic cardiomyopathy(HCM)patients and healthy controls(HC)using short-axis cine cardiac magnetic resonance(CMR)images-derived radiomics features.Methods·One hundred HCM subjects were included,and fifty HC were randomly selected at 2∶1 ratio during January 2018 to December 2021 in the Department of Cardiology,Renji Hospital,Shanghai Jiao Tong University School of Medicine.The CMR examinations were performed by experienced radiologists on these subjects.CVI 42 post-processing software was used to obtain left ventricular morphology and function measurements,including left ventricular ejection fraction(LVEF),left ventricular end-diastolic volume(LVEDV)and left ventricular end-diastolic mass(LVEDM).The 3D radiomic features of the end-diastolic myocardial region were extracted from short-axis images CMR cine.The distribution of the radiomic features in the two groups was analysed and machine learning models were constructed to classify the two groups.Results·One hundred and seven 3D radiomic features were selected and extracted.After exclusion of highly correlated features,least absolute shrinkage and selection operator(LASSO)was used,and a 5-fold cross-validation was performed.There were still 11 characteristics with non-zero coefficients.The K-best method was used to decide the top 8 features for subsequent analysis.Among them,four features were significantly different between the two groups(all P<0.05).Support vector machine(SVM)and random forest(RF)models were constructed to discriminate the two groups.The results showed that the maximum area under the curve(AUC)for the single-feature model(first order grayscale:entropy)was 0.833(95%CI 0.685?0.968)and the maximum accuracy for the multi-feature model was 83.3%with an AUC of 0.882(95%CI 0.705?0.980).Conclusion·There are significant differences in both left ventricular function and left ventricular morphology between HCM and HC.The 3D myocardial radiomic features of the two groups are also significantly different.Although single feature is able to distinguish the two groups,the combination of multi-features show better classification performance.
RESUMEN
Cardiac magnetic resonance imaging (CMR) is a widely used non-invasive imaging modality for evaluating cardiovascular diseases. CMR is the gold standard method for left and right ventricular functional assessment due to its ability to characterize myocardial structure and function and low intra- and inter-observer variability. However the post-processing segmentation during the functional evaluation is time-consuming and challenging. A fully automated segmentation method can assist the experts; therefore, they can do more efficient work. In this paper, a regression-based fully automated method is presented for the right- and left ventricle segmentation. For training and evaluation, our dataset contained MRI short-axis scans of 5570 patients, who underwent CMR examinations at Heart and Vascular Center, Semmelweis University Budapest. Our approach is novel and after training the state-of-the-art algorithm on our dataset, our algorithm proved to be superior on both of the ventricles. The evaluation metrics were the Dice index, Hausdorff distance and volume related parameters. We have achieved average Dice index for the left endocardium: 0.927, left epicardium: 0.940 and right endocardium: 0.873 on our dataset. We have also compared the performance of the algorithm to the human-level segmentation on both ventricles and it is similar to experienced readers for the left, and comparable for the right ventricle. We also evaluated the proposed algorithm on the ACDC dataset, which is publicly available, with and without transfer learning. The results on ACDC were also satisfying and similar to human observers. Our method is lightweight, fast to train and does not require more than 2 GB GPU memory for execution and training.