Deep learning based low-cost high-accuracy diagnostic framework for dementia using comprehensive neuropsychological assessment profiles.

BACKGROUND: The conventional scores of the neuropsychological batteries are not fully optimized for diagnosing dementia despite their variety and abundance of information. To achieve low-cost high-accuracy diagnose performance for dementia using a neuropsychological battery, a novel framework is proposed using the response profiles of 2666 cognitively normal elderly individuals and 435 dementia patients who have participated in the Korean Longitudinal Study on Cognitive Aging and Dementia (KLOSCAD). METHODS: The key idea of the proposed framework is to propose a cost-effective and precise two-stage classification procedure that employed Mini Mental Status Examination (MMSE) as a screening test and the KLOSCAD Neuropsychological Assessment Battery as a diagnostic test using deep learning. In addition, an evaluation procedure of redundant variables is introduced to prevent performance degradation. A missing data imputation method is also presented to increase the robustness by recovering information loss. The proposed deep neural networks (DNNs) architecture for the classification is validated through rigorous evaluation in comparison with various classifiers. RESULTS: The k-nearest-neighbor imputation has been induced according to the proposed framework, and the proposed DNNs for two stage classification show the best accuracy compared to the other classifiers. Also, 49 redundant variables were removed, which improved diagnostic performance and suggested the potential of simplifying the assessment. Using this two-stage framework, we could get 8.06% higher diagnostic accuracy of dementia than MMSE alone and 64.13% less cost than KLOSCAD-N alone. CONCLUSION: The proposed framework could be applied to general dementia early detection programs to improve robustness, preciseness, and cost-effectiveness.

Interactive influences of demographics on the Mini-Mental State Examination (MMSE) and the demographics-adjusted norms for MMSE in elderly Koreans.

BACKGROUND: The influences of demographics, culture, language, and environmental changes on Mini-Mental State Examination (MMSE) scores are considerable. METHODS: Using a sample of 7452 healthy, community-dwelling elderly Koreans, aged 55 to 94 years, who participated in the four ongoing geriatric cohorts in Korea, we investigated demographic influences on MMSE scores and derived normative data for this population. Geropsychiatrists strictly excluded subjects with cognitive disorders according to the protocol of the Korean version of the Consortium to Establish a Registry for Alzheimer's Disease Assessment Packet (CERAD-K) Clinical Assessment Battery (CERAD-K-C). RESULTS: Education (standardized ß = 0.463), age (standardized ß = -0.303), and gender (standardized ß = -0.057) had significant effects on MMSE scores (p < 0.001). The score of MMSE increase 0.379 point per 1-year education, decrease 0.188 per 1-year older, and decrease 0.491 in women compared to men. Education explained 30.4% of the scores' total variance, which was much larger than the variances explained by age (8.4%) or gender (0.3%). Accordingly, we present normative data for the MMSE stratified by education (0, 1-3, 4-6, 7-9, 10-12, and ≥ 13 years), age (60-69, 70-79, and 80-89 years), and gender. CONCLUSIONS: We provide contemporary education-, age-, and gender-stratified norms for the MMSE, derived from a large, community-dwelling elderly Korean population sample, which could be useful in evaluating individual MMSE scores.

Association of Central Serous Chorioretinopathy with Psychosocial Factors is Dependent on Its Phase and Subtype.

PURPOSE: To analyze the psychosocial factors associated with central serous chorioretinopathy (CSC) according to its phases and subtypes and to correlate the factors with the extent of choroidal hyperpermeability. METHODS: Age- and sex-matched CSC patients and controls (n = 37 in each group) were enrolled, and their psychosocial factors were compared. CSC was divided into two phases (active and inactive), and active CSC was further divided into two subtypes (acute and chronic). The correlations between the size of the hyperpermeable choroidal lesion identified on indocyanine green angiography and psychosocial factors were examined. RESULTS: Active CSC patients experienced more stressful events (p = 0.030), were more depressive (p = 0.037), and felt less emotional (p = 0.014) and informational (p = 0.014) support than the matched controls, whereas inactive CSC patients were comparable to the matched controls in all psychosocial factors. Among the active CSC patients, acute patients were more depressive (p = 0.029), while chronic patients experienced more stressful events (p = 0.024) than their matched controls. The size of the hyperpermeable choroidal lesion was correlated with the severity of depression in acute patients. CONCLUSIONS: Association of CSC with psychosocial factors was dependent on the phase and subtype of CSC. Psychosocial factors were associated with CSC in the active phase, and severity of depression was correlated with the size of the choroidal pathology in acute active CSC. Further prospective studies to investigate if psychosocial factors can trigger CSC are warranted.

Development of the ubiquitous spaced retrieval-based memory advancement and rehabilitation training program.

OBJECTIVE: The Ubiquitous Spaced Retrieval-based Memory Advancement and Rehabilitation Training (USMART) program was developed by transforming the spaced retrieval-based memory training which consisted of 24 face-to-face sessions into a self-administered program with an iPAD app. The objective of this study was to evaluate the feasibility and efficacy of USMART in elderly subjects with mild cognitive impairment (MCI). METHODS: Feasibility was evaluated by checking the satisfaction of the participants with a 5-point Likert scale. The efficacy of the program on cognitive functions was evaluated by the Korean version of the Consortium to Establish a Registry for Alzheimer's Disease Neuropsychological Assessment Battery before and after USMART. RESULTS: Among the 10 participants, 7 completed both pre- and post-USMART assessments. The overall satisfaction score was 8.0±1.0 out of 10. The mean Word List Memory Test (WLMT) scores significantly increased after USMART training after adjusting for age, educational levels, baseline Mini-Mental Status Examination scores, and the number of training sessions (pre-USMART, 16.0±4.1; post-USMART, 17.9±4.5; p=0.014, RM-ANOVA). The magnitude of the improvements in the WLMT scores significantly correlated with the number of training sessions during 4 weeks (r=0.793; p=0.033). CONCLUSION: USMART was effective in improving memory and was well tolerated by most participants with MCI, suggesting that it may be a convenient and cost-effective alternative for the cognitive rehabilitation of elderly subjects with cognitive impairments. Further studies with large numbers of participants are necessary to examine the relationship between the number of training sessions and the improvements in memory function.

