AI Improves MR Image Quality with Higher Acquisition Speeds
Amish Doshi, MD presents insights into the Clinical Benefits of AI in MR Imaging.
AI Improves MR Image Quality with Higher Acquisition Speeds
Artificial intelligence (AI) solutions are regularly developed and introduced into medical imaging. In MR imaging, these technologies are designed to enhance acquisition time and image quality. Mount Sinai Health Systems in New York City has implemented several AI applications for MR imaging. These applications leverage super-resolution and denoising techniques to enhance the quality of images obtained with accelerated MR protocols.
Amish Doshi, MD, chief of neuroradiology at Mount Sinai, reports improvements in MR acquisition times and patient throughput where they have over 20 MR scanners. “AI can translate into a significant increase in volume with multiple benefits, including improving the patient experience by reducing the amount of time a patient needs to be in the MRI scanner,” said Dr. Doshi.
Quantitative and Qualitative Benefits
Citing a case of a patient with multiple sclerosis, Dr Doshi said this AI software reduced the length of a sequence from four minutes to two minutes. In a lumbar spine study, a conventional MR scan took one minute 20 seconds, while the AI-enhanced reconstruction took less than a minute, with decreased noise compared to the conventional scan.
“The noise on these images is improved with an AI denoising algorithm that delivers significant improvement in time savings. This allows us to maintain and, in some circumstances, enhance image quality,” Dr. Doshi said.
This AI solution also enables radiologists to add sequences and increase data collection without increasing exam duration significantly; several of the protocols Dr. Doshi and his team use have shown an overall 20% reduction in scan time between conventional and AI-based reconstructions. In particular, lumbar spine MR studies have demonstrated a 46% reduction in scan time between conventional and AI studies.
“When we look at the durations of the protocols between AI and conventional neuro protocols, there’s a significant time savings with the AI-based reconstruction protocols compared to the conventional,” he said.
As a result, the radiology department has reduced its MR exam times from regularly scheduled 30-minute slots down to 20 minutes, and from regularly scheduled 60-minute time slots down to 40 minutes. For example, a MR brain previously took 30 minutes to complete. With this AI tool, it now takes 20 minutes. Now, Mount Sinai Health Systems can potentially complete three MR neuro scans in one hour, an increase of an additional scan every hour.
“This resulted in a significant increase in the number of scans performed on the various AI-based scanners,” he said. “Now we can offer more examination times for patients and better accommodate the demand for imaging.” Dr. Doshi said. “We also have the option to perform two exams an hour, providing additional time for technologists, staff and patients.”
As AI continues to impact medical imaging, clinicians can expect to see continuous improvements in every part of the radiology department, from scheduling and reporting to image quality and scan time.