2.1. Patients and controls
A case-control study was conducted from June 2019 to November 2020 in the Guangji Hospital affiliated with Soochow University (Suzhou City, China). Recruited patients met the following inclusion criteria: (1) inpatient adults (20 to 65 years-old) and (2) depressive diagnosis meeting DSM-5 criteria. Subjects were excluded if they were: (1) unable to undergo MRI scanning, (2) left-handed, (3) pregnant females, (4) febrile within 2 weeks, (5) subjects with immune dysfunction, or (5) subjects with organic psychosis.
Healthy volunteers were recruited as controls through matching demographic data on enrolled patients. These volunteers were (1) adults (20 to 65 years old) who did not meet any DSM-5 diagnostic criteria and (2) of normal intelligence. Subjects were excluded if they were: (1) unable to undergo MRI scanning, (2) left-handed, (3) pregnant females, (4) febrile within 2 weeks, (5) subjects with immune dysfunction, or (5) subjects with substance abuse or dependence history within one year, (6) subjects with a HAMD-17 score greater than or equal to 8 points, or (7) subjects with mental or genetic neurological disorders in first-degree relatives.
This study was approved by the ethics committee of the Guangji Hospital affiliated to Soochow University (ethics number 2017YFE0103700). All participants provided the written informed consent. All study procedures were in compliance with the Declaration of Helsinki.
2.2. Blood sampling and ELISA test
Fasting peripheral blood was collected from the enrolled patients and controls. After centrifugation at 3000 rpm for 10 minutes, the supernatant was stored at -80 °C and the ELISA test was performed according to the kit manual to obtain the IL-18 concentration.
2.3. Clinical assessment
The Chinese HAMD-17 Depression Rating Scale was used to assess depression due to good reliability and validity [12]. Subjects with depressive disorders were defined as having a HAMD-17 score no less than 8 points. Clinical assessments were completed by the same psychiatrist.
2.4. Image data collection
Resting-state imaging data was collected using Siemens SKYPE 3.0T MRI equipment. Prior to resting-state imaging, patients were scanned for anatomical positioning using parameters of 192 slices, 1 mm slice thickness, TR of 2530.0 ms, TE of 2.98 ms, FOV of 256 mm x 256 mm, and an element size of 1.0 mm x 1.0 mm x 1.0 mm. EPI sequence was used to acquire rs-MRI BOLD signal with parameters of 32 layers, 3.5 mm layer thickness, TR 2000.0 ms, TE 30.0 ms, FOV 224 mm x 224 mm, 64 x 64 matrix, 90° flip angle, and voxel size of 3.5 mm x 3.5 mm x 3.5 mm.
2.5. Image data pre-processing
Data was pre-processed by the cortex-based resting-state fMRI data analysis package DPABISurf developed by DPABI/DPABISurf. DPABISurf calls fMRIprep to pre-process the structural and functional magnetic resonance data and provides a series of statistical and visualization tools through fMRIprep.
The process involved: (1) removing the first 10 time points to ensure the magnetic field signal uniformity, (2) converting the data to BIDS format and call fmriprep1.5.0 docker, (3) pre-processing of structural image data, (4) pre-processing of functional image data, (5) nuisance regression, and (6) filtering and smoothing.
Pre-processed image results were processed through online quality control to eliminate subjects with large T1 structural image translations and fuzzy functional images with poor quality or incomplete functional image coverage. Subject data with head movements greater than 3 mm and 3 degrees was excluded and the remaining data was retained for statistical analysis.
2.6. Statistical analysis
The differences between groups of normally distributed continuous, non-normally distributed continuous, and categorical variables used two independent sample t-tests, the Mann-Whitney-rank sum test, and chi-square analysis, respectively. Using SPSS 24.0, data was statistically analyzed with a two-tailed statistical significance set at p<0.05.
The relationship between resting-state fMRI data and serum IL-18 levels was analyzed with DPABISurf statistical analysis using Matlab 2017b software. Pearson correlation was performed and corrected through threshold-free cluster enhancement (TFCE) to determine whether brain region differences between groups were present at the whole brain level. The false positive rate (family wise error rate) level was controlled at p<0.025 and the Bonferroni correction was performed on the two brain hemispheres.