Investigators: MD, PhD Yauhen Statsenko, Prof. Nazar Zaki, Tetiana Habuza, Dr. Klaus Gorkom, Dr. Taleb Al Mansoori.
Objective: We intend to find a sensitive biomarker of the early cognitive decline. By implementing the psychophysiological screening of the population into daily practice, we intend to improve the early detection of degeneration. The same biomarkers may serve as the quantitative measure of the cognitive status of elderly people and patients with neurodegenerative disease in follow-up studies.
Anticipated outcomes: Currently, disease-related brain atrophy is diagnosed or detected primarily at late stages, while early neuronal functional impairment is less detected and treated. Research on this issue will contribute to the early diagnosis of functional impairment. Supposedly, the implementation of advanced computer-aided imaging will contribute to the early diagnosis of neurodegenerative disorders.
Hypothesis: Psychophysiological test results reflect the level of brain atrophy, which can serve as a sensitive marker of a wide range of pathologies.
Methods:
We use the following groups of methods and tools for data acquisition and analysis:
- Brain MRI: Structural image acquisition with T1W-3D sequence, FLAIR, DWI. Image
- The human-driven visual analysis: An experienced neuroradiologist conducts it by using a semiquantitative protocol to estimate the level of brain atrophy in different regions.
- Voxel-based morphometry: CAT12, LST frameworks for SPM, 3DSlicer.
- Physiological-psychological testing and dynamometry: with diagnostic complex Neuro-Science Psychotest (NSP).
- Basic Statistical analysis: Descriptive and comparative statistics, principal component analysis, multivariate logistic regression analysis, pairwise correlation.
The machine learning techniques used:
- To estimate the onset of age-related cognitive changes in age groups, we use unsupervised ML algorithms. We start our exploratory analysis by looking at the separability measure of predefined age groups utilizing clustering methods, such as Simple K-Means, Canopy, Expectation Maximization, and GenClus++. Testing different numbers of clusters in terms of performance allows us to develop the possible onset of executive functioning decline. Also, we use the knee point detection algorithm to get the optimal value of clusters (age groups). Then we build pairwise distributions of every attribute in the dataset with age, and thoroughly study the features related to time. To enhance the performance of the clustering algorithms, we utilize feature selection methods.
- To predict the age group and determine the informative value of the psychophysiological tests’ components concerning their possibility to reflect age-related cognitive changes, we apply the supervised traditional ML methods. We implement several binary classification algorithms such as Support Vector Machines with linear and non-linear (radial basis function) kernels, Gaussian Naive Bayes, Bagging meta-estimator, Extra-trees classifier, Random Forest, and Multi-layer Perceptron.
Ethics: The study received an ethical review and approval by UAEU Human Research Ethics Committee (Notice Number: ERH_2019_4006 19_11).
Grant support: This work is supported by the UAEU StartUp grant 31M442 “Psychophysiological Outcomes of Brain Atrophy”; UAEU SUREgrant 1148 “Radiological and functional signs of brain aging.”