Image post-processing
The DICOM format of high-quality 3D-TEE ASD images at the isovolumetric diastole stored in the EchoPAC workstation were imported into specialized post-processing software Mimics Innovation Suite 17.0 (Materialise, Leuven, Belgium), where the source imaging datasets can be converted to 3D digital models (Fig.1B ). Various post-processing methods including threshold segmentation, interactive segmentation and surface smoothing were used to create an accurate 3D digital ASD model. Firstly, threshold segmentation was applied to distinguish the blood pool border of the ASD from the endocardium, from which the blood pool could be filtered and the mask of the ASD were obtain initially. The threshold was adjusted with reference to the endocardial in different cases. After threshold segmentation, there still exist much noise and dropout gaps of the mask, manual editing is necessary to correct segmentations errors. Interactive segmentation was utilized to artificially exclude the noise and fill of the dropout gaps. Finally, surface smoothing function key was used to smooth the surface of the anatomy before exporting to the software 3-matic(Materialise, Leuven, Belgium), where the ASD mask could be calculated for a 3D hollow mold. (Fig. 1C )