Radiomics-Based Detection of Radionecrosis Using Harmonized Multiparametric MRI

Abstract

In this study, a radiomics analysis was conducted to provide insights into the differentiation of radionecrosis and tumor progression in multiparametric MRI in the context of a multicentric clinical trial. First, the sensitivity of radiomic features to the unwanted variability caused by different protocol settings was assessed for each modality. Then, the ability of image normalization and ComBat-based harmonization to reduce the scanner-related variability was evaluated. Finally, the performances of several radiomic models dedicated to the classification of MRI examinations were measured. Our results showed that using radiomic models trained on harmonized data achieved better predictive performance for the investigated clinical outcome (balanced accuracy of 0.61 with the model based on raw data and 0.72 with ComBat harmonization). A comparison of several models based on information extracted from different MR modalities showed that the best classification accuracy was achieved with a model based on MR perfusion features in conjunction with clinical observation (balanced accuracy of 0.76 using LASSO feature selection and a Random Forest classifier). Although multimodality did not provide additional benefit in predictive power, the model based on T1-weighted MRI before injection provided an accuracy close to the performance achieved with perfusion.

Keywords: multicenter harmonization; multiparametric MRI; radiation-induced necrosis; radiomics analysis.

Figure 1

Radiomics pipeline.

Figure 2

Dimensionality reduction of acquisition parameters TE, TR, flip angle, slice thickness and pixel spacing (left) and radiomics features extracted from the healthy ROIs in T1w images (right) with UMAP showing the sensitivity of radiomic features to the scanner effects.

Figure 3

Effect of harmonization on the distribution of an example radiomic feature extracted from the healthy ROIs across the different scanners (namely, glcm-Correlation) before (A,C) and after ComBat harmonization (B,D).

Figure 4

Dimension reduction of radiomic features extracted from the healthy ROIs on T1w and perfusion images. For each modality, a UMAP clustering is performed on raw data (A,C) and after harmonization (B,D).

Figure 4

Dimension reduction of radiomic features extracted from the healthy ROIs on T1w and perfusion images. For each modality, a UMAP clustering is performed on raw data (A,C) and after harmonization (B,D).

Figure 5

Classification accuracy obtained with the Logistic Regression model before (white) and after ComBat harmonization (gray). The p-values from the Wilcoxon test are shown for each pair. The x-axis refers to radiomics models based on features extracted from different combination of modalities. Anat refer to the radiomic model based on anatomical MR images (T1w, CE-T1w, T2w and FLAIR).

Figure 6

Selected features for the two best radiomics models dedicated to the detection of radionecrosis: perfusion-based model (top) and T1w-based model (bottom).