Individualized Prognostication Based on Deep-Learning Models Using Computed Tomography as an Imaging Biomarker After...

Area of Science:

• Hepatocellular Carcinoma Research
• Medical Imaging Analysis
• Artificial Intelligence in Oncology

Background:

• No prior studies utilized deep-learning (DL) models with computed tomography (CT) as an imaging biomarker for predicting long-term outcomes in hepatocellular carcinoma (HCC) patients post-surgery.
• This study aimed to validate DL models for individualized prognostication in HCC patients undergoing resection, using CT as the imaging biomarker.

Purpose of the Study:

• To develop and validate deep-learning (DL) models for predicting postoperative recurrence and overall survival (OS) in patients with hepatocellular carcinoma (HCC) after hepatic resection.
• To assess the utility of computed tomography (CT) imaging as a biomarker within these DL models for individualized prognostication.

Main Methods:

• A cohort of 1733 patients undergoing hepatic resection for solitary HCC was analyzed.
• Deep-learning (DL) predictive models were constructed using clinical variables and CT imaging data to forecast recurrence (2 and 5 years) and OS (>5 and >10 years).
• Permutation importance was employed to determine the significance of each predictive variable, with CT imaging analysis showing the highest importance.

Main Results:

• The developed DL models demonstrated predictive performance with areas under the curve (AUC) of 0.70 for 2-year recurrence, 0.70 for 5-year recurrence, 0.80 for >5-year OS, and 0.80 for >10-year OS.
• CT imaging analysis was identified as the most important explanatory variable by permutation importance.
• Significant differences in postoperative recurrence and mortality rates were observed between high-risk and low-risk groups identified by the DL models.

Conclusions:

• The validated deep-learning (DL) models utilizing CT as an imaging biomarker are effective for individualized prognostication in HCC patients.
• These CT-based DL models hold potential for optimizing treatment planning and improving patient outcomes in hepatocellular carcinoma.