The increasing availability of large datasets obtained through quantitative human motion analysis is rapidly opening new research pathways in human gait, biomechanics and motor control research.
For example, during gait, the locomotor system relies on higher order cognitive control involving attention, planning, memory and other motor, perceptual and cognitive processes. Emerging evidence shows that gait analysis can independently predict cognitive decline and other adverse outcomes such as disability, cardiovascular disease and survival. Although instrumented gait analysis provides a substantial amount of data, gait phenotype classification is still largely based on clinicians' subjective judgment. Moreover, gait and movement analysis data is multidimensional and highly nonlinear, and is particularly suitable to be handled with data-driven techniques in order to generate predictive and classification models.
For this reason, the use of machine learning in medicine has grown enormously in the last decade: artificial neural networks were used in the field of posture research for prediction, diagnosis and prognosis. Support vector machines, k-nearest neighbor classifiers, hidden Markov models and naive-Bayes classifier are other methods for gait classification. These techniques were often combined with data reduction procedures for feature extraction. Among them, principal component analysis enabled decomposition of the kinematics of postural movements into their fundamental components, offering an alternative way to investigate motor patterns and the associated motor control activity.
This Research Topic covers state-of-the-art applications of supervised and unsupervised techniques to analyze human movement for diagnosis, disease classification and pattern recognition. Motion data could belong both to the kinematics and kinetics domains, either from standard movement analysis laboratories (optical motion capture, ground reaction forces), or from wearable devices (inertial units). Machine learning applications to electromyographic signals will also be included.
Work has just begun to unlock the potential of applying data science methods to answer clinical and biomechanical questions. A data-driven approach applied to movement biomechanics will not only increase our understanding of neurological and cognitive-motor interactions but, most importantly, may be used to aid early diagnosis, prognosis and the development of new interventions.
The increasing availability of large datasets obtained through quantitative human motion analysis is rapidly opening new research pathways in human gait, biomechanics and motor control research.
For example, during gait, the locomotor system relies on higher order cognitive control involving attention, planning, memory and other motor, perceptual and cognitive processes. Emerging evidence shows that gait analysis can independently predict cognitive decline and other adverse outcomes such as disability, cardiovascular disease and survival. Although instrumented gait analysis provides a substantial amount of data, gait phenotype classification is still largely based on clinicians' subjective judgment. Moreover, gait and movement analysis data is multidimensional and highly nonlinear, and is particularly suitable to be handled with data-driven techniques in order to generate predictive and classification models.
For this reason, the use of machine learning in medicine has grown enormously in the last decade: artificial neural networks were used in the field of posture research for prediction, diagnosis and prognosis. Support vector machines, k-nearest neighbor classifiers, hidden Markov models and naive-Bayes classifier are other methods for gait classification. These techniques were often combined with data reduction procedures for feature extraction. Among them, principal component analysis enabled decomposition of the kinematics of postural movements into their fundamental components, offering an alternative way to investigate motor patterns and the associated motor control activity.
This Research Topic covers state-of-the-art applications of supervised and unsupervised techniques to analyze human movement for diagnosis, disease classification and pattern recognition. Motion data could belong both to the kinematics and kinetics domains, either from standard movement analysis laboratories (optical motion capture, ground reaction forces), or from wearable devices (inertial units). Machine learning applications to electromyographic signals will also be included.
Work has just begun to unlock the potential of applying data science methods to answer clinical and biomechanical questions. A data-driven approach applied to movement biomechanics will not only increase our understanding of neurological and cognitive-motor interactions but, most importantly, may be used to aid early diagnosis, prognosis and the development of new interventions.