Early Detection of Dementia using Machine Learning Techniques

Rahul Singhal

Abstract

Today, as many as 7% of adults aged sixty and above suffer from dementia. Over four million individuals in India
suffer from dementia in some form. Dementia affects at least 44 million people worldwide, making it a global health
concern that must be tackled. Dementia is a condition of the mind rather than a disease. It is defined as a significant
deterioration in mental function from a prior higher level that interferes with a person’s everyday activities. Abnormal
brain alterations create the disorders included under the umbrella term "dementia." This condition causes a
deterioration in thinking abilities, also known as cognitive capacities, that is severe enough to interfere with everyday
living and independence. It also has an impact on one's conduct, emotions, and relationships. Early identification and
diagnosis of dementia can help in halting disease progression and minimize stress and morbidity in patients and
caregivers. The objective of this work is to implement different machine learning-based models to identify dementia
using gathered data and brain MRI scans in general practice. The approach might be beneficial for detecting persons
who may have dementia but have not been formally diagnosed or who have a tendency for it. To improve the results,
appropriate feature engineering and data preparation were used. Finally, using suitable performance assessment
parameters, the outcomes from all of the models were compared.