Low energy virtual monochromatic CT with deep learning image reconstruction to improve delineation of endoleaks.
Clin Radiol
; 79(10): e1260-e1267, 2024 Oct.
Article
en En
| MEDLINE
| ID: mdl-39079807
ABSTRACT
AIM:
This study aimed to investigate the utility of low-energy virtual monochromatic imaging (VMI) combined with deep-learning image reconstruction (DLIR) in improving the delineation of endoleaks (ELs) after endovascular aortic repair (EVAR) in contrast-enhanced dual-energy CT (DECT).METHODS:
A total of 61 consecutive patients (mean age, 77 years; 46 men) after EVAR who underwent contrast-enhanced DECT were enrolled. Virtual monochromatic 40- and 70-keV images were reconstructed using DLIR (TrueFidelity-H) and conventional hybrid iterative reconstruction (IR). Contrast-to-noise ratio (CNR) of the EL on the venous-phase CT was calculated. Four different reconstructed image series (hybrid IR and DLIR at two energy levels, 40- and 70-keV) were displayed side-by-side and visually assessed for EL conspicuity on a 5-point comparative scale from 0 (best) to -4 (significantly inferior). Two experienced radiologists independently conducted a qualitative evaluation of the CT images.RESULTS:
A total of 30 out of 61 patients presented with an EL. On both 40- and 70-keV images, the CNR of the EL was significantly higher in DLIR than in hybrid IR (40-keV, 14.5 ± 7.3 vs 8.6 ± 4.2, P<0.001; 70-keV, 8.7 ± 4.5 vs 5.5 ± 2.6, P<0.001). The comparative scale of EL conspicuity in the 40-keV DLIR images (Observer1, -0.2 ± 0.4; Observer2, 0.0 ± 0.0) was significantly higher than 40-keV hybrid IR (Observer1, -0.5 ± 0.5; Observer2, -1.0 ± 0.0; P<0.05), 70-keV DLIR (Observer1, -1.8 ± 0.4; Observer2, -2.0 ± 0.0; P<0.001) and 70-keV hybrid IR images (Observer1, -1.8 ± 0.4; Observer2, -2.4 ± 0.5; P<0.001), respectively.CONCLUSIONS:
Using 40-keV VMI in combination with DLIR improves EL delineation after EVAR compared with the 70-keV VMI with hybrid IR or DLIR.
Texto completo:
1
Colección:
01-internacional
Base de datos:
MEDLINE
Asunto principal:
Tomografía Computarizada por Rayos X
/
Endofuga
/
Aprendizaje Profundo
Límite:
Aged
/
Aged80
/
Female
/
Humans
/
Male
Idioma:
En
Revista:
Clin Radiol
Año:
2024
Tipo del documento:
Article
Pais de publicación:
Reino Unido