学术交流|最新动态| 华西磁共振研究中心

Machine learning prediction in MDD.

发布时间： 2024-05-27 11:20:48 作者：来源：

Topic: Machine learning prediction in MDD.

Moderator: Ling Lin, Post.doc

Speaker 1 : Fenghua Long, M.Med. Candidate

Supervisor: Prof. Fei Li

Speaker 2 : Mengyao Chen, M.Med. Candidate

Supervisor: Prof. Huaiqiang Sun

Date: 27/05/2024, 14:00

Location: The lab of HMRRC (10011, the 8th Teaching Building)

Speaker 1: Fenghua Long, M.Med. Candidate

Title: Mapping Neurophysiological Subtypes of Major Depressive Disorder Using Normative Models of the Functional Connectome.

Keypoints：

Question: What are the neurobiological subtypes of major depressive disorder (MDD) and how do they differ in terms of functional connectome abnormalities, depressive symptoms, and treatment outcomes?

Findings: An unsupervised clustering algorithm identified two reproducible neurophysiological subtypes of MDD from multiple sites. Subtype 1 showed severe functional connectome deviations, with increased connectivity in default mode, limbic, and subcortical areas, and decreased connectivity in sensorimotor and attention areas. Subtype 2 exhibited a moderate but opposite deviation pattern. These subtype differences were associated with depressive item scores and predictive ability for antidepressant treatment outcomes.

Meaning: The two subtypes reveal neurobiological heterogeneity in MDD, pointing to potential for precision diagnosis and personalized treatment.

Speaker 2 : Mengyao Chen, M.Med. Candidate

Title: Functional connectivity signatures of major depressive disorder: machine learning analysis of two multicenter neuroimaging studies.

Keypoints：

Question: Is resting-state functional connectivity (FC) a reliable biomarker for Major Depressive Disorder (MDD)?

Findings: In this study, which utilized two large multicenter neuroimaging datasets, machine learning techniques, including support vector machines (SVM) and graph convolutional networks (GCN), achieved a mean classification accuracy of 61% for differentiating MDD patients from healthy controls. The study identified thalamic hyperconnectivity as a significant feature in MDD. Classification accuracy for non-medicated vs. medicated subgroups was 62%, and for sex classification, it ranged from 73-81%.

Meaning: Resting-state FC shows potential as a biomarker for MDD, especially due to thalamic hyperconnectivity. However, moderate accuracy suggests FC alone may not be reliable for diagnosis, highlighting the need for refined models and larger datasets.