Stance phase event detection based on kinematic algorithms

[Farzaneh Yazdani]

Dear all,
I need to compare the peak values of the moment, kinematic and GRF time series during sub phases of the stance (first and second double support and single limb support) between two groups during gait. We use vertical component of the GRF to accurately determine the stance phase and all data are time normalized from 0% to 100% of the stance phase. The question is, as long as there is no difference between walking speed of the two groups can I simply determine the first double support at 0% to 10% of the gait cycle and the second double support at 50% to 60% of the gait cycle which means that the 10% to 50% of the cycle is single limb support??
Is there any kinematic based algorithm or specific kinematic event to detect these sub phases of the stance?

any help and comment will be greatly appreciated

Regards

F.Yazdani
PhD Candidate, Center for Human Motion Researches, SUMS.

[Allan Carman]

Re: Stance phase event detection based on kinematic algorithms

Farzaneh,

I wrote the original response bellow, and then realized I may have missed the point of the question which is about finding the maximum of kinematic and kinetic variables during specific phases of stance rather than normalization of kinematic and kinetic variables relative to the different phases.

So to keep this response short and to the point:
If you want peak magnitudes and occurrences do not normalize your gait data. Keep it in the original sample rate. Typically you normalize to produce a representative ensemble average curve and confidence interval for a group of curves.
Do not assume that the double-single-double support phases of stance for all participants and trials will be 0%-10%-50%-60% of the gait cycle (see below). Assuming everyone is the same will only introduce errors into your analysis. Establish the occurrence (sample number) of each heel-strike and toe-off for each foot and gait cycle (also see below). Use this to identify the occurrence of each stance phase and locate the maximum of the kinematic and kinetic variables during each phase.

Original response (normalizing kinematic and kinetic variables to different phases of gait):
As you are looking at potentially subtle variations in kinematics within the stance phase I would not assume that all double-single-double support phases in stance for every participant will occur at 0%-10%-50%-60% of the gait cycle. Although the overall mean velocities may be the same between groups there will be variations in walking speed between participants as well as some small variations in speed between and within each trial.

I would recommend identifying each heel strike and toe-off occurrence of both feet and for every trial. This may be time consuming but it will allow you to extract kinematic and kinetic data based on occurrence of each phase with stance as well as determine the overall mean of the double-single-double support durations for your group (and stance vs swing phase if you wish). The overall mean may well be close to 0%-10%-50%-60% of the gait. The difference now is that you extract the kinematic data based on the respective double-single-double support occurrences (from respective heel strike and toe-off data) and then normalized each phase to the representative mean phase duration of the group.

I have done something similar with stance and swing phases of gait, as there are large differences in variability and magnitude of knee joint angle data between stance and swing phases. This normalization within a gait cycle removes a source of temporal variability in gait data, meaning that when producing ensemble averages of joint angle data the underling stance and swing phase align within individual gait cycles. We remove a major source of temporal variability when we traditionally extract a gait cycle from heel-strike to heel-strike and normalize to 100%. Up until now I have not considered extracting data and normalizing to the different phases within stance?

If you have multiple force plates you have a criterion method for determining heel strikes and toe-offs, otherwise inertial sensors (IMU’s) are a possibility or using the vertical and/or horizontal position or acceleration data of heel and toe markers. I would not use kinematic (joint angle) data to identify heel strike and toe-off occurrences for normalizing gait data. If you are looking for changes in kinematic variables then you need an independent measure to identify the phases within gait. Otherwise you are changing the nature of the variable you are trying to measure based on patterns within the same or a related variable when normalizing the data.

Cheers
Allan

[Farzaneh Yazdani]

Re: Stance phase event detection based on kinematic algorithms

Dear Allan,
Based on your valuable comment I think the best method to extract peak values of kinematic and kinetic measures of my study will be using the vertical position of the heel and toe markers, so is it correct to assume that the period between HS frame and the first frame which has the lowest vertical position (zero z value I think) for both Rt heel marker and Lt toe marker is the first Double limb support, the period between the second frame with lowest z value for Rt toe marker and Lt heel marker and the TO frame is the second double limb support and the remaining frames will belong to single limb support?

I really appreciate your time and consideration

[Matyas Varga]

Re: Stance phase event detection based on kinematic algorithms

Hi,

I am a phd student myself and did some gait analysis using Vicon and gyroscopes. As long as you can identify healstrikes and toeoffs, you can work out the phases. These two events can be identified using the angular velocity of the shank in the saggital plane (the first peak after the midswing peak is healstrike and the first one before it is toe off) or simply by toe and heal vertical positions.

I hope this helps, if yoh have any question, please ask!

Thanks

Matt

Dear all,
I need to compare the peak values of the moment, kinematic and GRF time series during sub phases of the stance (first and second double support and single limb support) between two groups during gait. We use vertical component of the GRF to accurately determine the stance phase and all data are time normalized from 0% to 100% of the stance phase. The question is, as long as there is no difference between walking speed of the two groups can I simply determine the first double support at 0% to 10% of the gait cycle and the second double support at 50% to 60% of the gait cycle which means that the 10% to 50% of the cycle is single limb support??
Is there any kinematic based algorithm or specific kinematic event to detect these sub phases of the stance?

any help and comment will be greatly appreciated

Regards

F.Yazdani
PhD Candidate, Center for Human Motion Researches, SUMS.