Researchers led by presenter Dr. Lawrence Tanenbaum of imaging services provider RadNet prospectively evaluated a commercially available k-space-based reconstruction application (Air Recon DL, GE Healthcare) in a study involving 26 consecutive patients and two neuroradiologists. All patients received standard-of-care, as well as accelerated spine MRI exams -- 72% faster for cervical scans and 64% faster for lumbar scans.
Next, the two readers evaluated the different image series paired in side-by-side datasets, which were blinded and randomized in sequence and also left-to-right display order. They ranked the image features on a 5-point Liker scale.
The AI-reconstructed images not only were assessed to be qualitatively better but also yielded high interrater agreement. In addition, no aberrations were detected on these images.
"[Deep learning] enables 64% to 72% spine MRI scan time reduction as well as what radiologists perceive as enhanced image quality with benefits in [signal-to-noise ratio], image sharpness and artifact reduction over [standard-of-care] and [accelerated] images without [deep-learning] processing, providing gains in efficiency and portending practice utility for routine use," the authors wrote.
Accelerate your AI education by attending this Thursday talk.
This paper received a Roadie 2021 award for the most popular abstract by page views in this Road to RSNA section.
