Deep Learning-Based Screening Test for Cognitive Impairment Using Basic Blood Test Data for Health Examination

Importance of serum albumin in machine learning-based prediction of cognitive function in the elderly using a basic blood test

Introduction

In this study, we investigated the correlation between serum albumin levels and cognitive function, and examined the impact of including serum albumin values in the input layer on the prediction accuracy when forecasting cognitive function using deep learning and other machine learning models.

Methods

We analyzed the electronic health record data from Osaka Medical and Pharmaceutical University Hospital between 2014 and 2021. The study included patients who underwent cognitive function tests during this period; however, patients from whom blood test data was not obtained up to 30 days before the cognitive function tests and those with values due to measurement error in blood test results were excluded. The Mini-Mental State Examination (MMSE) was used as the cognitive function test, and albumin levels were examined as the explanatory variable. Furthermore, we estimated MMSE scores from blood test data using deep learning models (DLM), linear regression models, support vector machines (SVM), decision trees, random forests, extreme gradient boosting (XGBoost), and light gradient boosting machines (LightGBM).

Results

Out of 5,017 patients who underwent cognitive function tests, 3,663 patients from whom blood test data had not been obtained recently and two patients with values due to measurement error were excluded. The final study population included 1,352 patients, with 114 patients (8.4%) aged below 65 and 1,238 patients (91.6%) aged 65 and above. In patients aged 65 and above, the age and male sex showed significant associations with MMSE scores of less than 24, while albumin and potassium levels showed negative associations with MMSE scores of less than 24. Comparing MMSE estimation performance, in those aged below 65, the mean squared error (MSE) of DLM was improved with the inclusion of albumin. Similarly, the MSE improved when using SVM, random forest and XGBoost. In those aged 65 and above, the MSE improved in all models.

Discussion

Our study results indicated a positive correlation between serum albumin levels and cognitive function, suggesting a positive correlation between nutritional status and cognitive function in the elderly. Serum albumin levels were shown to be an important explanatory variable in the estimation of cognitive function for individuals aged 65 and above.

Predictive deep learning models for cognitive risk using accessible data

The early detection of mild cognitive impairment (MCI) is crucial to preventing the progression of dementia. However, it necessitates that patients voluntarily undergo cognitive function tests, which may be too late if symptoms are only recognized once they become apparent. Recent advances in deep learning have improved model performance, leading to applied research in various predictive problems. Studies attempting to estimate dementia and the risk of MCI based on readily available data are being conducted, with the hope of facilitating the early detection of MCI. The data used for these predictions vary widely, including facial imagery, voice recordings, blood tests, and inertial information during walking. Deep learning models that make predictions based on these data sources have been proposed. This article summarizes recent research efforts to predict the risk of dementia using easily accessible data. As research progresses and more accurate predictions become feasible, simple tests could be incorporated into daily life to monitor one's personal health status and to facilitate an early intervention.

Estimation of cognitive impairment in chronic pain patients and characteristics of estimated mild cognitive impairment

Background

Patients with chronic pain suffer from psychological effects such as anxiety due to the pain itself. Pain can not only impair activities of daily living (ADL) and quality of life (QOL), but also impair cognitive function. Therefore, in this study, we aimed to estimate the cognitive function of chronic pain patients using a deep neural network (DNN) model that has already been implemented in society. We investigated the characteristics of patients presumed to have mild cognitive impairment (MCI) and, at the same time, verified the relationship with the questionnaire commonly used in chronic pain research, which is administered by 43 university affiliated hospitals and medical institutions participating in the chronic pain research group of the Ministry of Health, Labor and Welfare in Japan (assessment batteries).

Method

The study included 114 outpatients from a multidisciplinary pain clinic, and we estimated their Mini-Mental State Examination (MMSE) scores based on age and basic blood test data (23 items). Furthermore, we classified the estimated MMSE scores of chronic pain patients into two groups based on a cutoff score of 27, which indicates MCI, and compared the blood data and assessment batteries. Additionally, we used a control group of 252 healthy adults aged 45 years or older who visited a dementia prevention outpatient clinic for comparison with the MMSE scores of chronic pain patients.

Result

The MMSE scores in chronic pain patients were below the cutoff for MCI. When classified into two groups based on the estimated MMSE score of 27 points, WBC, RBC, Hb, Hct, PLT, UA, BUN, creatinine, Triglyceride, and γ-GT were significantly higher in the blood data. In the MCI group, PDAS values were significantly lower. Furthermore, only in the non-MCI group, a significant correlation was found between the estimated MMSE value and BPI, PDAS, and Locomo. The estimated MMSE scores were significantly lower in chronic pain patients than in healthy adults (p = 0.04).

