Freezing of Gait Detection in Parkinson's Disease: A Subject-Independent Detector Using Anomaly Scores

Thuy T. Pham, Steven T. Moore, Simon John Geoffrey Lewis, Diep N. Nguyen, Eryk Dutkiewicz, Andrew J Fuglevand, Alistair L. McEwan, Philip H.W. Leong

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

Freezing of gait (FoG) is common in Parkinsonian gait and strongly relates to falls. Current clinical FoG assessments are patients' self-report diaries and experts' manual video analysis. Both are subjective and yield moderate reliability. Existing detection algorithms have been predominantly designed in subject-dependent settings. In this paper, we aim to develop an automated FoG detector for subject independent. After extracting highly relevant features, we apply anomaly detection techniques to detect FoG events. Specifically, feature selection is performed using correlation and clusterability metrics. From a list of 244 feature candidates, 36 candidates were selected using saliency and robustness criteria. We develop an anomaly score detector with adaptive thresholding to identify FoG events. Then, using accuracy metrics, we reduce the feature list to seven candidates. Our novel multichannel freezing index was the most selective across all window sizes, achieving sensitivity (specificity) of 96% (79%). On the other hand, freezing index from the vertical axis was the best choice for a single input, achieving sensitivity (specificity) of 94% (84%) for ankle and 89% (94%) for back sensors. Our subject-independent method is not only significantly more accurate than those previously reported, but also uses a much smaller window (e.g., 3 s versus 7.5 s) and/or lower tolerance (e.g., 0.4 s versus 2 s).

Original languageEnglish (US)
Article number7845616
Pages (from-to)2719-2728
Number of pages10
JournalIEEE Transactions on Biomedical Engineering
Volume64
Issue number11
DOIs
StatePublished - Nov 1 2017

Fingerprint

Freezing
Detectors
Feature extraction
Sensors

Keywords

  • Anomaly score
  • feature selection
  • gait freezing

ASJC Scopus subject areas

  • Biomedical Engineering

Cite this

Pham, T. T., Moore, S. T., Lewis, S. J. G., Nguyen, D. N., Dutkiewicz, E., Fuglevand, A. J., ... Leong, P. H. W. (2017). Freezing of Gait Detection in Parkinson's Disease: A Subject-Independent Detector Using Anomaly Scores. IEEE Transactions on Biomedical Engineering, 64(11), 2719-2728. [7845616]. https://doi.org/10.1109/TBME.2017.2665438

Freezing of Gait Detection in Parkinson's Disease : A Subject-Independent Detector Using Anomaly Scores. / Pham, Thuy T.; Moore, Steven T.; Lewis, Simon John Geoffrey; Nguyen, Diep N.; Dutkiewicz, Eryk; Fuglevand, Andrew J; McEwan, Alistair L.; Leong, Philip H.W.

In: IEEE Transactions on Biomedical Engineering, Vol. 64, No. 11, 7845616, 01.11.2017, p. 2719-2728.

Research output: Contribution to journalArticle

Pham, TT, Moore, ST, Lewis, SJG, Nguyen, DN, Dutkiewicz, E, Fuglevand, AJ, McEwan, AL & Leong, PHW 2017, 'Freezing of Gait Detection in Parkinson's Disease: A Subject-Independent Detector Using Anomaly Scores', IEEE Transactions on Biomedical Engineering, vol. 64, no. 11, 7845616, pp. 2719-2728. https://doi.org/10.1109/TBME.2017.2665438
Pham, Thuy T. ; Moore, Steven T. ; Lewis, Simon John Geoffrey ; Nguyen, Diep N. ; Dutkiewicz, Eryk ; Fuglevand, Andrew J ; McEwan, Alistair L. ; Leong, Philip H.W. / Freezing of Gait Detection in Parkinson's Disease : A Subject-Independent Detector Using Anomaly Scores. In: IEEE Transactions on Biomedical Engineering. 2017 ; Vol. 64, No. 11. pp. 2719-2728.
@article{18371c89f50c4263ac6ea8d663604c13,
title = "Freezing of Gait Detection in Parkinson's Disease: A Subject-Independent Detector Using Anomaly Scores",
abstract = "Freezing of gait (FoG) is common in Parkinsonian gait and strongly relates to falls. Current clinical FoG assessments are patients' self-report diaries and experts' manual video analysis. Both are subjective and yield moderate reliability. Existing detection algorithms have been predominantly designed in subject-dependent settings. In this paper, we aim to develop an automated FoG detector for subject independent. After extracting highly relevant features, we apply anomaly detection techniques to detect FoG events. Specifically, feature selection is performed using correlation and clusterability metrics. From a list of 244 feature candidates, 36 candidates were selected using saliency and robustness criteria. We develop an anomaly score detector with adaptive thresholding to identify FoG events. Then, using accuracy metrics, we reduce the feature list to seven candidates. Our novel multichannel freezing index was the most selective across all window sizes, achieving sensitivity (specificity) of 96{\%} (79{\%}). On the other hand, freezing index from the vertical axis was the best choice for a single input, achieving sensitivity (specificity) of 94{\%} (84{\%}) for ankle and 89{\%} (94{\%}) for back sensors. Our subject-independent method is not only significantly more accurate than those previously reported, but also uses a much smaller window (e.g., 3 s versus 7.5 s) and/or lower tolerance (e.g., 0.4 s versus 2 s).",
keywords = "Anomaly score, feature selection, gait freezing",
author = "Pham, {Thuy T.} and Moore, {Steven T.} and Lewis, {Simon John Geoffrey} and Nguyen, {Diep N.} and Eryk Dutkiewicz and Fuglevand, {Andrew J} and McEwan, {Alistair L.} and Leong, {Philip H.W.}",
year = "2017",
month = "11",
day = "1",
doi = "10.1109/TBME.2017.2665438",
language = "English (US)",
volume = "64",
pages = "2719--2728",
journal = "IEEE Transactions on Biomedical Engineering",
issn = "0018-9294",
publisher = "IEEE Computer Society",
number = "11",

