Contrast-Enhanced Brain MRI Synthesis With Deep Learning: Key Input Modalities and Asymptotic Performance | IEEE Conference Publication | IEEE Xplore

Contrast-Enhanced Brain MRI Synthesis With Deep Learning: Key Input Modalities and Asymptotic Performance


Abstract:

Contrast-enhanced medical images offer vital insights for the accurate diagnosis, characterization and treatment of tumors, and are routinely used worldwide. Acquiring su...Show More

Abstract:

Contrast-enhanced medical images offer vital insights for the accurate diagnosis, characterization and treatment of tumors, and are routinely used worldwide. Acquiring such images requires to inject the patient intravenously with a gadolinium-based contrast agent (GBCA). Although GBCAs are considered safe, recent concerns about their accumulation in the body tilted the medical consensus towards a more parsimonious usage. Focusing on the case of brain magnetic resonance imaging, this paper proposes a deep learning method that synthesizes virtual contrast-enhanced T1 images as if they had been acquired after the injection of a standard 0.100mmol/kg dose of GBCA, taking as inputs complementary imaging modalities obtained either after a reduced injection at 0.025mmol/kg or without any GBCA involved. The method achieves a competitive structural similarity index of 94.2%. Its asymptotic performance is estimated, and the most important input modalities are identified.
Date of Conference: 13-16 April 2021
Date Added to IEEE Xplore: 25 May 2021
ISBN Information:

ISSN Information:

Conference Location: Nice, France

1. Introduction

Gadolinium and biomedical imaging. Gadolinium-based contrast agents (GBCA) are routinely used in biomedical imaging. An approximate 40% of all magnetic resonance imaging (MRI) sessions in Europe and the United States rely on GBCAs to diagnose, characterize or monitor lesions that would remain otherwise poorly visible [1], [2]. In neuroimaging, contrast-enhanced T1-weighted imaging is the cornerstone modality for the detection and precise delineation of brain glioma and metastases, which cause each year hundreds of thousands of deaths worldwide [3], [4]. GBCAs are considered safe, with less than one patient in 100,000 that experiences an adverse reaction [1]. However, the linear sub-category of GCBAs has been withdrawn from European markets in 2017 on the suspicion of abnormal accumulation of gadolinium [5]. Gadolinium has also been identified as a possible trigger of the rare nephrogenic systemic fibrosis disease in patients with renal insufficiency [6]. Macrocyclic GBCAs remain recommended if not systematic in a large number of situations in order to achieve the best diagnosis performance, but recent guidelines suggest to aim for a more parsimonious usage, especially in the case of chronic disease monitoring [7].

Proposed deep learning approach for the synthesis of contrast-enhanced brain images.

References

References is not available for this document.