A Deep Learning 3D Super-Resolution Radiomics Model based on Gd-Enhanced MRI for Improving Pre-operative Prediction of HCC Pathological Grading

Motivation: Hepatocellular carcinoma (HCC) has a high recurrence rate post-surgery, with early recurrence linked to poor survival outcomes. Current radiomics models have limitations in predicting HCC pathological grading. Goal(s): To develop and validate multi-class radiomics models using super-resolution (SR) images to improve the accuracy of preoperative HCC pathological grading prediction. Approach: A retrospective study was conducted using deep learning-generated SR images to extract quantitative radiomic features, followed by classification using various machine learning models. Results: SR images significantly enhanced the diagnostic performance of radiomics models, particularly in distinguishing between moderately and poorly differentiated HCC, demonstrating higher accuracy and clinical utility. Impact: The study's findings suggest that super-resolution imaging can significantly enhance radiomics models' accuracy in predicting hepatocellular carcinoma grades, potentially leading to better patient stratification and personalized treatment plans, thus improving clinical outcomes and guiding future research in oncologic imaging.