AI analysis of nigral hyperintensity on susceptibility MRI to predict dopamine transporter PET abnormalities in REM sleep behavior disorder

Deep learning analysis of nigral hyperintensity on susceptibility MRI predicted dopamine transporter PET abnormalities in REM sleep behavior disorder with 78% sensitivity, 66.7% specificity, 73% accuracy. Bilateral MRI findings strongly linked to PET positivity (24/26, 92.3%).

• In 74 patients with idiopathic REM sleep behavior disorder, a deep learning model on susceptibility map-weighted imaging (SMwI) aimed to predict dopamine transporter (DAT) PET abnormalities.
• The model achieved 78% sensitivity, 66.7% specificity, and 73% accuracy at patient and hemisphere levels.
• Bilateral SMwI abnormalities predicted DAT PET positivity in 92.3% (24/26), while 35.5% of visually normal SMwI cases still had PET abnormalities.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.