ARRS 2022 Abstracts

RETURN TO ABSTRACT LISTING


E1308. Multi-region Radiomic Analysis Based on Multi-sequence MRI to Preoperatively Predict Microvascular Invasion in Hepatocellular Carcinoma
Authors
  1. Gao Lanmei; The First Affiliated Hospital of Fujian Medical University; The School of Medical Technology and Engineering, Fujian Medical University
  2. Li Yueming; The First Affiliated Hospital of Fujian Medical University; The School of Medical Technology and Engineering, Fujian Medical University
Objective:
Radiomics has shown great potential in providing valuable information for tumor pathophysiology. We aimed to construct and validate prediction models including radiomic models, clinico-radiological model, and fusion models to predict microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

Materials and Methods:
A total of 115 patients with pathologically confirmed HCC (training set: n = 80; validation set: n = 35) were retrospectively recruited. Radiomics models based on multi-sequence MRI and various regions were built using four classification algorithms. A clinico-radiological model was constructed individually or combined with a radiomics model to generate a fusion model, by multivariable logistic regression. The predictive efficacy of different models was assessed by the AUC value. An integrated discrimination improvement (IDI) value was calculated to compare the discriminative value of the fusion model with that of other models.

Results:
The clinico-radiological model had good efficacy with an AUC (95% CI) of 0.819 (0.732–0.905) in the training dataset and 0.717 (0.551–0.883) in the validation dataset, involving non-smooth margins and peritumoral hypointensity on HBP. Among all radiomics models, the model based on T2 weighted imaging and arterial phase (T2WI-AP model) in the volume of the liver-HCC interface (VOIinterface) showed the best predictive power, with an AUC (95% CI) of 0.866 (0.783–0.947) in the training group and 0.855 (0.731–0.963) in the validation group. The fusion model that incorporated the T2WI-AP radiomics model in VOIinterface and non-smooth tumor margins showed an excellent prediction efficacy (AUC [95% CI] was 0.915 [0.853–0.976] and 0.868 [0.749–0.988], respectively), outperforming the clinico-radiological model and T2WI-AP radiomics model in the training and validation set (IDI > 0).

Conclusion:
The fusion model of multi-region radiomics achieves an enhanced prediction of the individualized risk estimation of MVI in patients with HCC. This may be a beneficial tool for clinicians to improve decision-making in personalized medicine.