A data mining methodology for predicting early stage Parkinson's disease using non-invasive, high-dimensional gait sensor data

Conrad Tucker, Yixiang Han, Harriet Black Nembhard, Wang Chien Lee, Mechelle Lewis, Nicholas Sterling, Xuemei Huang

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

Parkinson's disease (PD) is the second most common neurological disorder after Alzheimer's disease. Key clinical features of PD are motor-related and are typically assessed by healthcare providers based on qualitative visual inspection of a patient's movement/gait/posture. More advanced diagnostic techniques such as computed tomography scans that measure brain function, can be cost prohibitive and may expose patients to radiation and other harmful effects. To mitigate these challenges, and open a pathway to remote patient-physician assessment, the authors of this work propose a data mining–driven methodology that uses low cost, non-invasive sensors to model and predict the presence (or lack therefore) of PD movement abnormalities and model clinical subtypes. The study presented here evaluates the discriminative ability of non-invasive hardware and data mining algorithms to classify PD cases and controls. A 10-fold cross-validation approach is used to compare several data mining algorithms in order to determine that which provides the most consistent results when varying the subject gait data. Next, the predictive accuracy of the data mining model is quantified by testing it against unseen data captured from a test pool of subjects. The proposed methodology demonstrates the feasibility of using non-invasive, low cost, hardware and data mining models to monitor the progression of gait features outside of the traditional healthcare facility, which may ultimately lead to earlier diagnosis of emerging neurological diseases.

Original languageEnglish (US)
Pages (from-to)238-254
Number of pages17
JournalIIE Transactions on Healthcare Systems Engineering
Volume5
Issue number4
DOIs
StatePublished - Oct 2 2015

Fingerprint

Data Mining
Gait
Parkinson Disease
Data mining
Disease
Sensors
methodology
Costs and Cost Analysis
Aptitude
hardware
Nervous System Diseases
Posture
Health Personnel
costs
Hardware
Costs
Early Diagnosis
Alzheimer Disease
early diagnosis
Tomography

All Science Journal Classification (ASJC) codes

  • Safety, Risk, Reliability and Quality
  • Safety Research
  • Public Health, Environmental and Occupational Health

Cite this

@article{048487c9b284473f8904ca8b529d625e,
title = "A data mining methodology for predicting early stage Parkinson's disease using non-invasive, high-dimensional gait sensor data",
abstract = "Parkinson's disease (PD) is the second most common neurological disorder after Alzheimer's disease. Key clinical features of PD are motor-related and are typically assessed by healthcare providers based on qualitative visual inspection of a patient's movement/gait/posture. More advanced diagnostic techniques such as computed tomography scans that measure brain function, can be cost prohibitive and may expose patients to radiation and other harmful effects. To mitigate these challenges, and open a pathway to remote patient-physician assessment, the authors of this work propose a data mining–driven methodology that uses low cost, non-invasive sensors to model and predict the presence (or lack therefore) of PD movement abnormalities and model clinical subtypes. The study presented here evaluates the discriminative ability of non-invasive hardware and data mining algorithms to classify PD cases and controls. A 10-fold cross-validation approach is used to compare several data mining algorithms in order to determine that which provides the most consistent results when varying the subject gait data. Next, the predictive accuracy of the data mining model is quantified by testing it against unseen data captured from a test pool of subjects. The proposed methodology demonstrates the feasibility of using non-invasive, low cost, hardware and data mining models to monitor the progression of gait features outside of the traditional healthcare facility, which may ultimately lead to earlier diagnosis of emerging neurological diseases.",
author = "Conrad Tucker and Yixiang Han and {Black Nembhard}, Harriet and Lee, {Wang Chien} and Mechelle Lewis and Nicholas Sterling and Xuemei Huang",
year = "2015",
month = "10",
day = "2",
doi = "10.1080/19488300.2015.1095256",
language = "English (US)",
volume = "5",
pages = "238--254",
journal = "IISE Transactions on Healthcare Systems Engineering",
issn = "2472-5579",
publisher = "Taylor and Francis Ltd.",
number = "4",

}

A data mining methodology for predicting early stage Parkinson's disease using non-invasive, high-dimensional gait sensor data. / Tucker, Conrad; Han, Yixiang; Black Nembhard, Harriet; Lee, Wang Chien; Lewis, Mechelle; Sterling, Nicholas; Huang, Xuemei.

