Ultrafast Brain MRI with Deep Learning Reconstruction for Suspected Acute Ischemic Stroke
Sebastian Altmann1, Nils F Grauhan1, Lavinia Brockstedt1
1From the Department of Neuroradiology (S.A., N.F.G., L.B., M.K., M.A.B., A.E.O.), Institute of Medical Biostatistics, Epidemiology and Informatics (I.S., R.P.), and Department of Neurology (T.U., S.G.), University Medical Center Mainz, Johannes Gutenberg University, Langenbeckstr 1, 55131 Mainz, Germany; Siemens Medical Solutions USA, Boston, Mass (B.C.); and Siemens Healthcare, Erlangen, Germany (T.F., Z.H.).
Deep learning-accelerated MRI significantly reduces scan times for suspected acute stroke patients. This faster technique is interchangeable with conventional MRI for detecting acute ischemic lesions, offering comparable diagnostic performance.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Medicine
Background:
- Deep learning (DL) acceleration in MRI can drastically cut examination durations.
- Prospective studies assessing DL-accelerated MRI diagnostic performance in acute stroke are limited.
- Investigating DL-accelerated MRI's role in acute ischemic stroke diagnosis is crucial.
Purpose of the Study:
- To evaluate the interchangeability of DL-accelerated MRI with conventional MRI for diagnosing acute ischemic stroke at 1.5 T.
- To compare diagnostic performance, including lesion detection and secondary findings, between the two MRI techniques.
- To assess image quality and diagnostic confidence using DL-accelerated MRI.
Main Methods:
- A prospective study involving 211 patients with suspected acute stroke at 1.5 T.
- Comparison of conventional MRI (14 min 18 sec) with DL-accelerated MRI (3 min 4 sec) using identical sequences.
- Analysis of primary endpoint (interchangeability for infarction detection) and secondary endpoints (vascular territory, secondary findings, image quality, confidence) by three readers.
Main Results:
- DL-accelerated MRI demonstrated interchangeability with conventional MRI for detecting acute ischemic infarction (Individual Equivalence Index [IEI] -0.002, within 5% margin).
- Interchangeability was also confirmed for affected vascular territory and clinically relevant secondary findings, with almost perfect interrater agreement (P > .91).
- DL-accelerated MRI yielded superior image quality (P < .001) and diagnostic confidence (P < .001), with good to excellent interrater reliability for lesion signal properties (ICC ≥0.8).
Conclusions:
- DL-accelerated brain MRI is a viable alternative to conventional MRI for acute ischemic lesion detection, offering significant time savings.
- The faster DL-accelerated technique maintains diagnostic accuracy and is interchangeable with standard MRI protocols.
- This advancement holds potential for improving efficiency in acute stroke imaging workflows.
