Data

A database is provided for the development of the gait event detection algorithm. The final evaluation will be done on a different database with a similar population. All the data are provided in a c3d format which can be easily managed with btk or ezc3d. Both of these toolbox are compatible with Python and Matlab.

In addition to the data, the methods proposed by Zeni and Higginson (2009) has been coded and shared as an example.

Population :

A set of pathological gait data containing marker trajectories and gold-standard gait events identified with forceplates has been made available.

Results are presented as mean (standard deviation)
  CP GMFS I CP GMFS2 CP GMFS III Idiopathic Toe-walker Foot Deformity
N 20 20 5 20 25
Age (years) 12.4 (4.0) 15.9 (8.5) 11.4 (4.0) 9.3 (3.0) 19.2 (13.3)
Gait profile Score 6.7 (1.7) 9.9 (3.2) 10.4 (1.7) 6.9 (2.1) 5.1 (1.2)

 Subjects were asked to walk bare-foot at their self-selected speed without external help on a 10m walkway.

Marker set :

The conventional gait model marker set has been chosen as it is one of the mostly used.

  • Trunk :
    • C7 : Spinal process of the 7th cervical vertebra
    • T10 : Spinal process of the 10th thoracic vertebra
    • STRN : Processus Xyphoideus
    • CLAV : Incisura Jugularis
  • Pelvis :
    • R/LASI : Right/Left Anterior Superior Iliac Spines
    • R/LPSI : Right/Left Posterior Superior Iliac SPines
  • R/L Thigh :
    • R/LTHI : Right/Left Mid-Thigh Wand
    • R/LKNE : Righ/Left Lateral Femoral Epicondyle
  • R/L Shank :
    • R/LTIB : Right/Left Mid-Tibia Wand
    • R/LANK : Right/Left Lateral Malleolus
  • R/L Foot :
    • R/LHEE : Right/Left Posterior Calcaneus
    • R/LTOE : Right/Left 2nd metatarso-cuneiform joint

Event detection with force plateform

All the event time were added in the c3d file using the following algorithm based on the raw ground Reaction Forces (GRF)

Algorithm :

    1.  
      Vertical ground reaction force.
      4th order Low-Pass Butterworth (10Hz) on raw vertical Ground Reaction Force (GRF)
    2. Identify the maximum of GRF
    3. Set threshold (10N)
    4. Identify events :
      • Foot Strike is the first frame before the max GRF that is below the threshold
      • Foot off is the first frame after GRFmax that is below the threshold
    5. Side detection :
      • Computation of the midpoint of the right and left foot markers
      • Computation of the distance between the midpoints and the centre of the forceplate (dR or dL)
      • if dR > dL then the events are associated to the left foot and conversely
    6.  Validation of event: The fact that only one foot was on the forceplate at the time of the event and that the foot was not between two forceplates was checked automatically. The foot was modeled as a rectangle. The distance L between the heel (HEE) and toe (TOE) marker was used as reference. The foot was considerd to have a lenght of 4/3 L and an width of 2/3 L (see figure below). Then it was tested that, at the time the event, only one foot was on the forceplate and that the considered foot was within the forceplate.
  Foot rectangle representation

J.A. Zeni, J.G. Richards, J.S. Higginson, Two simple methods for determining gait events during treadmill and overground walking using kinematic data, Gait Posture. 27 (2008) 710–714. doi:10.1016/j.gaitpost.2007.07.007