NORMA eResearch @NCI Library

Genetic Algorithm Optimized Deep Learning Model for Parkinson Disease Severity Detection

Srivastava, Shveta (2021) Genetic Algorithm Optimized Deep Learning Model for Parkinson Disease Severity Detection. Masters thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Master of Science)
Download (1MB) | Preview
[img]
Preview
PDF (Configuration manual)
Download (1MB) | Preview

Abstract

Parkinson’s disease (PD) is reflected by several motor function disabilities such as tremors, loss of balance, speech impairment, etc. due to loss of dopamine neurotransmitter. While researchers have been building models for diagnosing and classifying PD patients based on gait data, speech data or handwriting data accounting for visible symptoms only, PD patients show non-motor symptoms such as sleep disorder, neuropsychological symptoms, cognitive impairment, olfactory loss, much before the actual diagnosis. This study involves coupling both motor as well as non-motor symptoms of PD patients from up to 10 years longitudinal records in PPMI database and building an optimized deep learning model for PD severity classification based on the Hoehn & Yahr index. This longitudinal complex dataset brings along challenges of dealing with high volume of missing and inconsistent data in various assessments at different time points. To deal with such complexity, the proposed model for this study is a type of Recurrent Neural Network, Long Short-Term Memory (LSTM) model which learns well from data with long term dependencies. This multi-time step model gives high accuracy of 88% for multi-class severity prediction. The LSTM model is also coupled with heuristic evolutionary search algorithm, Genetic Algorithm (GA) considering the vast dimensionality of the longitudinal heterogenous records and to find the optimal window size and number of LSTM units to minimize the loss function, MSE. The results have been compared with baseline model, state-of-the-art approach, MLP (Multi-layer Perceptron) which is a feed forward network. The novel GA-LSTM model used in this project shows reduced RMSE score of 0.33 as compared to 0.72 in MLP. Multiple Machine learning algorithms have also been implemented where XGBoost shows the highest accuracy of 89%.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science

Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software

H Social Sciences > HM Sociology > Information Science > Communication > Medical Informatics
R Medicine > RC Internal medicine > RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry > Neurology. Diseases of the Nervous System.
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Clara Chan
Date Deposited: 14 Dec 2021 16:18
Last Modified: 14 Dec 2021 16:18
URI: https://norma.ncirl.ie/id/eprint/5226

Actions (login required)

View Item View Item