Conclusion

Patients with chronic pain may exhibit cognitive impairment due to systemic metabolic disturbances. This suggests that chronic pain affects activities of daily living, resulting in systemic metabolic disorders.

Recent deep learning models for dementia as point-of-care testing: Potential for early detection

Deep learning has been intensively researched over the last decade, yielding several new models for natural language processing, images, speech and time series processing that have dramatically improved performance. This wave of technological developments in deep learning is also spreading to medicine. The effective use of deep learning in medicine is concentrated in diagnostic imaging-related applications, but deep learning has the potential to lead to early detection and prevention of diseases. Physical aspects of disease that went unnoticed can now be used in diagnosis with deep learning. In particular, deep learning models for the early detection of dementia have been proposed to predict cognitive function based on various information such as blood test results, speech, and the appearance of the face, where the effects of dementia can be seen. Deep learning is a useful diagnostic tool, as it has the potential to detect diseases early based on trivial aspects before clear signs of disease appear. The ability to easily make a simple diagnosis based on information such as blood test results, voice, pictures of the body, and lifestyle is a method suited to point-of-cate testing, which requires immediate testing at the desired time and place. Over the past few years, the process of predicting disease can now be visualized using deep learning, providing insights into new methods of diagnosis.

Relationship Between Cognitive Function, Oral Conditions and Systemic Metabolic Function in the Elderly

Systemic metabolic disorders, including lifestyle-related diseases, are known risk factors for dementia. Furthermore, oral diseases such as periodontal disease and tooth decay are also associated with systemic metabolic disorders such as lifestyle-related diseases, and have also been reported to be indicators of risk factors for developing dementia. In this study, we investigated the relationship between cognitive function, oral conditions and systemic metabolic function in the elderly. We investigated the number of healthy teeth, the number of prosthetic teeth fitted, the number of missing prosthetic teeth, etc., in 41 elderly patients (69.7 ± 5.6 years old). Cognitive function was evaluated by the Mini Mental State Examination (MMSE). We also estimated MMSE scores for each subject using deep learning-based assessment of MMSE scores. This deep learning method enables the estimation of the MMSE score based on basic blood test data from medical examinations and reflects the systemic metabolic state including lifestyle-related diseases. The estimated MMSE score correlated negatively with age (r = -0.381), correlated positively with the number of healthy teeth (r = 0.37), and correlated negatively with the number of missing prosthetic teeth (r = -0.39). This relationship was not found in the measured MMSE scores. A negative correlation (r = -0.36) was found between age and the current number of teeth and a positive correlation (r = 0.37) was found between age and the number of missing prosthetic teeth. A positive correlation was found between the number of teeth requiring prosthesis and lifestyle-related diseases. The deep learning-based estimation method of cognitive function clearly demonstrated the close relationship between oral health condition, systemic metabolic function and the risk of cognitive impairment. It was determined that the smaller the number of existing teeth and the larger the number of missing prosthetic teeth, the higher is the risk of cognitive impairment. Systemic metabolic function is presumed to affect oral health and cognitive function. Interestingly, no such relationship was found in the measured MMSE scores. There are two possible reasons for this. The first is that MMSE is a subjective test and is less accurate in assessing cognitive function. The second is that because the MMSE estimated based on blood data using deep learning is calculated based on the metabolic function, it has a stronger correlation with the oral health condition affected by the metabolic function. In conclusion, oral health condition may predict cognitive impairment in the elderly.

Estimation of Human Cerebral Atrophy Based on Systemic Metabolic Status Using Machine Learning

Background

Based on the assumption that systemic metabolic disorders affect cognitive function, we have developed a deep neural network (DNN) model that can estimate cognitive function based on basic blood test data that do not contain dementia-specific biomarkers. In this study, we used the same DNN model to assess whether basic blood data can be used to estimate cerebral atrophy.

Methods

We used data from 1,310 subjects (58.32 ± 12.91years old) enrolled in the Brain Doc Bank. The average Mini Mental State Examination score was 28.6 ± 1.9. The degree of cerebral atrophy was determined using the MRI-based index (GM-BHQ). First, we evaluated the correlations between the subjects' age, blood data, and GM-BHQ. Next, we developed DNN models to assess the GM-BHQ: one used subjects' age and blood data, while the other used only blood data for input items.

