Early Diagnosis of Alzheimer's Disease Using Deep Feedforward Neural Networks

Overview

Alzheimer’s Disease (AD), the most common cause of death among people over the age of 65, currently has no prevention methods or cures. However, there are several drugs that can delay or prevent symptoms from worsening, making early diagnosis very important in maintaining a patient’s quality of life. The aim of this project was to develop a Supervised Machine Learning model that can classify patients as Normal Control, having Mild Cognitive Impairment (MCI) or AD with a high accuracy. Clinical/genetic data including age, family history of AD, and APOE genotype as well as sMRI imaging data were acquired for 229 Normal, 254 MCI (due-to-AD), and 153 AD subjects. Non-brain regions were removed from the sMRI volume scans and 2D images in the sagittal, axial, and coronal orientations were obtained. Features were extracted from images using transfer learning with the Inception-V3 Convolutional Neural Network. A machine learning model was created by merging the sMRI image feature input branches with a multi-layer perceptron containing the AD risk factor inputs. Data augmentation, k-fold cross-validation for hyperparameter optimization, and dropout regularization were used to reduce model overfitting. These comprehensive steps resulted in model prediction accuracy of 91% with a sensitivity of 82.86% for MCI and 88% for AD. A user interface was developed for entering model inputs and getting predictions. Overall, this model can serve as a non-invasive early diagnostic tool for AD and its prodromal form, MCI, which is essential for coming up with better patient care and treatment options.

Competition history

• AJAS 2019 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• AJAS None projects
• Every AJAS 2019 project