Development of an Algorithm using Machine Learning in Early Diagnosis of Dementia with the help of the Clinical and Multimodality Structural and Functional Volumetric Data

Background: Dementia is the most prevalent degenerative disease in elderly whose progression can only be prevented or delayed by early diagnosis. In this study, we proposed a two-layer model algorithm using machine learning techniques. Materials and Methods: Clinical and multimodality structural and functional volumetric data were collected from patients who received dementia screening from May 2016 to May 2017 at our institute and were stored in the programme. Now, from June 2017 to October 2017, imaging data of clinically normal patients having strong risk of dementia were analysed and a predesigned questionnaire was applied on them. They were categorised by the machine as normal or abnormal according to the previously fed data. Random Forest, Bayes Network, Logistic Regression and F-measure were used for analysis of the algorithm. Now, a year later, from June 2018 to October 2018 those people were again followed up for incidence of Mild Cognitive Impairment (MCI) and Dementia. Results: It was found that using the proposed algorithm the program could diagnose 23.8 % preclinical dementia cases, saving a year of lead time. Conclusion: Hence, this programme can save time and economic burden and can take a crucial role in early diagnosis of dementia.