Anterior Cruciate Ligament (ACL) injuries are a common threat during sporting tasks and their specific at-risk movement patterns have been assessed using dynamic tests such as the Vertical Drop Jump (VDJ), whose kinematic and kinetic variables have shown a good predictive ability to prospectively screen athletes with high-risk landing behaviours. Marker-based 3D motion capture systems in a controlled laboratory settings are the gold standard to evaluate kinematics and kinetics during the VDJ however, in order to overcome the downsides of such methods, the use of IMU (Inertial Measurements Unit) sensors together with machine learning-based algorithms is promising. The aim of this study is to develop and validate an Artificial Neural Network (ANN) allowing for the estimation of 3D moments at right knee joint and the right Ground Reaction Forces (GRFs) during the first landing phase of the VDJ by using data recorded from three IMUs, placed on the sacrum and on the right thigh and shank respectively. Due to the promising results, the developed methods represents a starting point for further improvements leading to a novel methods for in-field diagnosis in sports with respect to ACL risk of injury.
Le lesioni del legamento crociato anteriore (LCA) sono un rischio comune durante le attività sportive e i loro specifici pattern sono stati valutati utilizzando test dinamici come il Vertical Drop Jump (VDJ), le cui variabili cinematiche e cinetiche hanno mostrato una buona capacità predittiva per l’individuazione degli atleti ad alto rischio di infortunio durante l’atterraggio. Il gold standard per la valutazione della cinematica e della cinetica del VDJ sono i sistemi 3D di cattura del movimento impiegati in un ambiente di laboratorio controllato, tuttavia, al fine di superare gli aspetti negativi di tali metodi, l'utilizzo congiunto di sensori IMU (Inertial Measurement Unit) e algoritmi di machine learning è molto promettente. Lo scopo di questo studio è quello di sviluppare e validare una rete neurale artificiale (Artificial Neural Network – ANN) che consenta la stima dei momenti 3D all'articolazione del ginocchio destro e delle forze di reazione del suolo (Ground Reaction Forces - GRFs) durante la prima fase di atterraggio del VDJ utilizzando i dati registrati da tre IMU, rispettivamente posizionati sulla coscia e sulla gamba destra e sul sacrum. In seguito ai promettenti risultati ottenuti, il presente modello rappresenta un punto di partenza per ulteriori miglioramenti per lo sviluppo di nuovi metodo per la diagnosi sul campo per lo screening degli atleti rispetto al rischio di infortunio al LCA.
Machine learning-based estimation of ground reaction forces and knee joint kinetics from inertial sensors while performing a vertical drop jump
Cerfoglio, Serena
2020/2021
Abstract
Anterior Cruciate Ligament (ACL) injuries are a common threat during sporting tasks and their specific at-risk movement patterns have been assessed using dynamic tests such as the Vertical Drop Jump (VDJ), whose kinematic and kinetic variables have shown a good predictive ability to prospectively screen athletes with high-risk landing behaviours. Marker-based 3D motion capture systems in a controlled laboratory settings are the gold standard to evaluate kinematics and kinetics during the VDJ however, in order to overcome the downsides of such methods, the use of IMU (Inertial Measurements Unit) sensors together with machine learning-based algorithms is promising. The aim of this study is to develop and validate an Artificial Neural Network (ANN) allowing for the estimation of 3D moments at right knee joint and the right Ground Reaction Forces (GRFs) during the first landing phase of the VDJ by using data recorded from three IMUs, placed on the sacrum and on the right thigh and shank respectively. Due to the promising results, the developed methods represents a starting point for further improvements leading to a novel methods for in-field diagnosis in sports with respect to ACL risk of injury.File | Dimensione | Formato | |
---|---|---|---|
2021_07_Cerfoglio.pdf
solo utenti autorizzati dal 03/07/2022
Dimensione
5.4 MB
Formato
Adobe PDF
|
5.4 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/177314