Harmonization of radiomic features of breast lesions across international DCE-MRI datasets

Abstract

Purpose: Radiomic features extracted from medical images acquired in different countries may demonstrate a batch effect. Thus, we investigated the effect of harmonization on a database of radiomic features extracted from dynamic contrast-enhanced magnetic resonance (DCE-MR) breast imaging studies of 3150 benign lesions and cancers collected from international datasets, as well as the potential of harmonization to improve classification of malignancy.Approach: Eligible features were harmonized by category using the ComBat method. Harmonization effect on features was evaluated using the Davies-Bouldin index for degree of clustering between populations for both benign lesions and cancers. Performance in distinguishing between cancers and benign lesions was evaluated for each dataset using 10-fold cross validation with the area under the receiver operating characteristic curve (AUC) determined on the pre-and postharmonization sets of radiomic features in each dataset and a combined one. Differences in AUCs were evaluated for statistical significance.Results: The Davies-Bouldin index increased by 27% for benign lesions and by 43% for cancers, indicating that the postharmonization features were more similar. Classification performance using postharmonization features performed better than that using preharmonization features (p < 0.001 for all three). Conclusion:Harmonization of radiomic features may enable combining databases from different populations for more comprehensive computer-aided diagnosis models of breast cancer.

How can we combat multicenter variability in MR radiomics? Validation of a correction procedure

“…Although promising results have been reported in the literature [35,36], these techniques require access to the images, unlike ComBat. The ComBat realignment method has been previously used in MR radiomic studies [20][21][22][23] without any explicit validation or investigation of the respective role of the image standardization and of the scanner/protocol effect compensation as studied here (Figures 1 and 2). In [20], authors reported an increased accuracy of Entropy extracted from apparent diffusion coefficient MR images to predict the locoregional control in cervical cancer after ComBat, fully consistent with our findings.…”

“…Although it has been used in MR radiomic studies [20][21][22][23], it has never been validated in that highly challenging context.…”

Radiomic analysis of HTR-DCE MR sequences improves diagnostic performance compared to BI-RADS analysis of breast MR lesions

“…Furthermore, all the data were obtained from one center with a single MRI machine, software and protocol, and our results need to be externally validated probably using a harmonization of radiomic features method such as the ComBat method (54). Fourth, we only performed a cross validation of our results due to the relatively small sample size (174 lesions).…”

Harmonization of Quantitative Parenchymal Enhancement in T1‐Weighted Breast MRI

“…The alternative would be a data‐driven approach. Such approaches exist; for example, by harmonizing given feature distributions, 39 or by harmonizing MR images using deep learning, from which features can then be calculated 40 . Our model‐based approach may have several advantages over data‐driven approaches.…”