Abstract
The reliable estimation of left ventricle (LV) volume from 3D medical data enables both visible shape exhibition and accurate volume report, for efficiently clinical cardiac diagnosis. However, it is still open challenging as: 1) intermediate volume error caused by fully depending on segmentation, 2) non-visualized result via solo using direct regression, 3) ambiguous conclusion from the unmatching multi-type outputs, 4) large spatial scene of stereo domain, 5) unbalanced information in 3D medical data between depth and slice plane, 6) large regression output space. We proposed RE-3DLVNet the first powerful work for the refined volume estimation of 3D LV in a novel way segmentation volume±reinforced volume. It uses segmentation with advantage of intuitive voxel-to-voxel optimization to fast stabilize a rough volume. Reciprocally, it makes reinforced quantification (ReinQuan) with advantage of global inference to focus on checking scene misunderstanding and precisely remedy accumulative error caused by massive local optimization, for refining the volume. 1) In segmentation, a probe convolution creatively perceives global scene distribution, and a low-to-high depth segmentation net with high depth connection releases latent structure in low-resolution depth. 2) In ReinQuan, a regression net is conducted to refine estimating the reinforced volume of the segmentation, with focused contracted output space. 3) Interactively, a consistent constraint transfers the ReinQuan deviation between segmented result and its ground truth for penetratively penalizing. Extensive experiments show RE-3DLVNet clinical potential in assessing cardiac function and anatomy. It 1) enables exact 3D LV volume estimation with mean absolute error down to 1.03ml, increasing measurement precise by 30.41% compared to segmentation volume, and 2) achieves accurate visualization segmentation with Dice coefficient up to 95.13%.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.