An artificial neural network approach to detect presence and severity of Parkinson’s disease via gait parameters

Abstract

Introduction Gait deficits are debilitating in people with Parkinson’s disease (PwPD), which inevitably deteriorate over time. Gait analysis is a valuable method to assess disease-specific gait patterns and their relationship with the clinical features and progression of the disease. Objectives Our study aimed to i) develop an automated diagnostic algorithm based on machine-learning techniques (artificial neural networks [ANNs]) to classify the gait deficits of PwPD according to disease progression in the Hoehn and Yahr (H-Y) staging system, and ii) identify a minimum set of gait classifiers. Methods We evaluated 76 PwPD (H-Y stage 1–4) and 67 healthy controls (HCs) by computerized gait analysis. We computed the time-distance parameters and the ranges of angular motion (RoMs) of the hip, knee, ankle, trunk, and pelvis. Principal component analysis was used to define a subset of features including all gait variables. An ANN approach was used to identify gait deficits according to the H-Y stage. Results We identified a combination of a small number of features that distinguished PwPDs from HCs (one combination of two features: knee and trunk rotation RoMs) and identified the gait patterns between different H-Y stages (two combinations of four features: walking speed and hip, knee, and ankle RoMs; walking speed and hip, knee, and trunk rotation RoMs). Conclusion The ANN approach enabled automated diagnosis of gait deficits in several symptomatic stages of Parkinson’s disease. These results will inspire future studies to test the utility of gait classifiers for the evaluation of treatments that could modify disease progression.

Current Perspectives on the Assessment and Management of Gait Disorders in Parkinson’s Disease

“…Machine-learning techniques have been assessed to assist in processing this type of data, and have the ability to differentiate controls from PD cases, as well as different progressive stages of the condition. 104 AUC values of 0.76-0.90 using wearable devices have been obtained, 105 with particular correlation at earlier PD stages, further reinforcing the potential of gait assessment as an objective biomarker. Advanced gait-assessment techniques processed in this way are also able to predict in which patients FoG may arise based on gait parameters, allowing therapies to be targeted early.…”

Ability of a Set of Trunk Inertial Indexes of Gait to Identify Gait Instability and Recurrent Fallers in Parkinson’s Disease

“…In particular, the correlation coefficients revealed that HR AP was lower in subjects with lower pelvic obliquity and pelvic rotation. Alterations in trunk rotation and pelvic kinematics have been consistently described as characteristics of pwPD [ 55 , 88 , 89 ]. Based on our results, we can argue that the trunk rotation rigidity reflected by pelvic rotation and pelvic obliquity leads to alterations in AP trunk smoothness during the gait.…”

Machine Learning Approach to Support the Detection of Parkinson’s Disease in IMU-Based Gait Analysis

“…However, there is no single optimal classification tool; rather, the best algorithm performance is defined by the studied features [ 32 , 33 ]. In this way, RF appears to be the most robust in the case of a significant reduction in data [ 31 ], ANN is considered an adaptable algorithm with the ability to address nonlinear data [ 34 , 35 ], requiring a large number of parameters for a correct generalization [ 31 ], and DT and KNN benefit from dividing the problem of context recognition into smaller subproblems, which are approached one by one intuitively [ 32 , 36 ]. When assessing the gait of pwPD using IMUs, reducing the number of wearable devices to a single lumbar-mounted sensor allows for sufficient identification of gait abnormalities without significant information loss, reducing the wearability burden in nonlaboratory conditions [ 5 , 37 , 38 ].…”

“…Ideally, supervised ML algorithms should be trained on a set represented by a number of subjects that is approximately 20 times the number of features input into the model [ 47 , 49 ]. When this ratio of features to sample size is respected in studies on gait data and supervised ML, the accuracies of the retrieved models are accurate but less than 90%, reflecting the partial clinical relevance of gait parameters on the multifactorial etiology of pathologies leading to gait disorders [ 10 , 34 , 50 ]. Therefore, particularly for small samples, reducing the number of features in the model may save computational expenditure, reduce the required size of the training set to avoid overfitting [ 47 ], and simplify the medical interpretation of the classifier.…”