The last decade Artificial Intelligence (AI) applications grew rapidly as both data col-lection/storage and machine learning modeling constantly evolve. The goal of these applications is to automate predictions and decisions without human interaction, in a wide range of fields, with the medical field being one of the most important ones.
The aim of this research is to develop Deep Learning models that can accurately pre-dict whether a subject is Healthy (HC), suffers from Mild Cognitive Impairment (MCI), or Alzheimer’s Disease (AD) with the use of Electroencephalogram (EEG) da-ta. EEG is a test that measures the electrical brain activity by attaching multiple elec-trodes on the scalp. Identifying AD in an early stage increases the chances of efficient treatment of dementia. And identifying MCI, is important for identifying AD early, as it is considered a possible early stage of AD. EEG
The data were split into 2-second no-overlapping segments and preprocessing methods were used on the segments like filtering and Fast Fourier Transform (FFT). For this research only RNN with Long Short-Term Memory (LSTM) layers were used with some of the models including a convolutional layer. One of the models manage to achieve a 99.0% accuracy on the validation set segments. Finally a different method was developed to split the data by patient, so training and validation datasets do not include segments from the same patient.
Collections
Show Collections
