Subjects
This prospective study was approved by the Institutional Review Board of Severance Hospital, and informed written consent was obtained from all subjects before each procedure. Male gastric cancer patients between age 40 and 60 who underwent total gastrectomy or partial gastrectomy (distal gastrectomy, subtotal gastrectomy) were enrolled in this study. Candidates were divided into two groups: patients with scheduled adjuvant chemotherapy (CTx+ group) and patients who do not need adjuvant chemotherapy (CTx- group). In addition, Age- and sex-matched healthy controls (HC group) without cognitive impairment or active neurological disorders were also recruited as a control group. Participants who had (1) history of other malignancy, (2) history of metastatic malignancy, (3) history of any neurologic condition that could impair cognitive function (neurodegenerative disease, stroke, brain injury, etc.), (4) history of any neurologic condition that could impair cognitive function (dementia, stroke, brain injury, etc.), (5) history of alcohol, nicotine, caffeine, or other drug dependence or addiction and (6) psychiatric Axis I disorder were excluded in this study.
Initial baseline assessment was performed on CTx+ group (n = 19, age 49.2±5.5) and CTx- group (n = 14 age 49.2±6.8) and HC group (n=10, age 51.5±7.0). Baseline assessment was performed by carry out self-report questionnaires assessment, neurocognitive assessment, and magnetic resonance imaging (MRI) scan all on the same day, after gastric cancer surgery and before adjuvant chemotherapy. In CTx+ group, follow-up assessment was performed around 3 months past baseline assessment, after the subject underwent adjuvant chemotherapy. CTx- group was also assessed on matched intervals.
Self-report questionnaires and neurocognitive assessment
Self-report questionnaires contained the Cognitive Failure Questionnaire (CFQ) (36) to assess subjective cognitive decline; the Beck Depression Inventory (BDI) (37) to assess depressive symptoms; and the Beck Anxiety Inventory (BAI) (38) to assess anxiety symptoms. All subjects carried out Structured Clinical Interview, and assessment on major psychiatric illness was performed according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (39). Four cognitive domains were assessed using a set of neurocognitive tests: (1) performance and verbal intelligence [Korean version of Wechsler Adult Intelligence Scale (K-WAIS) performance and verbal subtests]; (2) memory (Rey-Kim Memory Test); (3) attention (K-WAIS digit span and spatial span subtest); and (4) executive function (stroop Test). Scores for the neurocognitive tests were expressed as age corrected scaled scores (AgeSS), standardized scores (SS), or percentile ranks for raw scores.
Image acquisition
All subjects were imaged on a 3.0 T MRI scanner (Tim Trio, Siemens Healthcare, Erlangen, Germany) and images including T1-weighted imaging, DTI and resting state fMRI were obtained. A 3D T1-weighted anatomical image was obtained using a spoiled gradient-echo sequence (sagittal acquisition with TR = 1900 msec; TE = 2.52 msec; FOV = 256 × 256 mm2; voxel size = 1 × 1 × 1 mm3; flip angle 9°; slice number = 176; and total acquisition time 4 min 26 sec) to serve as an anatomical underlay for the brain activity and to be used for GM volume analysis. DTI were acquired with the following echo planar acquisition parameters: diffusion‐weighted gradients applied in 30 non‐linear directions, number of average = 2, TR = 6200 ms, TE = 85 ms, flip angle = 90°, acquisition matrix = 128 × 128, FOV = 230 × 230 mm2, slice thickness = 3 mm, and b value = 600 s/mm2. Resting state functional images were acquired using gradient echo-planar pulse imaging (EPI). For each subject, 150 axial volume scans were obtained with the following parameters: TR = 3000 msec, TE = 30 msec, FOV = 192 × 192 mm2, voxel size = 3×3×3 mm3, slice number = 50 (interleaved). All participants were instructed to rest and keep their eye closed without sleeping, moving, or thinking about anything during the scan (7min 30sec). Vacuum-molded cushions and soft pads were used to support the head and minimize head movement.
