Tracking Foot Drop Recovery Following Lumbar-Spine Surgery, Applying Multiclass Gait Classification Using Machine Learning Techniques

Shiva Sharif Bidabadi, Tele Tan, Iain Murray, Gabriel Lee

Research output: Contribution to journalArticle

Abstract

The ability to accurately perform human gait evaluation is critical for orthopedic foot and ankle surgeons in tracking the recovery process of their patients. The assessment of gait in an objective and accurate manner can lead to improvement in diagnoses, treatments, and recovery. Currently, visual inspection is the most common clinical method for evaluating the gait, but this method can be subjective and inaccurate. The aim of this study is to evaluate the foot drop condition in an accurate and clinically applicable manner. The gait data were collected from 56 patients suffering from foot drop with L5 origin gathered via a system based on inertial measurement unit sensors at different stages of surgical treatment. Various machine learning (ML) algorithms were applied to categorize the data into specific groups associated with the recovery stages. The results revealed that the random forest algorithm performed best out of the selected ML algorithms, with an overall 84.89% classification accuracy and 0.3785 mean absolute error for regression.

Original languageEnglish
Article number2542
JournalSensors (Basel, Switzerland)
Volume19
Issue number11
DOIs
Publication statusPublished - 4 Jun 2019

Fingerprint

gait
machine learning
spine
Gait
surgery
Surgery
Learning systems
Foot
Spine
recovery
Recovery
Learning algorithms
Units of measurement
Orthopedics
surgeons
orthopedics
Inspection
Ankle
regression analysis
inspection

Cite this

@article{10efba543c4e4216972990156b377d67,
title = "Tracking Foot Drop Recovery Following Lumbar-Spine Surgery, Applying Multiclass Gait Classification Using Machine Learning Techniques",
abstract = "The ability to accurately perform human gait evaluation is critical for orthopedic foot and ankle surgeons in tracking the recovery process of their patients. The assessment of gait in an objective and accurate manner can lead to improvement in diagnoses, treatments, and recovery. Currently, visual inspection is the most common clinical method for evaluating the gait, but this method can be subjective and inaccurate. The aim of this study is to evaluate the foot drop condition in an accurate and clinically applicable manner. The gait data were collected from 56 patients suffering from foot drop with L5 origin gathered via a system based on inertial measurement unit sensors at different stages of surgical treatment. Various machine learning (ML) algorithms were applied to categorize the data into specific groups associated with the recovery stages. The results revealed that the random forest algorithm performed best out of the selected ML algorithms, with an overall 84.89{\%} classification accuracy and 0.3785 mean absolute error for regression.",
keywords = "foot drop, gait classification, inertial measurement unit, machine learning",
author = "{Sharif Bidabadi}, Shiva and Tele Tan and Iain Murray and Gabriel Lee",
year = "2019",
month = "6",
day = "4",
doi = "10.3390/s19112542",
language = "English",
volume = "19",
journal = "Sensors (Switzerland)",
issn = "1424-8220",
publisher = "MDPI AG",
number = "11",

}

Tracking Foot Drop Recovery Following Lumbar-Spine Surgery, Applying Multiclass Gait Classification Using Machine Learning Techniques. / Sharif Bidabadi, Shiva; Tan, Tele; Murray, Iain; Lee, Gabriel.

In: Sensors (Basel, Switzerland), Vol. 19, No. 11, 2542, 04.06.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Tracking Foot Drop Recovery Following Lumbar-Spine Surgery, Applying Multiclass Gait Classification Using Machine Learning Techniques

AU - Sharif Bidabadi, Shiva

AU - Tan, Tele

AU - Murray, Iain

AU - Lee, Gabriel

PY - 2019/6/4

Y1 - 2019/6/4

N2 - The ability to accurately perform human gait evaluation is critical for orthopedic foot and ankle surgeons in tracking the recovery process of their patients. The assessment of gait in an objective and accurate manner can lead to improvement in diagnoses, treatments, and recovery. Currently, visual inspection is the most common clinical method for evaluating the gait, but this method can be subjective and inaccurate. The aim of this study is to evaluate the foot drop condition in an accurate and clinically applicable manner. The gait data were collected from 56 patients suffering from foot drop with L5 origin gathered via a system based on inertial measurement unit sensors at different stages of surgical treatment. Various machine learning (ML) algorithms were applied to categorize the data into specific groups associated with the recovery stages. The results revealed that the random forest algorithm performed best out of the selected ML algorithms, with an overall 84.89% classification accuracy and 0.3785 mean absolute error for regression.

AB - The ability to accurately perform human gait evaluation is critical for orthopedic foot and ankle surgeons in tracking the recovery process of their patients. The assessment of gait in an objective and accurate manner can lead to improvement in diagnoses, treatments, and recovery. Currently, visual inspection is the most common clinical method for evaluating the gait, but this method can be subjective and inaccurate. The aim of this study is to evaluate the foot drop condition in an accurate and clinically applicable manner. The gait data were collected from 56 patients suffering from foot drop with L5 origin gathered via a system based on inertial measurement unit sensors at different stages of surgical treatment. Various machine learning (ML) algorithms were applied to categorize the data into specific groups associated with the recovery stages. The results revealed that the random forest algorithm performed best out of the selected ML algorithms, with an overall 84.89% classification accuracy and 0.3785 mean absolute error for regression.

KW - foot drop

KW - gait classification

KW - inertial measurement unit

KW - machine learning

UR - http://www.scopus.com/inward/record.url?scp=85067489360&partnerID=8YFLogxK

U2 - 10.3390/s19112542

DO - 10.3390/s19112542

M3 - Article

VL - 19

JO - Sensors (Switzerland)

JF - Sensors (Switzerland)

SN - 1424-8220

IS - 11

M1 - 2542

ER -