Towards Real-Time Prediction of Freezing of Gait in Patients With Parkinson's Disease: Addressing the Class Imbalance Problem

Abstract

Freezing of gait (FoG) is a common motor symptom in patients with Parkinson's disease (PD). FoG impairs gait initiation and walking and increases fall risk. Intelligent external cueing systems implementing FoG detection algorithms have been developed to help patients recover gait after freezing. However, predicting FoG before its occurrence enables preemptive cueing and may prevent FoG. Such prediction remains challenging given the relative infrequency of freezing compared to non-freezing events. In this study, we investigated the ability of individual and ensemble classifiers to predict FoG. We also studied the effect of the ADAptive SYNthetic (ADASYN) sampling algorithm and classification cost on classifier performance. Eighteen PD patients performed a series of daily walking tasks wearing accelerometers on their ankles, with nine experiencing FoG. The ensemble classifier formed by Support Vector Machines, K-Nearest Neighbors, and Multi-Layer Perceptron using bagging techniques demonstrated highest performance (F1 = 90.7) when synthetic FoG samples were added to the training set and class cost was set as twice that of normal gait. The model identified 97.4% of the events, with 66.7% being predicted. This study demonstrates our algorithm's potential for accurate prediction of gait events and the provision of preventive cueing in spite of limited event frequency.

Keywords: ADASYN; Parkinson’s disease; cost of classification; data synthesis; ensemble classifier; freezing of gait; wearable sensors.

Conflict of interest statement

The authors declare no conflict of interest. The funding agency had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Figures

Figure 1

Experiment layout. Number of obstacles in the object area varied between 0, 1 and 2. The width of walking path in the object area (w) varied between 150% and 100% of shoulder width of participants.

Figure 2

The process of creating samples from acceleration signals. (a) Extracting features from six successive windows at time t = T (left) and the next time step, t = T+1 (right). Red highlighted area shows FoG labeled period using recorded videos, green boxes show length of windows used to extract features from acceleration signal; (b) Combining arrays of features from different combinations of sensor-axis.

Figure 3

Performance measures of classifiers for patient-dependent models using synthetic data and cost of misclassification.

Figure 4

Performance measures of classifiers for patient-independent models using synthetic data and cost of misclassification.

Figure 5

Acceleration signal collected from the right ankle sensor and the corresponding labeled and detected events using of ClsfBagging with β=0.2 and CFoG=2.

Figure 6

Average FoG detection latency of ClsfBagging in patient-dependent models with β=0.2 and CFoG=2. Negative values of time represent duration before FoG onset.