Approved Research

Automatic and Trustworthy Radiomics for Cardiovascular Diseases

Lay summary

Aim:

The project aims at building an automatic model for trustworthy radiomics feature extraction and correlate it with a specific disease. The numerical quantifiers of shape and tissue characters extracted from the model can be combined with qualitative descriptors to facilitate more promising and reproducible clinical reports.

Scientific rationale:

The images we acquired under the clinical settings are not just pictures but data including information about disease-specific processes. However, the cardiac and respiratory motion makes CMR-based radiomics extremely challenging. By incorporating novel registration and segmentation, we intend to develop novel algorithms to extract, analyse and model many medical features correlated with specific diseases, improving medical decision-making.

Project Duration:

36 months

Public Health Impact:

With advanced registration and segmentation ability, the proposed methods will extract novel imaging biomarkers with biological context with informative radiomics. We hope the proposed model will enjoy strong reproducibility and explainability, hence having the potential to be translated into the clinical setting.