Classification of gait signals into different neurodegenerative diseases using statistical analysis and recurrence quantification analysis

Pooja Prabhu, A. K. Karunakar, H. Anitha, N. Pradhan

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Among all the biological signals, gait signal is one of the better features to detect movement disorders caused by a malfunction in parts of the brain and nervous system. Usually, identifying and evaluating movement disorders caused due to neurodegenerative diseases solely depends on a physicians experience. Different diseases having gait abnormalities generate a unique gait characteristic. Traditionally, Fourier analysis is used to understand the gait characteristic, thereby predicting potential diseases. Fourier analysis assumes the gait signal to be stationary, linear and noiseless which is not a reality. To overcome this, Recurrence Quantification Analysis (RQA) is used in this study to quantify gait parameters. RQA has proved to be one of the best tools for non-linear, non-stationary and short length data. It is used to quantify heart rate variability, ventricular fibrillation, wrist pulse and growth of bladder. This paper uses RQA in understanding the dynamics of human gait and the parameters obtained are used as a feature for classification using Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). This study considered thirteen subjects for the classification of gait signals of patients with Neurodegenerative diseases (Amyotrophic Lateral Sclerosis, Huntington and Parkinson) and thirteen healthy control subjects using two different classification models like Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). Features were extracted after statistical analysis and RQA, and Hill-climbing feature selection method was used to optimize the feature set. The accuracy deduced after binary classification using SVM and PNN ranged from 96% to 100%.

Original languageEnglish
JournalPattern Recognition Letters
DOIs
Publication statusAccepted/In press - 01-01-2018

Fingerprint

Neurodegenerative diseases
Statistical methods
Support vector machines
Fourier analysis
Neural networks
Neurology
Feature extraction
Brain

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

@article{9b7e1b475b85475da99b8c358cb5a3c7,
title = "Classification of gait signals into different neurodegenerative diseases using statistical analysis and recurrence quantification analysis",
abstract = "Among all the biological signals, gait signal is one of the better features to detect movement disorders caused by a malfunction in parts of the brain and nervous system. Usually, identifying and evaluating movement disorders caused due to neurodegenerative diseases solely depends on a physicians experience. Different diseases having gait abnormalities generate a unique gait characteristic. Traditionally, Fourier analysis is used to understand the gait characteristic, thereby predicting potential diseases. Fourier analysis assumes the gait signal to be stationary, linear and noiseless which is not a reality. To overcome this, Recurrence Quantification Analysis (RQA) is used in this study to quantify gait parameters. RQA has proved to be one of the best tools for non-linear, non-stationary and short length data. It is used to quantify heart rate variability, ventricular fibrillation, wrist pulse and growth of bladder. This paper uses RQA in understanding the dynamics of human gait and the parameters obtained are used as a feature for classification using Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). This study considered thirteen subjects for the classification of gait signals of patients with Neurodegenerative diseases (Amyotrophic Lateral Sclerosis, Huntington and Parkinson) and thirteen healthy control subjects using two different classification models like Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). Features were extracted after statistical analysis and RQA, and Hill-climbing feature selection method was used to optimize the feature set. The accuracy deduced after binary classification using SVM and PNN ranged from 96{\%} to 100{\%}.",
author = "Pooja Prabhu and Karunakar, {A. K.} and H. Anitha and N. Pradhan",
year = "2018",
month = "1",
day = "1",
doi = "10.1016/j.patrec.2018.05.006",
language = "English",
journal = "Pattern Recognition Letters",
issn = "0167-8655",
publisher = "Elsevier",

}

TY - JOUR

T1 - Classification of gait signals into different neurodegenerative diseases using statistical analysis and recurrence quantification analysis

AU - Prabhu, Pooja

AU - Karunakar, A. K.

AU - Anitha, H.

AU - Pradhan, N.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Among all the biological signals, gait signal is one of the better features to detect movement disorders caused by a malfunction in parts of the brain and nervous system. Usually, identifying and evaluating movement disorders caused due to neurodegenerative diseases solely depends on a physicians experience. Different diseases having gait abnormalities generate a unique gait characteristic. Traditionally, Fourier analysis is used to understand the gait characteristic, thereby predicting potential diseases. Fourier analysis assumes the gait signal to be stationary, linear and noiseless which is not a reality. To overcome this, Recurrence Quantification Analysis (RQA) is used in this study to quantify gait parameters. RQA has proved to be one of the best tools for non-linear, non-stationary and short length data. It is used to quantify heart rate variability, ventricular fibrillation, wrist pulse and growth of bladder. This paper uses RQA in understanding the dynamics of human gait and the parameters obtained are used as a feature for classification using Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). This study considered thirteen subjects for the classification of gait signals of patients with Neurodegenerative diseases (Amyotrophic Lateral Sclerosis, Huntington and Parkinson) and thirteen healthy control subjects using two different classification models like Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). Features were extracted after statistical analysis and RQA, and Hill-climbing feature selection method was used to optimize the feature set. The accuracy deduced after binary classification using SVM and PNN ranged from 96% to 100%.

AB - Among all the biological signals, gait signal is one of the better features to detect movement disorders caused by a malfunction in parts of the brain and nervous system. Usually, identifying and evaluating movement disorders caused due to neurodegenerative diseases solely depends on a physicians experience. Different diseases having gait abnormalities generate a unique gait characteristic. Traditionally, Fourier analysis is used to understand the gait characteristic, thereby predicting potential diseases. Fourier analysis assumes the gait signal to be stationary, linear and noiseless which is not a reality. To overcome this, Recurrence Quantification Analysis (RQA) is used in this study to quantify gait parameters. RQA has proved to be one of the best tools for non-linear, non-stationary and short length data. It is used to quantify heart rate variability, ventricular fibrillation, wrist pulse and growth of bladder. This paper uses RQA in understanding the dynamics of human gait and the parameters obtained are used as a feature for classification using Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). This study considered thirteen subjects for the classification of gait signals of patients with Neurodegenerative diseases (Amyotrophic Lateral Sclerosis, Huntington and Parkinson) and thirteen healthy control subjects using two different classification models like Support Vector Machine (SVM) and Probabilistic Neural Network (PNN). Features were extracted after statistical analysis and RQA, and Hill-climbing feature selection method was used to optimize the feature set. The accuracy deduced after binary classification using SVM and PNN ranged from 96% to 100%.

UR - http://www.scopus.com/inward/record.url?scp=85047179147&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85047179147&partnerID=8YFLogxK

U2 - 10.1016/j.patrec.2018.05.006

DO - 10.1016/j.patrec.2018.05.006

M3 - Article

JO - Pattern Recognition Letters

JF - Pattern Recognition Letters

SN - 0167-8655

ER -