Motivation: Diagnosing parkinsonism is challenging, especially in early stages, with misdiagnosis rates around 25%. Nigrosome-1 MRI and deep learning (DL) models may improve diagnostic accuracy. Goal(s): To report the clinical utility of nigrosome-1 MRI and DL models for automated nigrosome-1 classification and quantification against long-term clinical outcomes in patients presenting with parkinsonism. Approach: Retrospective review of patient records for clinical outcomes at year-3. Results: Radiological assessment, Heuron IPD binary classification and Heuron NI nigrosome-1 quantification showed strong concordance with clinical outcomes (AUC: radiologist=0.87, Heuron NI=0.84, Heuron IPD=0.82). Volume analysis with Heuron NI distinguished idiopathic Parkinson's disease from non-neurodegenerative conditions (p<0.005). Impact: Nigrosome-1 MRI in the radiologic clinic aids in differentiating idiopathic Parkinson's disease from non-neurodegenerative conditions when clinical presentation is unclear. Deep learning models enhance accessibility and show good potential as objective adjunctive imaging tools easing radiological workflow for accurate diagnosis.
Patidar et al. (Tue,) studied this question.