Computational Modeling of Dementia Prediction Using Deep Neural Network: Analysis on OASIS Dataset

Alzheimer is a progressive disease and it is the most prevalent neurodegenerative disorder. It is believed that the people with mild cognitive impairment are at high risk of developing this disease. According to the annual report released by the Alzheimer's Association (R) 2020, Alzheimer is the sixth leading cause of death in the United States. Thus, there is a need of educating people about this disease, reducing the risks by militating the necessary precautions to disseminate its affect by diagnosing it at early stages. It is also important to propose some recent advancement in this research which can help in early prediction of the disease using machine learning techniques. This paper intends to develop the novel algorithm by proposing changes in the designing of capsule network for best prediction results and making the model computationally efficient. The research is conducted on the Open Access Series of Imaging Studies (OASIS) dataset with dimensions (373 X 15) to diagnose the labels into two groups, as demented and non-demented. The novelty lies in conducting the in-depth research in identifying the importance of features, correlation study between factors and density of data showing status of factors by studying hierarchical examination of all the data points available using exploratory data analysis. Several optimization functions are conducted on the variables and feature selection is done to make the model faster and more accurate. The claims have been validated by showing the correlation accuracy at several iterations and layers with an admissible accuracy of 92.39%. The model is compared with state-of-art deep learning classifiers taken as benchmarks using different performance metrics. The ablation study is conducted on the proposed model using OASIS dataset to justify the predictions of the model.

Automated summary

Alzheimer's disease is a common neurodegenerative disorder that requires early diagnosis and research to reduce risks and impact. This study utilizes machine learning techniques for early prediction and improves model accuracy and efficiency through optimized algorithm design and feature selection.