A recent study from NYU Grossman School of Medicine and Meta AI Research shows that artificial intelligence (AI) can reconstruct fast, thick-sample MRI images into high-quality images with the same diagnostic value as those created with standard MRI. According to the study, MRI images may be available to more patients, and waiting times for appointments could be cut in half if they were reconstructed using AI instead of traditional methods. Meta AI researchers and NYU Langone imaging experts and radiologists collaborated on an AI model to speed up MRI. It also generated the world’s largest repository of raw MRI data, which researchers and developers have used in various fields.
NYU Langone scientists removed nearly three-quarters of the raw data collected by traditional, slow MRIs to replicate the sped-up scans in an earlier “proof-of-principle” study. Faster MRI scans were used to train an artificial intelligence model that produced images that would otherwise be indistinguishable from those produced by slower scans. Similar to how the brain builds images by filling in missing visual information from local context and past experiences, the researchers in this new study performed accelerated scans with only a quarter of the actual data and used the AI model to “fill in images.” “The missing information. The fastMRI scans were accurate and of higher quality than conventional scans in both experiments.
We used 298 clinical 3-T knee evaluation images to train a DL reconstruction model. Between January 2020 and February 2021, patients clinically referred for knee MRI completed a 3T accelerated conventional knee MRI protocol followed by an accelerated LD procedure for a prospective study. Whether or not the reconstructed DL images were interchangeable with traditional images in identifying abnormalities was determined. Six musculoskeletal radiologists observed each exam. Ordinal scores from 0 to 4 were used in analyzes evaluating the probability of abnormalities in meniscal or ligament tears, bone marrow, and cartilage. Overall image quality, artifacts, sharpness, and signal-to-noise ratio were evaluated, and four-point ordinal values were used to compare the methods.
170 people (mean age SD: 45 16; 76 men) were evaluated. The DL-reconstructed photos were found to be as good as traditional images at detecting anomalies. On average, DL photos received a higher quality rating among six readers than traditional photos (P .001). Radiologists agreed that AI-reconstructed images were just as good as traditional images for diagnosing tears and abnormalities. They also decided that faster scans had much higher image quality overall.
The researchers emphasize that unique tools are not needed to perform FastMRI. Standard MRI machines can be programmed to collect less data than is often required. The fastMRI effort has published its data, models, and code as an open source project for use by other researchers and manufacturers of commercial MRI systems.
With fastMRI, the time it takes to perform an MRI, which can take up to 30 minutes, is reduced to less than 5 minutes, which is equal to the time it takes to perform an X-ray or CT scan. Unlike these other imaging modalities, however, MRI offers a wealth of information, including visualization of soft tissue from numerous angles, identification of microscopic cartilage abnormalities, and identification of abdominal malignancies.
In conclusion, deep-learning reconstruction of prospectively accelerated knee MRI enabled a nearly two-fold reduction in scan time, improved image quality, and had the same diagnostic utility compared with traditional reconstruction. Deep learning reconstruction has been shown to cut the scan time for a knee MRI in half compared to the standard procedure without sacrificing diagnostic accuracy.
review the Paper Y Reference article. All credit for this research goes to the researchers of this project. Also, don’t forget to join our reddit page, discord channel, Y electronic newsletterwhere we share the latest AI research news, exciting AI projects, and more.
Dhanshree Shenwai is a Computer Engineer and has good experience in FinTech companies covering Finance, Cards & Payments and Banking domain with strong interest in AI applications. She is enthusiastic about exploring new technologies and advancements in today’s changing world, making everyone’s life easier.
