Classification of foot drop gait characteristic due to lumbar radiculopathy using machine learning algorithms

Area of Science:

• Biomedical Engineering
• Neurology
• Data Science

Background:

• Gait pattern analysis is crucial for identifying and managing neurological disorders.
• Current visual inspection methods for gait disorders are subjective and inaccurate.
• Foot drop, often caused by L5 radiculopathy, significantly impacts walking gait.

Purpose of the Study:

• To investigate the feasibility of using inertial measurement unit (IMU) sensors and supervised learning to differentiate foot drop gait from normal gait.
• To establish an objective, data-driven method for diagnosing foot drop.

Main Methods:

• Collected gait data from 56 adults with L5 radiculopathy-induced foot drop and 30 healthy adults.
• Utilized commercial off-the-shelf IMU sensors during walking trials.
• Applied machine learning algorithms, including Random Forest, SVM, and Naive Bayes, for classification.

Main Results:

• Initial classification accuracies reached up to 88.45% (Random Forest).
• Feature selection improved performance, with Random Forest achieving 93.18% accuracy.
• Demonstrated high feasibility in distinguishing foot drop gait from normal gait.

Conclusions:

• The combination of IMU sensors and machine learning provides a feasible and promising objective method for identifying foot drop gait.
• This approach can aid clinical decision-making and improve diagnosis and prognosis assessment for gait disorders.
• Highlights the potential of wearable sensor technology in neurological diagnostics.