Exploring the Application of Pattern Recognition and Machine Learning for Identifying Movement Phenotypes During Deep Squat and Hurdle Step Movements

Abstract

Background: Movement screens are increasingly used in sport and rehabilitation to evaluate movement competency. However, common screens are often evaluated using subjective visual detection of a priori prescribed discrete movement features (e.g., spine angle at maximum squat depth) and may not account for whole-body movement coordination, or associations between different discrete features.

Objective: To apply pattern recognition and machine learning techniques to identify whole-body movement pattern phenotypes during the performance of exemplar functional movement screening tasks; the deep squat and hurdle step. Additionally, we also aimed to compare how discrete kinematic measures, commonly used to score movement competency, differed between emergent groups identified via pattern recognition and machine learning.

Methods: Principal component analysis (PCA) was applied to 3-dimensional (3D) trajectory data from participant's deep squat (DS) and hurdle step performance, identifying emerging features that describe orthogonal modes of inter-trial variance in the data. A gaussian mixture model (GMM) was fit and used to cluster the principal component scores as an unsupervised machine learning approach to identify emergent movement phenotypes. Between group features were analyzed using a one-way ANOVA to determine if the objective classifications were significantly different from one another.

Results: Three clusters (i.e., phenotypes) emerged for the DS and right hurdle step (RHS) and 4 phenotypes emerged for the left hurdle step (LHS). Selected discrete points commonly used to score DS and hurdle step movements were different between emergent groups. In regard to the select discrete kinematic measures, 4 out of 5, 7 out of 7 and 4 out of 7, demonstrated a main effect (p < 0.05) between phenotypes for the DS, RHS, and LHS respectively.

Conclusion: Findings support that whole-body movement analysis, pattern recognition and machine learning techniques can objectively identify movement behavior phenotypes without the need to a priori prescribe movement features. However, we also highlight important considerations that can influence outcomes when using machine learning for this purpose.

Keywords: cluster; functional movement screen; gaussian mixture model; movement phenotypes; principal component analysis.

FIGURE 1

Whole-body reflective marker set up. Marker clusters were placed bilaterally on the shank, thigh, forearm, and upper arm as well as on the pelvis and trunk. Anatomical markers were placed on the anterior and posterior lateral aspects of the head, suprasternal notch, xiphoid process, 7^th cervical vertebra; and bilaterally on the acromion, lateral and medial epicondyles, radial and ulnar styloid processes, 2nd and 5th metacarpals, dorsum of the hand, iliac crest, anterior superior iliac spine, greater trochanter of femur, lateral and medial condyles, lateral and medial malleoli, 1st and 5th metatarsal, dorsal tarsal midline, and calcaneus.

FIGURE 2

Deep squat movement and hurdle step movement as adapted from Cook et al. (2006a).

FIGURE 3

(A) Represents the phases of the DS movement from 0 to 100%. The “move into” phase is contained by the red and black bars and in between the black and blue bars is the “move out” phase. The below graph represents tracking of the suprasternal notch marker to determine 50% of the movement (maximum squat depth) as well as define “start” and “end” points of the movement. (B) Represents the phases of the hurdle step movement from 0 to 100%. The “move into” phase is contained by the red and black bars and in between the black and blue bars is the “move out” phase. The graph below represents tracking of the right (LHS) or left (RHS) calcaneus marker to determine 50% of the movement (end of heel touch) as well as define “start” and “end” points of the RHS and LHS, respectively.

FIGURE 4

BIC values for k = 1–10 for the DS, RHS, and LHS demonstrating minimum values at k = 3 for the DS and RHS, and k = 4 for the LHS.

FIGURE 5

Reconstructed movement phenotypes using the centroid PC scores from each cluster considering the deep squat movement. Black, movement phenotype 1; red, movement phenotype 2; blue, movement phenotype 3.

FIGURE 6

Reconstructed movement phenotypes using the centroid PC scores from each cluster considering the right hurdle step movement phenotypes identified. Black, movement phenotype 1; red, movement phenotype 2; blue, movement phenotype 3.

FIGURE 7

Reconstructed movement phenotypes using the centroid PC scores from each cluster considering the left hurdle step movement phenotypes identified. Black, movement phenotype 1; red, movement phenotype 2; blue, movement phenotype 3; gray, movement phenotype 4.

FIGURE 8

Violin plot (Holger Hoffmann, 2020) demonstrating the distribution shape of each phenotype for kinematic measures commonly used to score the DS. The mean is represented by the white dotted line and median with the solid white line. (A) The femur is at or below horizontal; (B) the torso remains upright and/or is parallel with the tibia; (C) the dowel remains aligned over the feet; (D) the knees remain aligned with the feet. *The mean difference is significant at the 0.05 level.

FIGURE 9

Violin plot (Holger Hoffmann, 2020) demonstrating the distribution shape of each phenotype for kinematic measures commonly used to score the RHS. The mean is represented by the white dotted line and median with the solid white line. (A) Hip, knee, and ankle remain aligned; (B) there is little to no movement in the lumbar spine; (C) the hands/dowel remains parallel to the string. ^∗The mean difference is significant at the 0.05 level.

FIGURE 10

Violin plot (Holger Hoffmann, 2020) demonstrating the distribution shape of each phenotype for kinematic measures commonly used to score the LHS. The mean is represented by the white dotted line and median with the solid white line. (A) Hip, knee, and ankle remain aligned; (B) there is little to no movement in the lumbar spine. (C) The hands/dowel remains parallel to the string. ^∗The mean difference is significant at the 0.05 level.