CT-based Machine Learning Radiomics Modeling: Survival Prediction and Mechanism Exploration in Ovarian Cancer...

Area of Science:

• Oncology
• Radiology
• Genomics

Background:

• Ovarian cancer prognosis remains challenging.
• Accurate prediction of overall survival is crucial for patient management.

Purpose of the Study:

• To develop a radiomics model using CT to predict overall survival in ovarian cancer.
• To integrate genomic data with the radiomics score (Rad-score) to understand gene expression associations.

Main Methods:

• Retrospective analysis of 455 ovarian cancer patient CT and clinical data.
• Development of a Rad-score using Cox regression and LASSO methods.
• Genomic data analysis from TCGA to identify differential gene expression and hub genes.

Main Results:

• Prognostic models incorporating FIGO stage, residual disease, and Rad-score demonstrated high predictive accuracy (AUCs up to 0.906).
• The clinical-radiomics model showed good clinical applicability and net benefit.
• EMP1 was identified as a hub gene, potentially involved in extracellular matrix organization and focal adhesion.

Conclusions:

• FIGO stage, residual disease, and Rad-score effectively predict overall survival in ovarian cancer.
• The Rad-score may predict prognosis by reflecting EMP1 expression and its biological functions.