Interpretable machine learning models for classifying low back pain status using functional physiological variables

Bernard X W Liew,Deborah Falla,David Rugamer,Alessandro Marco De Nunzio

doi:10.1007/s00586-020-06356-0

Bernard X W Liew, Deborah Falla + Show 2 more

Open Access

https://doi.org/10.1007/s00586-020-06356-0

Copy DOI

Abstract

PurposeTo evaluate the predictive performance of statistical models which distinguishes different low back pain (LBP) sub-types and healthy controls, using as input predictors the time-varying signals of electromyographic and kinematic variables, collected during low-load lifting.MethodsMotion capture with electromyography (EMG) assessment was performed on 49 participants [healthy control (con) = 16, remission LBP (rmLBP) = 16, current LBP (LBP) = 17], whilst performing a low-load lifting task, to extract a total of 40 predictors (kinematic and electromyographic variables). Three statistical models were developed using functional data boosting (FDboost), for binary classification of LBP statuses (model 1: con vs. LBP; model 2: con vs. rmLBP; model 3: rmLBP vs. LBP). After removing collinear predictors (i.e. a correlation of > 0.7 with other predictors) and inclusion of the covariate sex, 31 predictors were included for fitting model 1, 31 predictors for model 2, and 32 predictors for model 3.ResultsSeven EMG predictors were selected in model 1 (area under the receiver operator curve [AUC] of 90.4%), nine predictors in model 2 (AUC of 91.2%), and seven predictors in model 3 (AUC of 96.7%). The most influential predictor was the biceps femoris muscle (peak beta = 0.047) in model 1, the deltoid muscle (peak beta = 0.052) in model 2, and the iliocostalis muscle (peak beta = 0.16) in model 3.ConclusionThe ability to transform time-varying physiological differences into clinical differences could be used in future prospective prognostic research to identify the dominant movement impairments that drive the increased risk.Graphic abstractThese slides can be retrieved under Electronic Supplementary Material.

Highlights

Low back pain (LBP) has a global prevalence of close to half a billion [1] and ranks as the number one cause of years lived with disability [1]
The excellent predictive accuracy of current models developed using a spectrum of biological signals and statistical techniques has demonstrated the potential for such methods to assist clinical decision-making [3,4,5,6,7,8]
The primary purpose of the present study is to develop and determine the predictive performance of statistical models to distinguish different low back pain (LBP) sub-types and healthy controls from each other, using whole-body electromyographic and kinematic variables as predictors collected during a functional lifting task

Summary

Introduction

Low back pain (LBP) has a global prevalence of close to half a billion [1] and ranks as the number one cause of years lived with disability [1]. Researchers are turning towards advance statistical learning techniques to develop accurate prediction models for people with LBP using information from high-dimensional, multivariate biological signals [3]. Existing studies have used biological signals such as surface electromyography (sEMG) [4, 5], kinematics [6, 7], brain neuroimaging [3], and spine neuroimages [8] as candidate predictors; feeding into statistical learning techniques such as support vector machine (SVM) [3,4,5,6], neural networks [7], and natural language processing [8]. The excellent predictive accuracy of current models developed using a spectrum of biological signals and statistical techniques has demonstrated the potential for such methods to assist clinical decision-making [3,4,5,6,7,8]

Objectives

Results

Discussion

Conclusion