On Discovering the Correlated Relationship between Static and Dynamic Data in Clinical Gait Analysis

‘Gait’ is a person’s manner of walking. Patients may have an abnormal gait due to a range of physical impairment or brain damage. Clinical gait analysis (CGA) is a technique for identifying the underlying impairments that affect a patient’s gait pattern. The CGA is critical for treatment planning. Essentially, CGA tries to use patients’ physical examination results, known as

static

data, to interpret the dynamic characteristics in an abnormal gait, known as

dynamic

data. This process is carried out by gait analysis experts, mainly based on their experience which may lead to subjective diagnoses. To facilitate the automation of this process and form a relatively objective diagnosis, this paper proposes a new probabilistic correlated static-dynamic model (CSDM) to discover correlated relationships between the dynamic characteristics of gait and their root cause in the static data space. We propose an EM-based algorithm to learn the parameters of the CSDM. One of the main advantages of the CSDM is its ability to provide intuitive knowledge. For example, the CSDM can describe what kinds of static data will lead to what kinds of hidden gait patterns in the form of a decision tree, which helps us to infer dynamic characteristics based on static data. Our initial experiments indicate that the CSDM is promising for discovering the correlated relationship between physical examination (static) and gait (dynamic) data.