Deep learning-based super-resolution and denoising algorithm improves reliability of dynamic contrast-enhanced MRI in diffuse glioma

Abstract

Dynamic contrast-enhanced MRI (DCE-MRI) is increasingly used to non-invasively image blood-brain barrier leakage, yet its clinical utility has been hampered by issues such as noise and partial volume artifacts. In this retrospective study involving 306 adult patients with diffuse glioma, we applied deep learning-based super-resolution and denoising (DLSD) techniques to enhance the signal-to-noise ratio (SNR) and resolution of DCE-MRI. Quantitative analysis comparing standard DCE-MRI (std-DCE) and DL-enhanced DCE-MRI (DL-DCE) revealed that DL-DCE achieved significantly higher SNR and contrast-to-noise ratio (CNR) compared to std-DCE (SNR, 52.09 vs 27.21; CNR, 9.40 vs 4.71; P < 0.001 for all). Diagnostic performance assessed by the area under the receiver operating characteristic curve (AUROC) showed improved differentiation of WHO grades based on a pharmacokinetic parameter [Formula: see text] (AUC, 0.88 vs 0.83, P = 0.02), while remaining comparable to std-DCE in other parameters. Analysis of arterial input function (AIF) reliability demonstrated that [Formula: see text] exhibited superior agreement compared to [Formula: see text], as indicated by mostly higher intraclass correlation coefficients (Time to peak, 0.79 vs 0.43, P < 0.001). In conclusion, DLSD significantly enhances both the image quality and reliability of DCE-MRI in patients with diffuse glioma, while maintaining or improving diagnostic performance.

Keywords: Deep learning; Glioma; Image enhancement; Magnetic resonance imaging; Perfusion.

Fig. 1

Boxplots illustrating SNR and CNR measurements across various ROIs for both std-DCE and DL-DCE. SNR and CNR values are compared between the two images for CE, NE, and WT regions. Asterisks (****) indicate statistically significant differences (P < 0.001).

Fig. 2

Representative images of DCE-MRI at baseline (a) and after contrast agent arrival (b) from std-DCE and DL-DCE. 43-year-old male patient with glioblastoma, IDH-wildtype in the left frontal-parietal lobe: The alternating columns provide an enlarged view of the tumor ROI extracted from the axial, sagittal and coronal image. The DL-DCE images demonstrate enhanced resolution and reduced noise compared to std-DCE, particularly highlighting detailed tumor features and vascular structures.

Fig. 3

Representative images of the DCE-MRI (first column), (second column), (third column), and (last column) of std-DCE and DL-DCE. A 52-year-old male patient with glioblastoma, IDH-wildtype in the right temporal lobe: The alternating rows provide an enlarged view of the tumor ROI extracted from the complete axial image. The DL-DCE images exhibited reveal heightened structural complexity, achieved through enhanced resolution and concurrent reduction of the heterogeneous noise seen in std-DCE images.

Fig. 4

(A) ROC curves for distinguishing low-grade and high-grades using PK parameters acquired from both std-DCE and DL-DCE. Boxplots display the distribution of PK parameter values. (B) ROC curves for discriminating IDH mutation using PK parameters obtained from both std-DCE and DL-DCE. Boxplots illustrate the distribution of PK parameter values. The AUC values are shown for , , and , indicating their diagnostic performance. Asterisks (****) indicate statistically significant differences (P < 0.001).

Fig. 5

The upper row displays the AIF curves for all 306 patients. Graphs of the lower row are plotted using the median, first quartile, and third quartile values from and which are AIFs obtained from std-DCE and DL-DCE, respectively. The median exhibits a sharper wash-in slope and higher maximum signal intensity compared to the median .