Diplegic forms of Cerebral Palsy: validation of Ferrari classification via artificial neural network

What is it about?

Gait classifications of diplegic forms of Cerebral Palsy (dCP) have been extensively proposed in literature to assist in diagnosis and prognosis formulation. However, qualitative classification strategies so far conceived by clinical experts, lack of statistical validity and ability to infer functional outcomes. Instead, quantitative approaches have been generating groups not clinically interpretable. Conversely Ferrari et al., in order to improve prognosis formulation, have proposed to analyse top-down components of motor organization distinguishing four forms corresponding to four distinct kinematics patterns of gait. Aim of this study is to provide a contribution to validation of this classification by demonstrating through a supervised artificial neural network that the four clinical forms are recognizable also on a quantitative basis in the space of gait analysis data. On average 5 gait analysis trials were acquired 125 children with dCP walking indistinctively either barefoot, with shoes and/or canes, pre- or post- surgery or botulinum injections.

Featured Image

Why is it important?

Classification systems of diplegic forms of Cerebral Palsy children so far proposed have been based either on qualitative approaches, lacking statistical evidence and inferences to functional outcomes, or quantitative, lacking meaningful value for prognosis formulation and clinical decision-making. Ferrari et al. proposed a qualitative classification system differentiating dCP children in four forms. This study provided a contribution to the validation of this classification by training a supervised neural network to recognize the four forms from gait analysis data and demonstrate how the four clinical forms proposed by Ferrari et al. are recognizable also on a quantitative basis in the space of gait analysis data