Results

There was a negative correlation between age and GM-BHQ scores (r = -0.71). The subjects' age was positively correlated with blood urea nitrogen (BUN) (r = 0.40), alkaline phosphatase (ALP) (r = 0.22), glucose (GLU) (r = 0.22), and negative correlations with red blood cell counts (RBC) (r = −0.29) and platelet counts (PLT) (r = −0.26). GM-BHQ correlated with BUN (r = −0.30), GLU (r = −0.26), PLT (r = 0.26), and ALP (r = 0.22). The GM-BHQ estimated by the DNN model with subject age exhibited a positive correlation with the ground truth GM-BHQ (r = 0.70). Furthermore, even if the DNN model without subject age was used, the estimated GM-BHQ showed a significant positive correlation with ground truth GM-BHQ (r = 0.58). Age was the most important variable for estimating GM-BHQ.

Discussion

Aging had the greatest effect on cerebral atrophy. Aging also affects various organs, such as the kidney, and causes changes in systemic metabolic status, which may contribute to cerebral atrophy and cognitive impairment. The DNN model may serve as a new screening test for dementia using basic blood tests for health examinations. Finally, the blood data reflect systemic metabolic disorders in each subject—this method may thus contribute to personalized care.

Machine Learning-Based Assessment of Cognitive Impairment Using Time-Resolved Near-Infrared Spectroscopy and Basic Blood Test

We have demonstrated that machine learning allows us to predict cognitive function in aged people using near-infrared spectroscopy (NIRS) data or basic blood test data. However, the following points are not yet clear: first, whether there are differences in prediction accuracy between NIRS and blood test data; second, whether there are differences in prediction accuracy for cognitive function in linear models and non-linear models; and third, whether there are changes in prediction accuracy when both NIRS and blood test data are added to the input layer. We used a linear regression model (LR) for the linear model and random forest (RF) and deep neural network (DNN) for the non-linear model. We studied 250 participants (mean age = 73.3 ± 12.6 years) and assessed cognitive function using the Mini Mental State Examination (MMSE) (mean MMSE scores = 22.9 ± 6.1). We used time-resolved NIRS (TNIRS) to measure absolute concentrations of hemoglobin and optical pathlength at rest in the bilateral prefrontal cortices. A basic blood test was performed on the same day. We compared predicted MMSE scores and grand truth MMSE scores; prediction accuracies were evaluated using mean absolute error (MAE) and mean absolute percentage error (MAPE). We found that (1) the DNN-based prediction using TNIRS data exhibited lower MAE and MAPE compared with those using blood test data, (2) the difference in MAPE between TNIRS and blood test data was only 0.3%, (3) adding TNIRS data to the blood test data of the input layer only improved MAPE by 1.0% compared to the use of blood test data alone, whereas the use of the blood test data alone exhibited the prediction accuracy with 81.8% sensitivity and 91.3% specificity (N = 202, repeated five-fold cross validation). Given these findings and the benefits of using blood test data (low cost and large-scale screening possible), we concluded that the DNN model using blood test data is still the most suitable for mass screening.

Effects of Exercise-Diet Therapy on Cognitive Function in Healthy Elderly People Evaluated by Deep Learning Based on Basic Blood Test Data

BACKGROUND

Recent studies reported that vascular cognitive impairment in the elderly caused by arteriosclerosis plays an important role in cognitive disorders in both vascular dementia and Alzheimer's disease. In addition, systemic metabolic disorders such as oxygen metabolism dysfunction could be risk of dementia. Based on these findings, we have developed a deep neural network-based screening test (DNN-based test) of cognitive function using basic blood test data, which allowed prediction of cognitive function expressed by Mini Mental State Examination (MMSE) scores.

AIM

Here, we investigated whether the DNN-based test could be applicable to assessment of the effects of exercise-diet on cognitive function.

METHODS

We studied the following two groups: (1) seven subjects of the local fitness gym (68.6 ± 3.2 years old). We evaluated cognitive function by the DNN-based test using blood data before and after the intervention (for 3 months). These results were compared with the measured MMSE score. (2) we studied a total of 230 subjects (67.9 ± 7.4 years old) who were members of the Tsuminory health class (Apple classroom). We assessed cognitive function by the DNN-based test before and after the intervention (for 2 months). We compared the predicted MMSE scores by the DNN-based test before and after the 2-month intervention.

RESULTS

In the first group, the MMSE score predicted by the DNN-based test increased from 27.1 ± 0.8 to 27.6 ± 0.7 after the intervention period (p = 0.024). The measured MMSE score also increased after exercise, but not significant (P = 0.28). In the second group, the exercise-diet therapy increased the predicted MMSE scores in 189 cases (p < 0.001). In contrast, the therapy significantly reduced the mean MMSE score (p < 0.001).

DISCUSSION

The MMSE score predicted by the DNN-based test were increased by exercise-diet therapy in most subjects. The DNN-based test may be useful to monitor the effect of exercise-diet therapy on cognitive function aged people.