Hierarchical Transformer for Early Detection of Alzheimer’s Disease Open Access

Abstract

Alzheimer’s disease is an irreversible disease that severely affect the brain functions and life

quality of the patients. For now, there is no effective cure for the disease. Therefore, this unfortunate fact makes the early detection of Alzheimer’s disease vital. The early stage of the Alzheimer’s disease, Mild Cognitive Impairment (MCI), normally involve loss in memory, language ability, and object recognition ability. In this paper, we present a new dataset that includes the transcribed audio of the MCI patients and healthy subject. We also present a hierarchical transformer-based model and the corresponding analysis for the MCI/health classification task on our dataset

Table of Contents

Contents

1 Introduction .................1

2 Related Works .................4

2.1 DetectionofAlzheimer’sDisease................. 4

2.2 Transformers for Natural Language Understanding . . . . . . 6

3 Dataset..............................9

3.1 B-SHARP: Brain, Stress, Hypertension, and Aging Research Program.............................. 9

3.2 SpeechTaskProtocol....................... 11

3.3 ComparisontoDementiaBank.................. 13

4 Approaches 16

4.1 BaselinesApproaches....................... 16

4.1.1 Convolutional Neural Network(CNN) . . . . . . . . . . 16

4.1.2 CNN-LSTM........................ 16

4.2 HierarchicalTransformer..................... 17

5 Experiments 21

5.1 DataSplit............................. 21

5.2 Transformers ........................... 22

5.3 PerformanceonIndividualTasks ................ 23

5.4 PerformancewithEnsembleApproaches . . . . . . . . . . . . 24

6 Analysis 26

6.1 AttentionAnalysis ........................ 26

6.1.1 MethodsandMeasures.................. 26

6.1.2 RoBERTaAnalysis.................... 28

6.1.3 BERT Analysis and Comparison to RoBERTa . . . . . 36

6.1.4 Findingvalidation .................... 40

6.2 EnsembleAnalysis ........................ 45

6.2.1 RoBERTaEnsembleAnalysis .............. 45

7 Conclusion 53

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Computer Science
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Computer Science
Keyword	Natural Language Processing Alzheimer's disease detection Deep Learning
Committee Chair / Thesis Advisor	Jinho Choi, Emory University
Committee Members	Michelangelo Grigni, Emory University Hiram Maxim, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Hierarchical Transformer for Early Detection of Alzheimer’s Disease ()	2020-04-13 15:31:42 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions