959P - Machine learning-based MRI radiomics predicts overall survival of unresectable hepatocellular carcinoma undergoing transarterial chemoembolization plus PD-(L)1 inhibitors and molecular targeted therapy

Background

The prognosis of patients with unresectable hepatocellular carcinoma (HCC) undergoing transarterial chemoembolization (TACE) plus PD-(L)1 inhibitors and molecular targeted therapy (MTT) varies widely on an individualized basis. The purpose of this study was to develop and validate an MRI-based radiomics model for predicting overall survival of such patients.

Methods

This bi-center, retrospective, cohort study included 119 unresectable HCC patients undergoing the combination therapy between November 2018 and March 2023. Patients were randomly assigned in a 7:3 ratio to form a training cohort (n=83) and a testing cohort (n=36). Study endpoint was overall survival (OS). The radiomics features were extracted from MRI images of T1WI arterial phase, T2WI, DWI and DELAY sequences, and a radiomics signature was constructed based on machine learning algorithm. Adding the radiomics signature to the clinical model, a combined model was developed and validated. The predictive performance of the combined and clinical models was evaluated and compared based on discrimination, calibration and clinical decision.

Results

Age (p=0.013), Child-Pugh grade (p<0.001), MELD score (p<0.001), ECOG performance status (p=0.048), and tumor size (p=0.006) were included the clinical model. With radiomics signature added, the combined model showed improved discrimination performance (Area under curve [AUC] 0.756 vs. 0.608, 12-month survival probability; AUC 0.679 vs. 0.573, 18-month survival probability, in the testing cohort). The calibration curves of the two models displayed good concordance between predicted and observed probabilities (p>0.05). Decision curve analysis showed that when predicting the 12-month survival probability, the combined model's net benefit was superior to that of the clinical model between a threshold probability range of 10% to 60%.

Conclusions

This study presents an MRI-based radiomics model that could provide a preoperative individualized prediction of survival probability of unresectable HCC patients undergoing TACE plus PD-(L)1 inhibitors and MTT.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The author.

Funding

Has not received any funding.

Disclosure

The author has declared no conflicts of interest.