Two-minute spine MRI tests AI at both ends of the imaging workflow

A prospective study of 70 patients tested two places where AI could shorten lumbar-spine imaging: reconstructing an accelerated scan and generating a structured report from it. The researchers reported diagnostic images acquired within two minutes, while the reporting system showed substantial to almost perfect agreement with radiologist consensus on the assessed findings.[1]

Seventy patients, 140 scans

Each patient received both protocols in the same session, producing 70 paired comparisons and 140 scans in total. Four blinded radiologists assessed image quality. The researchers examined interchangeability across five key pathological findings and reported acquisition-time reductions of approximately 80–86%, with comparable subjective image quality and improved signal-to-noise and contrast-to-noise measures.[1]

That is an operational gain within the tested acquisition protocol. It does not mean an entire MRI appointment takes two minutes: positioning, preparation, any additional sequences and reporting are separate parts of the visit. Nor did the study measure an 80–86% reduction in waiting lists.[1]

A useful image and a useful report

Fast image acquisition and automated interpretation use different evidence. Reconstruction must preserve clinically important anatomy. A reporting model must identify findings and their severity, while its output still needs to be considered alongside the patient’s clinical history. Comparing reports with radiologist consensus examines that second task separately from the speed of acquisition.[1]

The accessible author abstract describes agreement as substantial to almost perfect, without supplying the finding-by-finding coefficients. It supports the feasibility claim, but not a numerical sensitivity claim or a conclusion that automated reporting can replace radiologist review.[1]

The study appeared online in January 2026 and in the journal’s March volume. Its continuing relevance is the separation of two useful goals: acquiring diagnostic images sooner and producing a reliable account of what they show. A faster scan can improve one stage of care while the reporting stage still needs its own assessment.[1]

Report an error or suggest a correction