Imaging Biomarkers of Cognitive Impairment in Parkinson’s Disease

Neurology Medicine Medical Imaging
Parkinson's disease Cognitive decline Imaging biomarkers [18F]FDG PET MRI Neurodegeneration Diffusion-weighted imaging

Multimodal Neuroimaging Study on Cognitive Impairment in Parkinson’s Disease

Parkinson’s Disease (PD) is a common neurodegenerative disorder mainly characterized by motor impairments. However, cognitive impairment (CI), one of the non-motor symptoms, significantly affects patients’ quality of life. According to extensive epidemiological studies, approximately 20% of Parkinson’s Disease patients show symptoms of Mild Cognitive Impairment (MCI) at the early stages of the disease. As the disease progresses, about 80% of patients eventually develop Parkinson’s Disease Dementia (PDD). Despite these concerning statistics, the underlying mechanisms of cognitive decline and dementia in Parkinson’s Disease remain unclear. Identifying biomarkers indicative of the brain’s pathological changes associated with these conditions is pivotal for elucidating their pathophysiological processes and improving diagnostic and prognostic accuracy. Addressing these needs, neuroimaging techniques have been increasingly applied in recent years to detect early cortical changes in Parkinson’s patients.

This report is based on the scientific paper “Imaging biomarkers of cortical neurodegeneration underlying cognitive impairment in Parkinson’s disease,” published in the journal European Journal of Nuclear Medicine and Molecular Imaging. The study comprehensively investigates the sensitivity and diagnostic utility of neuroimaging biomarkers for Parkinson’s Disease-related CI. The paper, authored by Jesús Silva-Rodríguez, Miguel Ángel Labrador-Espinosa, and several collaborators, is primarily affiliated with the Hospital Universitario Virgen del Rocío and the Instituto de Biomedicina de Sevilla in Spain. The research focuses on directly comparing the performance of three widely utilized neuroimaging techniques in detecting cortical changes in Parkinson’s patients: Structural MRI (sMRI), Diffusion-Weighted MRI (dMRI), and [18F]Fluorodeoxyglucose Positron Emission Tomography ([18F]FDG PET).

Study Background and Design

Study Objectives

The study aimed to compare the sensitivity and diagnostic accuracy of the three aforementioned neuroimaging modalities in detecting cortical changes at different cognitive stages of Parkinson’s Disease. Additionally, the study explored whether combining data from multiple imaging modalities could enhance diagnostic performance.

Study Sample

The study recruited 120 Parkinson’s Disease patients, comprising 53 cognitively normal patients (PD-CN), 32 patients with Mild Cognitive Impairment (PD-MCI), and 35 patients with Parkinson’s Disease Dementia (PDD). All patients were diagnosed following the Movement Disorders Society (MDS) clinical diagnostic criteria. Cognitive function was evaluated using the Parkinson’s Disease Cognitive Rating Scale (PD-CRS), and motor symptoms were assessed using Part III of the Unified Parkinson’s Disease Rating Scale (UPDRS-III).

Imaging Acquisition and Processing

  • sMRI: Data were acquired using high-resolution 3D T1-weighted sequences. The sMRI images were segmented into gray matter, white matter, and cerebrospinal fluid using standard algorithms (based on the Computational Anatomy Toolbox, CAT12, and Statistical Parametric Mapping, SPM12). The images were standardized to Montreal Neurological Institute (MNI) space.
  • dMRI: Imaging data were acquired using diffusion-weighted pulsed sequences. Preprocessing steps, including head motion correction and other adjustments, resulted in free-water corrected diffusion metrics such as Mean Diffusivity (MD).
  • [18F]FDG PET: Brain glucose metabolism data were collected using two PET scanners (19 participants scanned on Siemens Biograph HiRez, and 101 participants on GE Discovery MI). PET data were also normalized to MNI space, with differential smoothing applied to correct for scanner differences in resolution.

Data Analysis and Machine Learning

Quantitative imaging parameters were extracted from 52 cortical regions of interest (ROIs) derived from the Harvard-Oxford Brain Atlas. Analysis of covariance (ANCOVA) was employed to evaluate group differences in gray matter volume, standardized uptake value ratios (SUVR), and MD values. Following this, machine learning techniques with cross-validation were used to construct classification models and assess the diagnostic performance of these metrics in distinguishing between different cognitive stages.

Key Findings

Group-Level Analysis for Different Imaging Modalities

  • sMRI:
    • PDD patients exhibited significant gray matter atrophy, primarily in posterior-parietal cortices (e.g., precuneus, posterior cingulate) and temporal regions.
    • In contrast, gray matter changes between PD-MCI and PD-CN groups were minimal and did not reach statistical significance.
  • dMRI:
    • PDD patients showed widespread increases in MD, particularly in the temporal pole, posterior cingulate, and angular gyrus.
    • PD-MCI patients also demonstrated significant abnormalities in the medial temporal lobe’s MD values.
  • [18F]FDG PET:
    • Both PDD and PD-MCI patients exhibited a characteristic pattern of hypometabolism in posterior-occipital regions.
    • This was particularly pronounced in the precuneus and angular gyrus, where the effect sizes for [18F]FDG PET far exceeded those for other modalities.

Comparative Classification Performance

The constructed classification models yielded the following results: - sMRI: - Moderately effective in distinguishing PDD from PD-CN (AUC = 0.77), but showed no diagnostic value in discriminating between PD-MCI and PD-CN (AUC = 0.57). - dMRI: - Performed well in distinguishing PDD from PD-CN (AUC = 0.87), with modest capability for PD-MCI (AUC = 0.71). - [18F]FDG PET: - Showed the best performance among the modalities, with an AUC of 0.89 for PDD classification and 0.78 for PD-MCI classification.

Multimodal Models

Integrating data from multiple modalities (sMRI, dMRI, and [18F]FDG PET) did not significantly outperform models based solely on [18F]FDG PET (AUC = 0.86 vs. 0.89). This suggests that multimodal combinations provide limited additional diagnostic value with the current imaging data.

Research Significance and Implications

This study is the first to systematically compare the performance of [18F]FDG PET with both dMRI and sMRI in detecting cognitive impairment in Parkinson’s Disease. It highlighted [18F]FDG PET as the most sensitive modality for early detection of CI. Additionally, the research underscored the potential of dMRI, particularly when [18F]FDG PET is unavailable, as MD measurements could serve as a clinically effective alternative. Furthermore, the findings suggest that sMRI likely reflects later-stage neurodegenerative changes rather than early microstructural abnormalities.

Notably, the study adopted a machine-learning-based analytical framework to robustly assess diagnostic performance, providing a valuable methodological reference for future large-scale, multicenter neuroimaging research.

This research not only advances the understanding of the pathophysiology of cognitive impairment in Parkinson’s Disease but also provides an evidence-based foundation for the selection of clinical diagnostic tools, thereby promoting the application of precision medicine in this field.