Neuro / Head & NeckAI / InformaticsResearch
Deep learning-accelerated 3D FLAIR reduces scan time and improves noise quality for cerebral white matter lesions
Acta radiologica (Stockholm, Sweden : 1987)today
Deep learning-accelerated 3D FLAIR reduced scan times by 15% at 1.5T and 35% at 3T versus standard 3D FLAIR, with significantly improved noise quality at both field strengths and comparable lesion conspicuity.
- At 1.5T, SDL FLAIR (4 min 53 sec) reduced scan time by 15% vs standard (5 min 43 sec), with significantly improved noise quality (P<0.05).
- At 3T, SDL FLAIR (2 min 45 sec) reduced scan time by 35% vs standard (4 min 15 sec); lesion conspicuity improved but not statistically significant except for noise.
- Automated lesion volumetric measurements were comparable between SDL and standard FLAIR; interrater agreement for image quality was moderate to almost perfect (κ=0.55-0.85).
Automated summary
RadPigeon summaries are original and for information only. They are not clinical advice.