Crouch gait patterns defined using k-means cluster analysis are related to underlying clinical pathology

Adam Rozumalski, Michael H. Schwartz

Research output: Contribution to journalArticlepeer-review

43 Scopus citations

Abstract

In this study a gait classification method was developed and applied to subjects with Cerebral palsy who walk with excessive knee flexion at initial contact. Sagittal plane gait data, simplified using the gait features method, is used as input into a k-means cluster analysis to determine homogeneous groups. Several clinical domains were explored to determine if the clusters are related to underlying pathology. These domains included age, joint range-of-motion, strength, selective motor control, and spasticity. Principal component analysis is used to determine one overall score for each of the multi-joint domains (strength, selective motor control, and spasticity). The current study shows that there are five clusters among children with excessive knee flexion at initial contact. These clusters were labeled, in order of increasing gait pathology: (1) mild crouch with mild equinus, (2) moderate crouch, (3) moderate crouch with anterior pelvic tilt, (4) moderate crouch with equinus, and (5) severe crouch. Further analysis showed that age, range-of-motion, strength, selective motor control, and spasticity were significantly different between the clusters (p < 0.001). The general tendency was for the clinical domains to worsen as gait pathology increased. This new classification tool can be used to define homogeneous groups of subjects in crouch gait, which can help guide treatment decisions and outcomes assessment.

Original languageEnglish (US)
Pages (from-to)155-160
Number of pages6
JournalGait and Posture
Volume30
Issue number2
DOIs
StatePublished - Aug 1 2009

Keywords

  • Cerebral palsy
  • Crouch gait
  • Principal component analysis
  • Singular value decomposition
  • k-means Cluster analysis

Fingerprint Dive into the research topics of 'Crouch gait patterns defined using k-means cluster analysis are related to underlying clinical pathology'. Together they form a unique fingerprint.

Cite this