}

TY - JOUR

T1 - Freezing of Gait Detection in Parkinson's Disease

T2 - A Subject-Independent Detector Using Anomaly Scores

AU - Pham, Thuy T.

AU - Moore, Steven T.

AU - Lewis, Simon John Geoffrey

AU - Nguyen, Diep N.

AU - Dutkiewicz, Eryk

AU - Fuglevand, Andrew J

AU - McEwan, Alistair L.

AU - Leong, Philip H.W.

PY - 2017/11/1

Y1 - 2017/11/1

N2 - Freezing of gait (FoG) is common in Parkinsonian gait and strongly relates to falls. Current clinical FoG assessments are patients' self-report diaries and experts' manual video analysis. Both are subjective and yield moderate reliability. Existing detection algorithms have been predominantly designed in subject-dependent settings. In this paper, we aim to develop an automated FoG detector for subject independent. After extracting highly relevant features, we apply anomaly detection techniques to detect FoG events. Specifically, feature selection is performed using correlation and clusterability metrics. From a list of 244 feature candidates, 36 candidates were selected using saliency and robustness criteria. We develop an anomaly score detector with adaptive thresholding to identify FoG events. Then, using accuracy metrics, we reduce the feature list to seven candidates. Our novel multichannel freezing index was the most selective across all window sizes, achieving sensitivity (specificity) of 96% (79%). On the other hand, freezing index from the vertical axis was the best choice for a single input, achieving sensitivity (specificity) of 94% (84%) for ankle and 89% (94%) for back sensors. Our subject-independent method is not only significantly more accurate than those previously reported, but also uses a much smaller window (e.g., 3 s versus 7.5 s) and/or lower tolerance (e.g., 0.4 s versus 2 s).

AB - Freezing of gait (FoG) is common in Parkinsonian gait and strongly relates to falls. Current clinical FoG assessments are patients' self-report diaries and experts' manual video analysis. Both are subjective and yield moderate reliability. Existing detection algorithms have been predominantly designed in subject-dependent settings. In this paper, we aim to develop an automated FoG detector for subject independent. After extracting highly relevant features, we apply anomaly detection techniques to detect FoG events. Specifically, feature selection is performed using correlation and clusterability metrics. From a list of 244 feature candidates, 36 candidates were selected using saliency and robustness criteria. We develop an anomaly score detector with adaptive thresholding to identify FoG events. Then, using accuracy metrics, we reduce the feature list to seven candidates. Our novel multichannel freezing index was the most selective across all window sizes, achieving sensitivity (specificity) of 96% (79%). On the other hand, freezing index from the vertical axis was the best choice for a single input, achieving sensitivity (specificity) of 94% (84%) for ankle and 89% (94%) for back sensors. Our subject-independent method is not only significantly more accurate than those previously reported, but also uses a much smaller window (e.g., 3 s versus 7.5 s) and/or lower tolerance (e.g., 0.4 s versus 2 s).

KW - Anomaly score

KW - feature selection

KW - gait freezing

UR - http://www.scopus.com/inward/record.url?scp=85018786554&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85018786554&partnerID=8YFLogxK

U2 - 10.1109/TBME.2017.2665438

DO - 10.1109/TBME.2017.2665438

M3 - Article

C2 - 28186875

AN - SCOPUS:85018786554

VL - 64

SP - 2719

EP - 2728

JO - IEEE Transactions on Biomedical Engineering

JF - IEEE Transactions on Biomedical Engineering

SN - 0018-9294

IS - 11

M1 - 7845616

ER -