Detecting Dopamine Release via PCA of Residuals

Dopamine release can be detected in vivo using [ ¹¹ C]raclopride dynamic PET imaging: raclopride competes for binding with dopamine and thus a voxel-level change in dopamine concentration alters the voxel-level PET temporal signal. Current methods to detect dopamine release suffer from low sensitivity, primarily due to image noise and relying on model comparison methods which, at present, poorly separate true positives from true negatives. We propose an alternative, data-driven method that regresses denoised voxel-level signals on regional signals, producing structured residuals where dopamine release is present. Principal component analysis of these residuals is then used to localize the release. Simulation studies demonstrate that our proposed PCA method yields more reliable and conformal detection of dopamine release than the model comparison method across a variety of denoising kernel sizes, with 1.5-2x higher sensitivity at equivalent false positive rate.