Cardiovascular disease is a leading cause of mortality worldwide. The application of Deep Learning algorithms in cardiac medicine is increasingly gaining attention in this field. Image segmentation and analysis of computed tomography images are crucial initial steps for important applications such as three-dimensional cardiac structure reconstruction, disease pre-diagnosis, and treatment planning. However, existing datasets often have one or more limitations: they may be outdated, lack comprehensive coverage of cardiac components, sourced from abroad, or lack labeling for cardiovascular diseases. To serve the Vietnamese cardiovascular patient group, the team conducted a comprehensive study of publicly available datasets worldwide, researched raw volumetric images from Toshiba Aquiline ONE scanner, and performed manual segmentation and labeling of cardiac diseases under the guidance, supervision, and verification of physicians and experts. The resulting VHSCDD dataset overcomes the aforementioned limitations, facilitating experimentation, comparison, and development of Deep Learning algorithms in this field.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |