2. Materials and Methods
2.1. Datasets and study population
★ 데이터셋 설명 방식 참고 ★
This study uses two datasets including the OASIS-3 (LaMontagne et al., 2019) cohort as the reference population and UK Biobank (Littlejohns et al., 2020) as the target one. OASIS-3 contains 1389 imaging samples (age: 67.18±8.71) and their associated demographic and clinical data.
We used Clinical Dementia Rating Scale Sum of Boxes Scores or CDR_SOB to identify samples with mild cognitive impairment or MCI with CDR_SOB>0 from control or CN with CDR_SOB=0. Overall, we have 1028 sample of CN and 361 sample of MCI in this dataset.
We additionally used the ADNI dataset (age:71.85± 6.99) which contain 382 CN and 347 MCI individuals to validate our proposed MCI risk score (Mueller et al., 2008; Weber et al., 2021).
2.2. Imaging protocol
★ 데이터셋 설명 방식 참고 ★
OASIS-3 data were collected from 3 different Siemens scanners (Siemens Medical Solutions USA, Inc), including one Vision 1.5T and two scanners of TIM Trio 3T with a 16-channel head coil on 1.5T scanners and 20-channel head coil on 3T scanners with foam pad stabilizers placed next to the ears to decrease motion. High-resolution T2*-weighted images were acquired using a gradient-echo EP sequence with TE =27 ms, TR = 2.5 s, flip angle = 90°, slice thickness = 4mm, slice gap = 4 mm, and matrix size = 64 for the Trio scanner. For the Vision scanner, the scanning parameters with gradient-echo EP sequence are TE= 27 ms, TR= 2.2 s, flip angle= 90°, slice thickness=4 mm, slice gap = 4 mm, and matrix size = 64 (LaMontagne et al., 2019).
The ADNI dataset comprises imaging data acquired using various MRI scanners, including 1.5T GE, 3T GE, 1.5T Philips, 3 T Philips, 1.5T Siemens, and 3T Siemens scanners. The specific details of the imaging protocols utilized in the acquisition of the ADNI dataset can be found in (Song et al., 2022).
2.3. Preprocessing and feature extraction
★ rs-fMRI 초반 시점 수집된 Volume 제거 참고 ★
사전전처리 flow도 참고 가능
(rs-fMRI Volume 제거에 OASIS-3 데이터도 포함된 내용인지는 명확히 기술되어 있지 않음)
For the resting-state fMRI analysis, the initial step involved removing the first five dummy scans prior to preprocessing. For the subsequent steps, we utilized the default slice timing routines of Statistical Parametric Mapping (SPM12, https://www.fil.ion.ucl.ac.uk/spm/). The slice that was obtained in the middle of the sequence was used as the reference slice in this procedure. For participant head movement correction, we used rigid body motion correction. After that, we used the echo-planar imaging (EPI) template to normalize the imaging data to the standard Montreal Neurological Institute (MNI) space.