Abstract
BACKGROUND: Deep-learning (DL) reconstructions could improve image quality and reduce acquisition time in diffusion-weighted imaging (DWI). This study assessed, qualitatively and quantitatively, DL-DWI in liver metastasis of colorectal cancer patients. METHODS: This prospective study enrolled 50 participants from June to November 2022. Phantom and participant data were acquired on a 1.5T MR scanner using a free-breathing DL-DWI research application sequence. Three DWIs were compared: a moderately-accelerated DL-DWI (DL-1), a corresponding standard reconstruction (Standard-1) and a highly-accelerated DL-DWI (DL-2). Image quality (four features on b750 images and one feature on ADC map) was assessed by two radiologists. Region of interest (ROI) based ADC measurements were performed at three locations: liver, spleen, liver metastasis. Across the three series, median scores and ADC values were assessed using a Friedman non-parametric test and post-hoc analysis (pairwise Wilcoxon tests with Bonferroni correction). A p-value < 0.05 was considered statistically significant. RESULTS: Fifty participants with metastatic colorectal cancer (mean age 62 years, range 36-88 years, 26 males) were evaluated. ROIs were delineated in liver (N = 50), spleen (N = 48), and liver metastasis (N = 11). Qualitatively, across both readers, DL-1 method received the highest scores for 5/8 features on the b750 images; all methods scored similarly on ADC maps for both readers. Quantitatively, ADCs were significantly different between DL-1 and Standard-1 series across all three organs, with DL-1-based ADC always higher (p < 0.01). This ADC increase was small: 8.9% (liver), 3.4% (spleen), 4.5% (liver metastasis). CONCLUSIONS: This study suggests that a DL-based reconstruction is a promising technique to enable acceleration of liver DWI considering both qualitative and quantitative results. TRIAL REGISTRATION: NCT05118555 (Evaluation of New Magnetic Resonance Techniques); study date of registration (first submitted: 2021-10-18).