Development of a screening algorithm for Alzheimer's disease using categorical verbal fluency.

We developed a weighted composite score of the categorical verbal fluency test (CVFT) that can more easily and widely screen Alzheimer's disease (AD) than the mini-mental status examination (MMSE). We administered the CVFT using animal category and MMSE to 423 community-dwelling mild probable AD patients and their age- and gender-matched cognitively normal controls. To enhance the diagnostic accuracy for AD of the CVFT, we obtained a weighted composite score from subindex scores of the CVFT using a logistic regression model: logit (case)  = 1.160+0.474× gender +0.003× age +0.226× education level - 0.089× first-half score - 0.516× switching score -0.303× clustering score +0.534× perseveration score. The area under the receiver operating curve (AUC) for AD of this composite score AD was 0.903 (95% CI = 0.883 - 0.923), and was larger than that of the age-, gender- and education-adjusted total score of the CVFT (p<0.001). In 100 bootstrapped re-samples, the composite score consistently showed better diagnostic accuracy, sensitivity and specificity for AD than the total score. Although AUC for AD of the CVFT composite score was slightly smaller than that of the MMSE (0.930, p = 0.006), the CVFT composite score may be a good alternative to the MMSE for screening AD since it is much briefer, cheaper, and more easily applicable over phone or internet than the MMSE.

A normative study of the disability assessment for dementia in community-dwelling elderly koreans.

OBJECTIVE: We investigated demographic influences on Korean version of Disability Assessment for Dementia (DAD-K) performance and developed normative data for DAD-K. METHODS: The DAD-K was administered to 2362 normal controls (NCs), 296 patients with mild cognitive impairment (MCI) and 293 patients with Alzheimer's disease (AD). MANOVA and ROC curve analysis were used to compare DAD-K performance and the diagnostic accuracy of DAD-K, respectively. The demographic influence on DAD-K scores was analyzed by multiple linear regression and ANOVA. Normative DAD-K data were calculated using natural logarithmic transformation. RESULTS: All DAD-K scores were significantly different among groups (p<0.001). Post hoc analysis showed that instrumental activities of daily living (IADL), executive function and DAD-K total scores began to decline significantly in the very mild stage of AD, whereas the basic activities of daily living (BADL) scores began to decline in the mild stage of AD. The area under the ROC curve differentiating MCI or AD from NC was 0.737 and that differentiating AD from MCI or NC was 0.911. IADL and planning and organization scores were influenced by age and education and performance and DAD-K total scores were influenced by education. CONCLUSION: The demographic influences on DAD-K scores are not conspicuous and are mainly limited to the IADL and planning and organization scores. Unitary or minimally stratified norms for a specific population were developed for DAD-K application. Our results suggest that the DAD-K is useful for differentiating NC or MCI from AD but not as powerful for differentiating NC from MCI.

Altered categorization of semantic knowledge in Korean patients with Alzheimer's disease.

Decreased semantic memory has been suggested as an early sign in Alzheimer's disease (AD). This study aimed to investigate qualitative differences in semantic network between AD patients and healthy community-dwelling elderly controls. Category fluency test (animal) was administered to 416 AD patients and 339 controls. Mini-Mental State Examination was administered to evaluate global cognition. Patterns of semantic organization were compared between the two groups using hierarchical clustering (HC) and multidimensional scaling (MDS). In HC analysis, whereas a logically organized semantic structure was observed in the control group with an elaborate use of general knowledge, the semantic structure of the patient group was not interpretable. In MDS mapping, two dimensions of domesticity and size were identified in the control group. The MDS map of AD patients showed mixed features of perception and emotion. Knowledge-based semantic associations may be impaired in AD patients, which can add an extra dimension to cognitive impairments in AD patients regardless of quantitative facets of cognitive dysfunction.

Impaired design fluency is a marker of pathological cognitive aging; results from the Korean longitudinal study on health and aging.

OBJECTIVE: We investigated neuropsychological markers that can be used to discriminate pathological cognitive aging from normal cognitive aging. METHODS: We administered frontal lobe function tests including the Wisconsin Card Sorting Test (WCST), digit span test, lexical fluency test, fixed condition design fluency test, and Trail Making Test B (TMT-B) to 92 individuals with pathological cognitive aging (PCA) and 222 individuals with normal cognitive aging (NCA). We examined the main effects of participants' diagnoses (PCA, NCA) and age (65-69 years old, 70-74 years old and 75 years old or over) on their test performance using multivariate analysis of variance. RESULTS: The main effects of both the diagnosis (F=2.860, p=0.002) and the age group (F=2.484, p<0.001) were significant. The PCA group showed lower performance on the backward digit span test (F=14.306, p<0.001), fixed condition design fluency test (F=8.347, p=0.004) and also exhibited perseverative errors in the WCST (F=4.19, p=0.042) compared with the NCA group. The main effect of the diagnosis on the backward digit span test and the fixed condition design fluency test remained significant after Bonferroni correction. The main effect of age remained significant in the TMT-B (F=8.737, p<0.001) after Bonferroni correction. Other test scores were not influenced by diagnosis or age. CONCLUSION: The design fluency task may be a good neuropsychological marker to assess pathological cognitive aging.