Multimodal Imaging Signatures of Parkinson's Disease

Abstract

Parkinson's disease (PD) is a complex neurodegenerative disorder that manifests through hallmark motor symptoms, often accompanied by a range of non-motor symptoms. There is a putative delay between the onset of the neurodegenerative process, marked by the death of dopamine-producing cells, and the onset of motor symptoms, creating an urgent need to develop biomarkers that may yield early PD detection. Neuroimaging offers a non-invasive approach to examining the potential utility of a vast number of functional and structural brain characteristics as biomarkers. We present a statistical framework for analyzing neuroimaging data from multiple modalities to determine features that reliably distinguish PD patients from healthy control (HC) subjects. Our approach builds on elastic net, performing regularization and variable selection, while introducing additional criteria centering on parsimony and reproducibility. We apply our method to data from 42 subjects (28 PD patients and 14 HC). Our approach demonstrates extremely high accuracy, assessed via cross-validation, and isolates brain regions that are implicated in the neurodegenerative PD process.

Keywords: MRI; Parkinson's disease; biomarker; classification; multimodal imaging; penalized regression; prediction.

Figure 1

Characterization of the onset and progression of Parkinson's disease neurodegeneration (green and yellow), which persists through the commencement of motor symptoms (orange) and ultimately clinical diagnosis and beyond (red).

Figure 2

Depiction of AAL-90 parcellation and a hierarchical subparcellation with 290 brain regions. The subregions are constructed from resting state fMRI data of healthy controls (outside of the current sample) based on functional characteristics with anatomical constraints to keep subregions contiguous and bounded within a single region.

Figure 3

Overview of the multiple modalities generating data, estimates obtained from each reflecting particular structural or functional properties, spatial scale for summary data representations, and ultimately the features constituting the global set of potential neuroimaging markers of PD.

Figure 4

(A) AUC for different tuning parameters, with each point averaged over 100 applications of two-fold cross-validation. The point A reflects the tuning parameter value yielding the maximum AUC, and is depicted in the curves in (B). The traces define a restricted space of tuning parameters. Above and to the right of the white trace yields no more than an average of 75 predictors, and below and to the left of the black trace reflects at least 0.90 AUC on average. (B) ROC curve (in black) reflecting high prediction accuracy based on 271 imaging predictors; AUC is 0.989. The colored curves highlight the variability associated with each separate CV sample.

Figure 5

(A) Plots of the mean absolute coefficient (standardized) vs. the proportion of times the feature is retained over 200 training samples at (α, λ) corresponding to points (B–E) in Figure 4A. The enlarged plot shown in (B) is point E from Figure 4A, with colors depicting the modality. The reference lines in all plots reveal the 10% of values with strongest predictive power over the training samples. At point E, modalities FC, SC, and VBM yield the most predictive features.

Figure 6

Models achieving perfect separation between PD patients and HC subjects with a minimum number of variables. Each three feature model is adjusted for age, sex, and head coil. The models are comprised of eight distinct features.