[Paper Review] The Reconfiguration Pattern of Individual Brain Metabolic Connectome for Parkinson's Disease Identification

Weikai Li, Yongxiang Tang+3 more|arXiv (Cornell University)|Apr 29, 2021

Parkinson's Disease Mechanisms and Treatments49 references11 citations

TL;DR

This study introduces a novel individual-level brain metabolic connectome method using 18F-FDG PET data and Jensen-Shannon Divergence Similarity Estimation (JSSE) to identify Parkinson's disease (PD) at the individual level. It reveals specific reconfiguration patterns in metabolic networks—particularly in PUT-PCG, THA-PCG, SMA, ACG-PCL, DCG-PHG, and ACG pathways—and achieves 91.84% accuracy in distinguishing PD patients from healthy controls using MK-SVM classification.

ABSTRACT

Background: Positron Emission Tomography (PET) with 18F-fluorodeoxyglucose (18F-FDG) reveals metabolic abnormalities in Parkinson's disease (PD) at a systemic level. Previous metabolic connectome studies derived from groups of patients have failed to identify the individual neurophysiological details. We aim to establish an individual metabolic connectome method to characterize the aberrant connectivity patterns and topological alterations of the individual-level brain metabolic connectome and their diagnostic value in PD. Methods: The 18F-FDG PET data of 49 PD patients and 49 healthy controls (HCs) were recruited. Each individual's metabolic brain network was ascertained using the proposed Jensen-Shannon Divergence Similarity Estimation (JSSE) method. The intergroup difference of the individual's metabolic brain network and its global and local graph metrics were analyzed to investigate the metabolic connectome's alterations. The identification of the PD from HC individuals was used by the multiple kernel support vector machine (MK-SVM) to combine the information from connection and topological metrics. The validation was conducted using the nest leave-one-out cross-validation strategy to confirm the performance of the methods. Results: The proposed JSSE metabolic connectome method showed the most involved metabolic motor networks were PUT-PCG, THA-PCG, and SMA pathways in PD, which was similar to the typical group-level method, and yielded another detailed individual pathological connectivity in ACG-PCL, DCG-PHG and ACG pathways. These aberrant functional network measures exhibited an ideal classification performance in the identifying of PD individuals from HC individuals at an accuracy of up to 91.84%.

Motivation & Objective

• To overcome limitations of group-level metabolic connectome studies by enabling individual-level analysis of brain metabolic networks in Parkinson’s disease (PD).
• To identify aberrant connectivity patterns and topological alterations in individual PD patients using 18F-FDG PET data.
• To develop a robust diagnostic framework that integrates connection-level and topological metrics for individual PD identification.
• To validate the method’s performance using rigorous cross-validation and assess its clinical potential for early PD detection.

Proposed method

• The Jensen-Shannon Divergence Similarity Estimation (JSSE) method is used to construct individual brain metabolic connectomes from 18F-FDG PET data.
• Each individual’s metabolic network is modeled as a graph, with brain regions as nodes and metabolic similarity between regions as weighted edges.
• Graph metrics—both global (e.g., efficiency, modularity) and local (e.g., degree, betweenness)—are computed to quantify topological alterations.
• A multiple kernel support vector machine (MK-SVM) integrates connection-level and topological features for improved classification performance.
• A nested leave-one-out cross-validation strategy is employed to validate the diagnostic accuracy and generalizability of the model.
• The method emphasizes individual-specific reconfiguration patterns rather than group-averaged effects, enhancing sensitivity to subtle pathological changes.

Experimental results

Research questions

• RQ1 Which individual-level metabolic network reconfiguration patterns are most prominent in Parkinson’s disease patients compared to healthy controls?
• RQ2 How do topological metrics of individual brain metabolic networks differ between PD patients and healthy controls?
• RQ3 Can a combination of connection-level and topological features from individual metabolic connectomes reliably distinguish PD patients from healthy individuals?
• RQ4 What is the diagnostic performance of the proposed method in classifying individual PD cases using cross-validation?
• RQ5 Which specific metabolic pathways show the most significant reconfiguration in PD, and how do they relate to known motor and cognitive circuits?

Key findings

• The JSSE-based individual metabolic connectome method successfully identified key pathological reconfiguration patterns in PD, including in PUT-PCG, THA-PCG, and SMA pathways.
• Additional individual-specific aberrant connections were detected in ACG-PCL, DCG-PHG, and ACG pathways, suggesting previously underappreciated network disruptions.
• The MK-SVM classifier achieved a classification accuracy of 91.84% in distinguishing PD patients from healthy controls using individual connectome features.
• Global and local graph metrics showed significant intergroup differences, with PD patients exhibiting altered network efficiency and integration patterns.
• The nested leave-one-out cross-validation confirmed the robustness and generalizability of the diagnostic model across individual subjects.
• The results demonstrate that individual-level metabolic connectome analysis captures more nuanced pathological signatures than group-level approaches.