Left Ventricle Wall Motion Abnormalities Detection in Cardiac MRI

Cardiovascular diseases are the main cause of death in the world. The diagnosis of these pathologies is based on physiological signals and various imaging modalities such as magnetic resonance imaging (MRI), ultrasound imaging and multi-arrays CT scanner. Among all these cardiac imaging modalities, MRI is a powerful tool that allows an accurate measurement of both global and regional myocardial contraction. The aim of this work is to propose an automatic method based on the combination of LSTM-U-Net architecture and the monogenic signal for detecting the dyskinesia and hypokinesia abnormalities. To detect wall motion abnormalities, a characterization and a localization of the left and right ventricles based on U-Net is first established. Once the left ventricle is detected, the monogenic signal is used to detect the direction of the endocardial contraction. The proposed algorithm was retrospectively tested on a population of 52 subjects, including 20 subjects with normal ventricular function and 32 pathological cases with hypokinesia and dyskinesia abnormalities. The results illustrate that the phase orientation is the most important parameters characterizing the left ventricle. Compared to the gold standard expert interpretation and LGE analysis, the results show that the accuracy of the proposed algorithm in the detection of hypokinesia and dyskinesia was respectively 77.27% and 82.95% while the sensitivity was 82.78% and 87.39 %. In addition, a higher specificity of 71.51% and 75.38 %,was achieved. The findings of this study illustrate the accuracy of the proposed algorithm in the detection of left ventricular wall motion abnormalities.