Hybrid Model for Alzheimer Disease Prediction from Electronic Health Records of patients

The chronic progressive neurodegenerative disorder Alzheimer's disease (AD) renders various challenges to diagnose and provide treatments early on. Standard diagnostic methods are based on cognitive
assessment, imaging, and clinical examination, each of which are costly and time consuming. In this research, a new method is introduced for utilizing Electronic Health Records
(EHR) in the early diagnosis of AD. For the
identification of relevant biomarkers in unstructured
as well as structured EHRs like patient history,
demographics,
administered
medication,
and
cognitive tests, the proposed method integrates ML
with feature selection methods. The study utilizes a
hybrid approach that integrates Transformer-based
natural language processing for structured data and
ensemble learning for processing text-based
information. Experimental results on a large-scale
EHR dataset illustrate that our model is superior to
conventional ML methods in terms of predictive
accuracy, sensitivity, and specificity. The proposed
system provides a scalable and interpretable solution
for physicians for the early detection