Skip to main content
. Author manuscript; available in PMC: 2025 Jan 1.
Published in final edited form as: Comput Med Imaging Graph. 2024 Apr 16;115:102382. doi: 10.1016/j.compmedimag.2024.102382

Table 6.

Assessment of CardSegNet Framework Segmentation Dice Performance on M&M-2 and Local Datasets Using 10-Fold Cross-Validation.

Dataset Label Cardiac Slices
 Mean Overall Dataset
Base Middle Apex
M&M-2 LV 95.1±0.3 96.5±0.5 94.3±0.9 95.3±0.6 95.4±0.6
Myo 95.4±0.7 97.8±0.8 95.1±1.1 96.1±0.8
RV 94.6±0.4 95.9±0.4 93.9±0.8 94.8±0.5
Rajaie CMRI Dataset* LV 95.4±0.7 96.8±0.6 94.6±0.9 95.3±0.7 94.5±0.7
Myo 93.1±0.7 94.4±0.5 92.4±0.8 96.1±0.6
*

Although images of the long axis were also available in our local database, only images of the short axis were utilized for the comparison purpose. "Base," "Middle," and "Apex" labels were assigned to the top 30 %, subsequent 40 %, and remaining 30 % of CMR slices, respectively.