Voxel-wise analysis of GM
All preprocessing steps were conducted in accordance with the standardized procedure (40). First, the structural images were aligned along the anterior–posterior commissure line and positioned so that the anterior commissure matched the origin. Afterward, the images were segmented into gray matter, white matter and cerebrospinal fluid probability maps by using a Bayesian image segmentation algorithm. Brain tissue probability maps for each subject were then used for intersubject alignment. In this study, we applied diffeomorphic anatomical registration by using an exponentiated Lie algebra algorithm (DARTEL) (40). The DARTEL has been suggested to enhance the accuracy of intersubject alignment, by modeling the shape of each brain by using a host of parameters. The DARTEL processing involves generating the flow fields that parameterize the deformations and creating the templates for all subjects. After the final study‐specific template was created, gray matter images for each subject were warped to the study‐specific template and then normalized into standard Montreal Neurological Institute space. The volumes were resampled to 1.5 × 1.5 × 1.5 mm3 voxel size. This spatial normalization step included Jacobian modulation in order to preserve regional volume data. Finally, the DARTEL‐warped, normalized and modulated gray matter images were smoothed by using 8‐mm full‐width at half maximum kernel.
Functional connectivity analysis
Spatial preprocessing and statistical analyses of functional images were performed using SPM12 (Wellcome Trust Centre for Neuroimaging). To analyze functional connectivity of resting state functional MRI, motion artifacts were assessed in individual subjects by visually inspecting realignment parameter estimations to confirm there were no abrupt head motions and the maximum head motion in each axis was <3 mm. Functional images were realigned and registered to structural images for each subject. The anatomical volume was segmented into gray matter, white matter, and cerebrospinal fluid. The gray matter image was used for determining the parameters of normalization onto the standard Montreal Neurological Institute (MNI) gray matter template provided with SPM12. The spatial parameters were then applied to the realigned functional volumes that were finally resampled to voxels of 2×2×2 mm3 and smoothed with an 8 mm full-width at half-maximum kernel.
The assessment of cortical networks was performed using a ROI seed-based correlation approach. Connectivity analysis was conducted with the “conn” toolbox, implemented in the SPM12 (http://www.fil.ion.ucl.ac.uk/spm/ext). Initially, the bilateral hippocampus seed regions were defined as MNI space taken from the AAL atlas(41). ROI for subfields of bilateral hippocampus (consisting of cornu ammonis(CA), including CA1,CA2, and CA3 subfields; dentate gyrus(DG), including fascia dentata and the CA4 subfield; and subiculum(SB), including the prosubiculum, subiculum proper,presubiculum and parasubiculum) was obtained from the the maximum probability map(MPM) (42) and was defined using the Anatomy toolbox v22c implemented on SPM12 (www.fil.ion.ucl.ac.uk/spm). The waveform of each brain voxel was temporally filtered by means of a bandpass filter (0.008 Hz < f < 0.09 Hz) to adjust for low-frequency drift and high-frequency noise effects. A linear regression analysis was conducted to remove signals from the ventricular area and the white matter(43). Movement parameters were added as first-level covariate. The between-group, within-group longitudinal comparisons, and group-by-time interactions analyses were compared with an uncorrected p-value height threshold of 0.001 and k=90 as extent threshold for the whole brain. To estimate the strength of an FC, correlation coefficients were computed and converted to z-values using Fisher's r-to-z transformation.
DTI imaging processing and analysis
DTI data were analyzed using diffusion MR toolbox ‘Explore DTI’ (44) and following steps were performed: (i) correction for subject motion and eddy current induced distortions (45); (ii) tensor estimation using the REKINDLE approach for outlier detection (46) with iteratively reweighted linear least squares estimation after identification and removal of data outliers (47); and (iii) automated atlas based analysis with the SRI24 Atlas (normal adult brain anatomy; (48) using affine and elastic registration based on ‘elastix’ (49). All DTI data were visually checked in terms of quality of tensor estimation and quality of registration. After these preprocessing steps, FA, AD, RD, and MD values were calculated in the 130 brain regions that are provided by the SRI24 Atlas (48).
Statistical analysis
Baseline demographic characteristics including age, years of education and results from self-report questionnaires were compared between gastric cancer patients with one-way ANOVA. In analyzing results from neurocognitive assessment, we compared changes in performance status of neurocognitive test before and 3 months after adjuvant chemotherapy with a repeated measures ANOVA for a significance level of P = 0.05 between CTx+ group and CTx- group. Statistical analyses were conducted by using SPSS 25.0 (IBM, Armonk, NY, USA).