Tuesday, December 1 | 3:20 p.m.- 3:40 p.m. | T7-SSNR09-3 | Room E350
Can the use of deep learning (DL) help reduce gadolinium-based contrast agent (GBCA) dose on MRI? It appears so, according to research to be presented Tuesday afternoon.
GBCAs are essential for lesion characterization, treatment planning, and monitoring in brain MRI, but their use has prompted concerns regarding long-term gadolinium retention in the brain and increasing environmental concentrations in groundwater, presenter Thomas C. Arnold, PhD, of Subtle Medical, and colleagues noted. To address these concerns, experts have made efforts to reduce GBCA doses -- without compromising image quality.
In his presentation, Arnold will share findings from a study that evaluated a deep-learning algorithm that was designed to restore contrast enhancement from 10% of the standard dose to match full-dose quality on brain MRI exams. The study included data from 63 patients; each individual received a low gadolinium dose (10%) and low-dose T1 imaging after a pre-contrast T1 sequence, then the remainder of the GBCA dose from which the researchers obtained full-dose T1 images.
The algorithm processed pre-contrast and low-dose images to produce what the team dubbed "enhanced low-dose" images, then two radiologists independently reviewed either the standard-of-care or the enhanced low-dose images in a blinded, randomized study design. The radiologists categorized lesions (enhancing, non-enhancing, or normal) and rated lesion characteristics of contrast enhancement, border delineation, and internal morphology on four-point Likert scale.
Interreader agreement for standard-of-care images and intrareader agreement between standard-of-care and enhanced low-dose images showed no significant difference (p = 0.128), the team reported. There were also no significant differences between standard-of-care GBCA doses and enhanced low-dose images when it came to lesion characteristics.
