Meta-Analysis Informed Functional Connectomes Representations for Depression Identification.

Area of Science:

• Neuroimaging
• Psychiatry
• Machine Learning

Background:

• Meta-analyses in neuroimaging are increasingly popular but lack clear clinical utility.
• Voxel-wise analyses face the curse of dimensionality, limiting diagnostic accuracy.
• Convergent masks in meta-analyses are often small and focal.

Purpose of the Study:

• To develop a functional connectome representation (FCR) by integrating meta-analytic neuroimaging data.
• To evaluate the FCR's performance in identifying depression.

Main Methods:

• A retrospective study utilizing resting-state functional MRI data.
• Developed FCR using community detection and principal component analysis.
• Evaluated model performance using accuracy, specificity, and sensitivity on principal and external datasets.

Main Results:

• The FCR model achieved high accuracy: 89.42% on the principal dataset and 83.35% on the external dataset.
• Effect sizes (Cohen's d) for FCR components ranged from -0.22 to 0.84, indicating significant differences between patients and controls.
• Permutation tests confirmed the model's accuracy was significantly above chance, with robustness demonstrated by a negative correlation between accuracy and noise.

Conclusions:

• The functional connectome representation (FCR) effectively discriminates between individuals with depression and healthy controls.
• The FCR demonstrates strong diagnostic performance, generalization, and robustness.
• This approach shows potential utility in the clinical identification of depression.