Multimodal Assessment of Cognitive Disorders

In this work, the study used data collected from the three centers of the Show Chwan Healthcare System. The data selected from the register-based database of the Show Chwan Health System were analyzed anonymously with the informed consent from all participants, and the study was designed retrospectively in accordance with relevant guidelines and regulations. The project was reviewed by the Medical Research Ethics Committee of Show Chwan Memorial Hospital, and the study was approved by the Data Inspectorate.
The data for the study consisted of samples of clinical and neuropsychological assessment obtained from 6701 patients. For detailed neuropsychological tests, we assessed the history of cognitive status and objective assessments including the Clinical Dementia Ratings (CDR), Mini Mental Status Examination (MMSE), Cognitive Abilities Screening Instrument (CASI) and Montreal Cognitive Assessment (MoCA) performed to evaluate memory, executive function, orientation, visual-spatial ability, and language function [11] . Along with the current scales such as CDR, MMSE, CASI, MoCA, we used a newly designed Informant-based questionnaire named HAICDDS which is applied in dementia registration in a health system with 9 regional hospitals in Taiwan. Clinical application of the HAICDDS had been published in journals [11] –[13] (link, link) or conferences [14] , [15] . The CDR determined the severity of dementia. Experienced neurologists evaluated the participants based on their clinical symptoms and reviews of medical/medication history, neuropsychological test results, and then classified the participants into six diagnostic groups: normal (535 participants), MCI (1687 participants), VMD (678 participants), Mild (1812 participants), Moderate (1309 participants), and Severe (680 participants). The six diagnostic groups were defined using the CDR staging. Among CDR 0.5, participants without significantly impaired activities of daily living were divided as CDR 0.5 MCI and those with significantly impaired activities of daily living were divided as CDR 0.5 VMD. Therefore, the 6 groups were CDR 0, CDR 0.5 MCI, CDR 0.5 VMD, CDR 1, CDR 2, and CDR 3. The operational diagnosis of a significant interfere with ADL is the IADL total score <7.
we randomly split the data with the ratio of 9:1, of which 90% are training data sets (6030 participants) and 10% are test data sets (671 participants) [16] (link). In order to estimate the generalization error, this procedure was repeated 10 times independently to avoid any deviation caused by randomly partitioning data sets. The average accuracy and F1-score were calculated for performance analysis. We finally obtained 10 training-test of different training set (6030 participants) and test set (671 participants). We also repeated the independent training-test procedure more than 10 times (k = 10) but the results were similar, so only the results with k = 10 were reported in the manuscript.