Harmonization of Multisite Dopamine Transporter Single-photon Emission Computed Tomography Data for Segregation of Patients with Parkinson's Disease from Healthy Older Adults

Abstract

Abstract Dopamine transporter single-photon emission computed tomography (DAT-SPECT) can aid in selecting and stratifying participants in large-scale Parkinson’s disease (PD) clinical studies. Whether differences in DAT-SPECT methodology across institutes in multicenter cohort studies affect outcomes remains unclear. We assessed the impact of multisite data harmonization on differentiating patients with PD from healthy older adults. We obtained DAT-SPECT specific binding ratios (SBRs) in 72 healthy older adults and 81 patients with PD from four centers. We evaluated three SBR correction techniques: prospective correction using standard phantom scanning (scanner correction), standardized SBR computation (operation standardization), a combination of the two, and data-driven correction using a statistical method. We examined changes in the SBRs and diagnostic accuracy using the area under the receiver operating characteristic curve (AUC-ROC). Multisite harmonization improved the outcome, although the SBRs without correction effectively distinguished patients with PD from healthy controls (Hedge’s g = 2.82; AUC-ROC = 0.926). Combined scanner and operation correction had the largest improvement, followed by data-driven correction (g = 4.32 and 3.99 and AUC-ROC = 0.992 and 0.987, respectively). Our findings suggest that harmonization further improves outcomes in multicenter cohorts. Prospective correction with phantom scanning and operation standardization is ideal for SBR robustness and interpretability. This data-driven correction is a potential alternative method.