Objective: To develop a fully automated quantitative method to analyse routinely acquired ¹²³I-ioflupane (DaTSCAN) data to aid diagnostic predictions in patients with Parkinson’s disease, non-degenerative Parkinsonism and tremor disorders.

Background: Diagnosis of Parkinson’s disease remains challenging with the lack of a biomarker, particularly in early disease or if atypical features are present. DaTSCANs have an established role in clinically uncertain Parkinsonian syndromes but they are usually subjectively assessed, which are rater dependent, limiting reliability and clinical utility.

Method: DaTSCAN data was acquired from a specialist movement disorder clinic and analysed (N=114).  We tested three hypotheses: i) qualitatively reported scans differed quantitatively at a group level, ii) individual scan reports and diagnosis could be predicted accurately using computational methods, and iii) degenerative vs non-degenerative disease could be predicted for clinically ambiguous scans reported as unilaterally abnormal.  A fully automated method was used to anatomically standardise each scan and extract relative tracer uptake values for striatal sub-regions.  Individual scan and patient predictions were done using linear and quadratic discriminant analyses with 5 fold cross-validation.

Results: Scans reported as bilaterally abnormal had significantly reduced uptake in all striatal sub-regions compared to normal reported scans for each hemisphere at a group level.  Inconsistent group level differences were found for unilateral reported scans.  It was possible to predict bilateral abnormal vs normal report at 85% accuracy.  Applying this predictive model to scans reported as unilaterally abnormal resulted in 71% diagnostic accuracy for degenerative vs non-degenerative disease.  Degenerative from non-degenerative tremor syndrome was predicted at 81% accuracy if only Parkinson’s disease and tremor disorder were considered.

Conclusion: Prediction of individual scan and diagnostic categories achieved relatively high accuracy and more sophisticated image processing and prediction methods would likely improve this significantly.  By increasing accuracy in predicting diagnosis and outcomes in clinically uncertain situations, it is hoped this will enhance patient management and cost-effectiveness, allowing earlier and better targeted treatments.

« Back to MDS Virtual Congress 2020

MDS Abstracts - https://www.mdsabstracts.org/abstract/quantification-of-123i-ioflupane-striatal-uptake-and-individual-patient-diagnostic-predictions-degenerative-parkinsons-disease-non-degenerative-parkinsonism-and-tremor-disorders/