TY - JOUR

T1 - Different preprocessing strategies lead to different conclusions

T2 - A [11C]DASB-PET reproducibility study

AU - Nørgaard, Martin

AU - Ganz, Melanie

AU - Svarer, Claus

AU - Frokjaer, Vibe G

AU - Greve, Douglas N

AU - Strother, Stephen C

AU - Knudsen, Gitte M

PY - 2020

Y1 - 2020

N2 - Positron emission tomography (PET) neuroimaging provides unique possibilities to study biological processes in vivo under basal and interventional conditions. For quantification of PET data, researchers commonly apply different arrays of sequential data analytic methods ("preprocessing pipeline"), but it is often unknown how the choice of preprocessing affects the final outcome. Here, we use an available data set from a double-blind, randomized, placebo-controlled [11C]DASB-PET study as a case to evaluate how the choice of preprocessing affects the outcome of the study. We tested the impact of 384 commonly used preprocessing strategies on a previously reported positive association between the change from baseline in neocortical serotonin transporter binding determined with [11C]DASB-PET, and change in depressive symptoms, following a pharmacological sex hormone manipulation intervention in 30 women. The two preprocessing steps that were most critical for the outcome were motion correction and kinetic modeling of the dynamic PET data. We found that 36% of the applied preprocessing strategies replicated the originally reported finding (p < 0.05). For preprocessing strategies with motion correction, the replication percentage was 72%, whereas it was 0% for strategies without motion correction. In conclusion, the choice of preprocessing strategy can have a major impact on a study outcome.

AB - Positron emission tomography (PET) neuroimaging provides unique possibilities to study biological processes in vivo under basal and interventional conditions. For quantification of PET data, researchers commonly apply different arrays of sequential data analytic methods ("preprocessing pipeline"), but it is often unknown how the choice of preprocessing affects the final outcome. Here, we use an available data set from a double-blind, randomized, placebo-controlled [11C]DASB-PET study as a case to evaluate how the choice of preprocessing affects the outcome of the study. We tested the impact of 384 commonly used preprocessing strategies on a previously reported positive association between the change from baseline in neocortical serotonin transporter binding determined with [11C]DASB-PET, and change in depressive symptoms, following a pharmacological sex hormone manipulation intervention in 30 women. The two preprocessing steps that were most critical for the outcome were motion correction and kinetic modeling of the dynamic PET data. We found that 36% of the applied preprocessing strategies replicated the originally reported finding (p < 0.05). For preprocessing strategies with motion correction, the replication percentage was 72%, whereas it was 0% for strategies without motion correction. In conclusion, the choice of preprocessing strategy can have a major impact on a study outcome.

U2 - 10.1177/0271678X19880450

DO - 10.1177/0271678X19880450

M3 - Journal article

C2 - 31575336

VL - 40

SP - 1902

EP - 1191

JO - Journal of Cerebral Blood Flow and Metabolism

JF - Journal of Cerebral Blood Flow and Metabolism

SN - 0271-678X

IS - 9

ER -