Homogenization of brain MRI from a clinical data warehouse using contrast-enhanced to non-contrast-enhanced image translation
Abstract
Purpose Clinical data warehouses (CDW) provide access to massive amounts of medical images, but these images are often heterogeneous. They can for instance include images acquired both with or without the injection of a gadolinium-based contrast agent. Harmonizing such data sets is thus fundamental to guarantee unbiased results, for example when performing differential diagnosis. Furthermore, classical neuroimaging software tools for feature extraction are typically applied only to images without gadolinium. The objective of this work is to homogenize images from a CWD and enable the extraction of consistent features from brain MR images, no matter the initial presence or absence of gadolinium. Approach We propose and compare different 3D U-Net and conditional GAN models to convert contrast-enhanced T1-weighted (T1w-ce) into non-contrast-enhanced (T1w-nce) brain MRI. These models were trained using 230 image pairs and tested on 77 image pairs from the CDW of the Greater Paris area. Results Validation using standard image similarity measures demonstrated that the similarity between real and synthetic T1w-nce images was higher than between real T1w-nce and T1w-ce images for all the models compared. The best performing models were further validated on a segmentation task. We showed that tissue volumes extracted from synthetic T1w-nce images were closer to those of real T1w-nce images than volumes extracted from T1-ce images. Conclusions We showed that deep learning models could synthesize T1w-nce from T1w-ce images and that reliable features could be extracted from the synthetic images, thus demonstrating the ability of such methods to homogenize a data set coming from a CDW.
Origin : Files produced by the author(s)