Our legacy-funded MRI service drastically reduces scan times, while enhancing image quality, with advanced new software
Cancer imaging is taking a leap forward, thanks to the fusion of advanced magnetic resonance imaging (MRI) techniques and deep learning being pioneered at Paul Strickland Scanner Centre. The software is being used on our two new MRI scanners, which were installed in 2022 and made possible by significant legacy donations.
Recent research co-authored by three members of our clinical team showcases a transformative approach to whole-body diffusion-weighted imaging (WB-DWI) that significantly enhances image quality while cutting scan times by more than half. Alongside a number of other experts in the field, our Deputy Superintendent Radiographer for MRI Mr Will ,McGuire, MRI Superintendent Radiographer Ms Marie Fennessy and our lead consultant for MRI, Prof Anwar Padhani, report on their findings in a paper that has just been published in European Radiology, a leading peer-reviewed academic journal.
The promise of Whole-Body MRI
Over the past decade, WB-MRI has emerged as a powerful, non-invasive tool for cancer imaging. It provides a comprehensive view of both anatomical structures and functional processes, making it invaluable for staging cancer and assessing treatment responses. The technique is increasingly recognised not only for its role in cancer but also for its potential in screening and evaluating inflammatory conditions.
One key component of WB-MRI is diffusion-weighted imaging (DWI). This technique measures the movement of water molecules
within tissues, which can indicate the presence of cancer cells. Traditional WB-DWI, while effective, requires long scan times and
therefore isn’t as readily available in other hospitals around the world.
Enter Deep Learning
The study delves into the efficacy of a deep learning-accelerated WBDWI technique, known as Deep Resolve Boost (DRB). This method employs advanced algorithms to rapidly produce high-quality images from MRI data.
In the study, 50 patients with cancer in the bone marrow underwent WB-MRI scans using both traditional and DRBaccelerated DWI sequences. Radiologists compared the two sets of images, evaluating them based on several criteria: noise levels, artifacts, signal suppression, and the visibility of lesions.
Superior Image Quality and reduced scan times
The results were striking. In nearly 80 per cent of cases, radiologists preferred the DRB images over the conventional ones. This preference was particularly pronounced in patients with a higher body mass index (BMI). The DRB technique consistently produced images with fewer artifacts and better signal-tonoise ratios, which are critical for accurately identifying cancerous lesions. Quantitative assessments backed up these findings. The signal-to-noise and contrast-tonoise ratios were significantly
higher in DRB images for all normal tissues. Although the apparent diffusion coefficient (ADC) values, which help quantify tissue
diffusivity, were slightly higher in normal tissues for DRB images, they did not differ for cancerous lesions. This consistency is crucial, as it ensures that the new technique does not compromise diagnostic accuracy.
Moreover, the acquisition time for DRB sequences was reduced by over 50%, dropping from 14 minutes to just under 7 minutes.
This reduction is not just a matter of convenience; shorter scan times can significantly improve patient comfort and throughput in busy clinical settings.
Implications for Clinical Practice
The integration of deep learning into WB-DWI represents a significant advancement in medical imaging. The improved image quality and reduced scan times can enhance patient experience and streamline workflow in radiology departments. For patients, especially those undergoing frequent scans, the reduction in time spent in the MRI machine can alleviate discomfort and anxiety.
By providing high-quality images more rapidly, this technology has the potential to make WB-MRI more accessible and practical for routine use in cancer care. It could also facilitate more widespread adoption of WB-MRI for screening and monitoring purposes and reinforces the ongoing mission of Paul Strickland Scanner Centre to improve practice beyond our organisation.
Looking Ahead
While the study is a promising step forward, it is just the beginning. As deep learning algorithms continue to evolve, we can expect even greater improvements in image quality and further reductions in scan times.
Future research will likely focus on expanding the application of these techniques to other types of MRI and further refining the algorithms to enhance diagnostic accuracy.