In: IIE Transactions on Healthcare Systems Engineering, Vol. 5, No. 4, 02.10.2015, p. 238-254.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A data mining methodology for predicting early stage Parkinson's disease using non-invasive, high-dimensional gait sensor data

AU - Tucker, Conrad

AU - Han, Yixiang

AU - Black Nembhard, Harriet

AU - Lee, Wang Chien

AU - Lewis, Mechelle

AU - Sterling, Nicholas

AU - Huang, Xuemei

PY - 2015/10/2

Y1 - 2015/10/2

N2 - Parkinson's disease (PD) is the second most common neurological disorder after Alzheimer's disease. Key clinical features of PD are motor-related and are typically assessed by healthcare providers based on qualitative visual inspection of a patient's movement/gait/posture. More advanced diagnostic techniques such as computed tomography scans that measure brain function, can be cost prohibitive and may expose patients to radiation and other harmful effects. To mitigate these challenges, and open a pathway to remote patient-physician assessment, the authors of this work propose a data mining–driven methodology that uses low cost, non-invasive sensors to model and predict the presence (or lack therefore) of PD movement abnormalities and model clinical subtypes. The study presented here evaluates the discriminative ability of non-invasive hardware and data mining algorithms to classify PD cases and controls. A 10-fold cross-validation approach is used to compare several data mining algorithms in order to determine that which provides the most consistent results when varying the subject gait data. Next, the predictive accuracy of the data mining model is quantified by testing it against unseen data captured from a test pool of subjects. The proposed methodology demonstrates the feasibility of using non-invasive, low cost, hardware and data mining models to monitor the progression of gait features outside of the traditional healthcare facility, which may ultimately lead to earlier diagnosis of emerging neurological diseases.

AB - Parkinson's disease (PD) is the second most common neurological disorder after Alzheimer's disease. Key clinical features of PD are motor-related and are typically assessed by healthcare providers based on qualitative visual inspection of a patient's movement/gait/posture. More advanced diagnostic techniques such as computed tomography scans that measure brain function, can be cost prohibitive and may expose patients to radiation and other harmful effects. To mitigate these challenges, and open a pathway to remote patient-physician assessment, the authors of this work propose a data mining–driven methodology that uses low cost, non-invasive sensors to model and predict the presence (or lack therefore) of PD movement abnormalities and model clinical subtypes. The study presented here evaluates the discriminative ability of non-invasive hardware and data mining algorithms to classify PD cases and controls. A 10-fold cross-validation approach is used to compare several data mining algorithms in order to determine that which provides the most consistent results when varying the subject gait data. Next, the predictive accuracy of the data mining model is quantified by testing it against unseen data captured from a test pool of subjects. The proposed methodology demonstrates the feasibility of using non-invasive, low cost, hardware and data mining models to monitor the progression of gait features outside of the traditional healthcare facility, which may ultimately lead to earlier diagnosis of emerging neurological diseases.

UR - http://www.scopus.com/inward/record.url?scp=84981237647&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84981237647&partnerID=8YFLogxK

U2 - 10.1080/19488300.2015.1095256

DO - 10.1080/19488300.2015.1095256

M3 - Article

AN - SCOPUS:84981237647

VL - 5

SP - 238

EP - 254

JO - IISE Transactions on Healthcare Systems Engineering

JF - IISE Transactions on Healthcare Systems Engineering

SN - 2472-5579

IS - 4

ER -