A total of 166 participants were included in this cross-sectional study. The participants were recruited as part of a prospective, community-based cohort study for MCI (PREcision medicine platform for mild cognitive impairment on multi-omics, imaging, evidence-based R&BD; PREMIER) and had visited the Center for Dementia in Gangnam, Bucheon, Incheon, Dongtan, Gijang, Dongtan, or Haewoondae, South Korea. Individuals who were cognitively unimpaired (CU) or had aMCI were included in the PREMIER cohort. The Centers of Dementia are public facilities that focus on integrated dementia management by providing services, including early screening, dementia prevention education, cognitive rehabilitation programs, and treatment support. All participants were assessed using the CERAD-K, which is a comprehensive NP test. Experienced neurologists reviewed all clinical information about the participants and divided them into CU and aMCI groups.

The criteria for CU were as follows: 1) normal cognitive function defined by all the scores of the four episodic memory tests (word list memory, word list recall, word list recognition, and constructional praxis recall tests) in the CERAD-K above −1.5 standard deviation (SD) of the age, sex, and education-based norm78; and 2) normal activities of daily living with or without subjective memory complaints.

The criteria for aMCI, which were based on Pertersen’s criteria,1 were as follows: 1) memory complaint corroborated by an informant; 2) objective memory impairment relative to age, education, and sex; 3) essentially preserved general cognitive function; 4) largely intact functional activities; and 5) no dementia. Objective memory impairment was defined by a performance score of 1.5 SD below the respective age-, education-, and sex-specific normative means for at least one of the four memory scores of episodic memory tests included in the CERAD-K NP battery, namely word list memory, word list recall, word list recognition, and constructional praxis recall tests.78 Of all the 166 participants, 162 underwent apolipoprotein genotyping, and a review of past medical history was missing in one participant due to follow-up loss.

The Inbrain CST was administered on a 12-inch tablet PC running Microsoft Windows 10. It consists of seven neuropsychological tests assessing five cognitive functions: the Visual Span Test included forward and backward tasks used to assess attention; the Difficult Naming Test and semantic (fruits)/phonemic (Korean alphabet digeut) word fluency test assessed language; the Block Design Test assessed visuospatial function; the time orientation questions and Word Place Association Test assessed memory; and the Korean-Trail Making Test-Elderly version evaluated executive function. Age-, education-, and sex-specific normative values for each cognitive domain were provided by the manufacturer based on a previous study of 480 cognitively normal, community-dwelling, older adults.5

Seventy-six participants (61 aMCI and 15 CU) underwent brain MRI scans. T1-weighted MRI scans were obtained on a 3.0 T magnetic resonance scanner (Siemens Skyra, Siemens Medical Solutions, Erlangen, Germany), and high-resolution anatomic magnetization-prepared rapid gradient-echo images were obtained with the following parameters: sagittal slice thickness, 1.0 mm with 16.7% overlap; no gap; repetition time, 2,300 ms; echo time, 2.01 ms; flip angle, 9°; 1-mm isotropic voxels; and 192 slices. Image processing was performed using FreeSurfer 5.1.0 (Athinoula A. Martinos Center at the Massachusetts General Hospital, Harvard Medical School; http://surfer.nmr.mgh.harvard.edu/). First, we constructed the cortical surface and the outer and inner surface meshes. The outer and inner meshes were isomorphic and identical in terms of connectivity and height. We adopted the manifold harmonic transform (MHT) to map the cortical thickness from the surface onto the frequency domain to remove noise.910 The high-frequency components of the transformed cortical thickness were regarded as noise by the MHT and were eliminated. This filtering of high-frequency components can remove noise and reduce the dimensionality of cortical thickness. For normalization, hippocampal and parahippocampal volumes were calculated as the sum of the right and left hippocampal or parahippocampal volumes divided by the estimated total intracranial volume. AD signature cortical thickness was measured as the average volume of the cortical thickness in the regions of interest: entorhinal, inferior temporal, middle temporal, and fusiform (summation of cortical thickness ROIs/8).11

