This website requires JavaScript.

Spatial-temporal V-Net for automatic segmentation and quantification of right ventricles in gated myocardial perfusion SPECT images

Chen ZhaoShi ShiZhuo He ...+4 Weihua Zhou
Background. Functional assessment of right ventricles (RV) using gatedmyocardial perfusion single-photon emission computed tomography (MPS) heavilyrelies on the precise extraction of right ventricular contours. In this paper,we present a new deep learning model integrating both the spatial and temporalfeatures in SPECT images to perform the segmentation of RV epicardium andendocardium. Methods. By integrating the spatial features from each cardiacframe of gated MPS and the temporal features from the sequential cardiac framesof the gated MPS, we develop a Spatial-Temporal V-Net (S-T-V-Net) for automaticextraction of RV endocardial and epicardial contours. In the S-T-V-Net, a V-Netis employed to hierarchically extract spatial features, and convolutionallong-term short-term memory (ConvLSTM) units are added to the skip-connectionpathway to extract the temporal features. The input of the S-T-V-Net is anECG-gated sequence of the SPECT images and the output is the probability map ofthe endocardial or epicardial masks. A Dice similarity coefficient (DSC) losswhich penalizes the discrepancy between the model prediction and the groundtruth is adopted to optimize the segmentation model. Results. Our segmentationmodel was trained and validated on a retrospective dataset with 34 subjects,and the cardiac cycle of each subject was divided into 8 gates. The proposedST-V-Net achieved a DSC of 0.7924 and 0.8227 for the RV endocardium andepicardium, respectively. The mean absolute error, the mean squared error, andthe Pearson correlation coefficient of the RV ejection fraction between theground truth and the model prediction are 0.0907, 0.0130 and 0.8411.Conclusion. The results demonstrate that the proposed ST-V-Net is an effectivemodel for RV segmentation. It has great promise for clinical use in RVfunctional assessment.