Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Linguistic-Based Mild Cognitive Impairment Detection Using Informative Loss

Ali Pourramezan Fard, Mohammad H. Mahoor, Muath Alsuhaibani, Hiroko H. Dodgec

TL;DR

A framework that analyzes transcripts generated from video interviews collected within the I-CONECT study project, a randomized controlled trial aimed at improving cognitive functions through video chats, to distinguish between Mild Cognitive Impairment (MCI) and Normal Cognitive (NC) conditions in older adults.

Abstract

This paper presents a deep learning method using Natural Language Processing (NLP) techniques, to distinguish between Mild Cognitive Impairment (MCI) and Normal Cognitive (NC) conditions in older adults. We propose a framework that analyzes transcripts generated from video interviews collected within the I-CONECT study project, a randomized controlled trial aimed at improving cognitive functions through video chats. Our proposed NLP framework consists of two Transformer-based modules, namely Sentence Embedding (SE) and Sentence Cross Attention (SCA). First, the SE module captures contextual relationships between words within each sentence. Subsequently, the SCA module extracts temporal features from a sequence of sentences. This feature is then used by a Multi-Layer Perceptron (MLP) for the classification of subjects into MCI or NC. To build a robust model, we propose a novel loss function, called InfoLoss, that considers the reduction in entropy by observing each sequence of sentences to ultimately enhance the classification accuracy. The results of our comprehensive model evaluation using the I-CONECT dataset show that our framework can distinguish between MCI and NC with an average area under the curve of 84.75%.

Linguistic-Based Mild Cognitive Impairment Detection Using Informative Loss

TL;DR

A framework that analyzes transcripts generated from video interviews collected within the I-CONECT study project, a randomized controlled trial aimed at improving cognitive functions through video chats, to distinguish between Mild Cognitive Impairment (MCI) and Normal Cognitive (NC) conditions in older adults.

Abstract

This paper presents a deep learning method using Natural Language Processing (NLP) techniques, to distinguish between Mild Cognitive Impairment (MCI) and Normal Cognitive (NC) conditions in older adults. We propose a framework that analyzes transcripts generated from video interviews collected within the I-CONECT study project, a randomized controlled trial aimed at improving cognitive functions through video chats. Our proposed NLP framework consists of two Transformer-based modules, namely Sentence Embedding (SE) and Sentence Cross Attention (SCA). First, the SE module captures contextual relationships between words within each sentence. Subsequently, the SCA module extracts temporal features from a sequence of sentences. This feature is then used by a Multi-Layer Perceptron (MLP) for the classification of subjects into MCI or NC. To build a robust model, we propose a novel loss function, called InfoLoss, that considers the reduction in entropy by observing each sequence of sentences to ultimately enhance the classification accuracy. The results of our comprehensive model evaluation using the I-CONECT dataset show that our framework can distinguish between MCI and NC with an average area under the curve of 84.75%.
Paper Structure (27 sections, 16 equations, 17 figures, 8 tables)

This paper contains 27 sections, 16 equations, 17 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Speech-base Detection of AD/ADRD
  4. Review of Research on I-CONECT
  5. Methodology
  6. Definition
  7. SE Module
  8. SCA Module
  9. Framework Architecture
  10. InfoLoss
  11. Objective Function and Inference Classification
  12. Frequency-Based Uncertainty
  13. Loss Function
  14. Evaluation and Experimental Results
  15. I-CONECT Dataset
  16. ...and 12 more sections

Figures (17)

  • Figure 1: The architecture of our proposed framework. The input to the framework is a sequence of $\gamma$ sentences. The output of the framework is a 2-dimensional vector representing the estimated probabilities of MCI and NC classes.
  • Figure 2: This figure shows the SCA module. The input to the SCA module is a sequence of sentential representation vectors, and the output is the corresponding temporal representation vector.
  • Figure 3: The figure shows an example of sequences of sentences for subjects in the I-CONECT dataset. Assuming m is the number of sequences for subject 1, and n is the number of sequences for subject 2. If m is smaller than n, observing the class label for each sequence related to Subject 1 provides more information about the ultimate class label for Subject 1 than observing each sequence related to Subject 2. In simpler terms, observing each sequence associated with Subject 1 contributes more to reducing the entropy of Subject 1 compared to the entropy reduction of Subject 2 performed by observing its related sequences.
  • Figure 4: The distribution of themes per subject among individuals with MCI and NC.
  • Figure 5: The distribution of sentences per subject for MCI (left chart) and NC (right chart) groups.
  • ...and 12 more figures