Of the participants who were enrolled through the seven centers (see Participants section), all 76 patients who visited the Bucheon center underwent amyloid PET-computed tomography (CT) scans, which involved [¹⁸F]-flutemetamol imaging using a Biograph 128mCT unit (Siemens Medical Solutions, Knoxville, TN, USA). Ninety minutes after a bolus injection of 370 MBq [¹⁸F] flutemetamol, the patients underwent a low-dose CT scan followed by a 20-minute (i.e., four frames of 5 minutes) PET scan. The scans were carefully checked for movement, and frames were summed to acquire a static (20-minute) image for each patient. The scans were visually examined and dichotomously assessed as either amyloid-positive or amyloid-negative by a nuclear medicine specialist who had undergone a training program for visual interpretation of [¹⁸F] flutemetamol images.

We evaluated the diagnostic performance of the Inbrain CST for identification of aMCI by comparing the findings with those of CERAD-K. Considering the diagnosis of aMCI based on CERAD-K as the gold standard, we assessed the overall accuracy, sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios of the Inbrain CST. Statistical analyses were performed using the Rex package (Rex: Excel-based statistical analysis software, version 3.0.3; RexSoft Inc., Seoul, Korea) and R (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria). Sensitivity/specificity, positive/negative likelihood ratios, positive/negative predictive values, overall accuracy, and 95% confidence intervals (CIs) were estimated. Sensitivity and specificity were compared using the McNemar test, whereas positive/negative likelihood ratios were compared with a regression model approach, and positive/negative predictive values were compared with a generalized score statistic, all of which were implemented in the Rex package. We determined the characteristics of aMCI patients as defined by CERAD-K and the Inbrain CST. The clinical and neuroimaging variables of the two groups were statistically compared using Rex software. Group comparisons were performed using t-tests for continuous variables and χ² tests for categorical variables.

This study protocol was reviewed and approved by the Institutional Review Board (IRB) of Soonchunhyang University Bucheon Hospital (IRB approval No. 2017-04-058). All participants signed an informed consent form approved by the IRB prior to participation.

Of the 166 participants, 93 were diagnosed with aMCI while 73 were CU. The average age of the aMCI group was 77.02 years, and the average number of years of education was 6.46 years. Thirty-six of the 162 participants (22.2%) were apolipoprotein E ε4 carriers, including 22/90 (24.4%) participants with aMCI, and 14/72 (19.4%) CU participants. Amyloid PET scans revealed positive results in 52/76 (68.4%) of the total participants, including 46/61 (75.4%) of those with aMCI and 6/15 (40.0%) CU participants (Table 1).

When objective memory impairment was defined as an Inbrain CST memory domain score of less than −1.0 SD of the reference scores for the same age and years of education, the sensitivity of the Inbrain CST for aMCI diagnosis was 81.7%, and its specificity was 84.9%. Positive and negative predictive values were 87.4% and 78.5%, respectively (Table 2), and the positive and negative likelihood ratios were 5.4233 and 0.2152, respectively. The diagnostic accuracy was 83.1%, and the error rate was 16.9%.

The clinical and neuroimaging characteristics of 93 patients with aMCI as defined by the CERAD-K and 87 patients with aMCI as defined by the Inbrain CST were compared (Table 1). Age, sex, education, APOE e4 carrier status, amyloid positivity, MMSE score, clinical dementia rating (CDR), CDR-sum of boxes (CDR-SB), depression and anxiety scale score, and underlying medical conditions were not significantly different between the two groups. Among participants who underwent brain MRI, the hippocampal and parahippocampal volumes and AD signature cortical thickness, which can be considered neuroimaging biomarkers of AD, were not different between the two groups (Table 3).

We obtained amyloid PET in 76 participants (61 aMCI and 15 CU), and 52 participants showed positive scans. Among the 61 patients with aMCI defined by CERAD who underwent amyloid PET, 52 (85.3%) showed amyloid positivity. In aMCI defined by the Inbrain CST group, 61 patients had amyloid PET and 46/61 (75.4%) revealed positive results. Only four of the participants with amyloid PET positivity had discrepant diagnosis; two patients were diagnosed with aMCI only by the Inbrain CST, while the other two were diagnosed with aMCI only by CERAD-K. The results of two diagnostic tests were congruent for the remaining 48 patients (46 aMCI, 2 CU).