Génération de scanners synthétiques à partir de L’IRM UTE, à l’aide de l’intelligence artificielle : évaluation initiale dans la mucoviscidose
Year: 2023
Type: dissertation
Abstract: Lung MRI with ultrashort echo times (UTEs) enables high-resolution and radiation-free morphologic imaging; however, its image quality is still lower than that of CT. This retrospective study aims to assess the image quality and clinical applicability of synthetic CT images generated from lungs UTE MRI by a generative adversarial network (GAN), in patients with cystic fibrosis. Paired MRI and CT slices acquired on the same day at one of six participant centers, including 156 patients suffering from cystic fibrosis were dispatched in a training (82 patients), test (28 patients) and validation data sets. Image quality was evaluated on quantitative et qualitative parameters, while CF-related structural abnormalities were depicted by two readers in order determine Bhalla severity scores. Synthetic CT Contrast-to-noise ratio was significatively higher than that of UTE MRI in both test and external data sets, signal-to-noise ratio was similar between synthetic CT and real CT, and overall noise level was inferior to that of real CT. Synthetic CT images had the lowest level of artifacts. Concordance between Bhalla scores on synthetic and real CT images was almost perfect. Thus, synthetic CT images showed an almost perfect concordance with real CT images for depiction of cystic fibrosis related structural lung abnormalities, and had a better image quality than UTE IRM.
Source: HAL
Author: Raoult, Julien, Charles Roger
Institution: Université de Bordeaux
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7010030298
