Prediction of TACE Treatment Response in a Preoperative MRI via Analysis of Integrating Deep Learning and Radiomics Features

PurposeTo evaluate the efficiency of an integrated model on MRI scans of hepatocellular carcinoma (HCC) patients for preoperative prediction of transcatheter arterial chemoembolization (TACE) treatment response.MethodsRadiomics and deep learning features were integrated to build a prediction model for preoperative procedures so as to obtain a fast and accurate prediction of TACE treatment response. This is a retrospective study and the data consists of 71 HCC patients who underwent TACE treatment in a single center. These patients were divided into two groups: progressive disease (PD) response (20 patients) and non-progressive disease (N-PD) response (51 patients). fivefold cross-validation was applied to the data set to validate model performance. A receiver operating characteristic (ROC) curve was used to assess the predictive ability of the model. Quantification of its results was performed by calculating the area under the receiver operating characteristic curve (AUC). The accuracy, recall, specificity, precision and f1_score were also calculated for the cutoff value that maximized the AUC value.ResultsAs assessed by the fivefold cross-validation, the integrated model had the best prediction ability, with a value of AUC 0.947 ± 0.069, an accuracy of 0.893 ± 0.088, f1-score of 0.700 ± 0.245, specificity of 0.700 ± 0.245, precision of 0.700 ± 0.245 and a recall of 0.600 ± 0.279. This was followed by the deep learning-based model with an AUC of 0.867 ± 0.121 and the radiomics-based model with an AUC of 0.848 ± 0.128.ConclusionThe experiment results demonstrate that a feature set that combines radiomics and deep learning features tends to be effective in predicting TACE treatment response as opposed to using only one feature. However, due to the limited amount of data, more data will be needed to verify the effectiveness of this method in the future.