Effect of sensor number and location on accelerometry-based vertical ground reaction force estimation during walking

• DOI: 10.1371/journal.pdig.0000343
• Publication Date: 2024-05-14T17:23:16Z
• Authors: Ricky Pimentel; Cortney Armitano-...; Ryan MacPherson; Anoop Sathyan; Jack Twiddy; Kaila Peterson; Michael Daniele; Adam W. Kiefer; Edgar Lobaton; Brian Pietrosimone; Jason R. Franz
• Keywords: Ground reaction force; Biomechanics; Treadmill

Knee osteoarthritis is a major cause of global disability and cost for the healthcare system. Lower extremity loading determinant knee onset progression; however, technology that assists rehabilitative clinicians in optimizing key metrics lower significantly limited. The peak vertical component ground reaction force (vGRF) first 50% stance highly associated with biological patient-reported outcomes linked to symptoms. Monitoring maintaining typical vGRF profiles may support healthy gait biomechanics joint tissue prevent progression osteoarthritis. Yet, optimal number sensors sensor placements predicting accurate from accelerometry remains unknown. Our goals were to: 1) determine how many what locations yielded most estimates during walking; 2) characterize prescribing different conditions affected estimates. We asked 20 young adult participants wear 5 accelerometers on their waist, shanks, feet walk force-instrumented treadmill control targeted biofeedback prompting 5% underloading overloading vGRFs. trained tested machine learning models estimate various accelerometer inputs identified which combinations accurate. found neural network using one at waist walking, average errors 4.4% body weight. waist-only configuration was able distinguish between prescribed biofeedback, matching measured outcomes. Including foot or shank acceleration signals model reduced accuracy, particularly condition. results suggest system designed monitor changes walking deploy only need single located participants.