kirtley24

01-16-2008, 09:34 AM

Dear Bill/others who might be interested,

Your questions...

1. Stansfield (2003) and Stansfield (2006) papers in Gait & Posture did

not include time-resolved data for slope m, intercept c, and correlation

r, all of which you have nicely presented at 2%-of-the-gait-cycle

increments. Did the authors provide those data to you by private

communication, or were they published elsewhere?

> No - I simply manually digitised their graphs (very tedious!) and then a

student (Anji Wall)calculated all the regression coefficients with a

MATLAB/Excel program. At Hof provided his actual EMG data (it's on the CGA

site) so this was much simpler to crunch.

2. The data used to compute the regression coefficients were collected

from children of average age 9.6 yrs, range 7-12. In the spreadsheet,

the regression data are used to predict gait parameters as functions of

normalized velocity for an adult. The authors warn against such an

"extrapolation". Since the normalized velocities of adults and children

are similar, maybe it is OK, but is there evidence to support it?

>Yes, you're right. But until someone else publishes their data it's all

we've got. I doubt there will be many differences, but of course there may

be. Perhaps more importantly, the Edinburgh group use an unusual marker

attachment which results in a flexion offset on th eknee angle and other

discrepancies with other labs' normative data. I really did this as an

exercise to encourage people to generate the same spreadsheet for their own

dataset.

3. In the spreadsheet, velocity is normalized by leg length, but

Stansfield et al (2006) normalize velocity by height. Stansfield et al

(2003) compare normalization of gait data by height and by leg length

and conclude that neither is perfect and that the choice between them is

difficult; in that paper they do not report regressions of gait

parameters vs velocity. If the regression parameters m and c in the

spreadsheet are based on Stansfield's 2006 paper, in which velocity is

normalized by height, then the velocity in the spreadsheet should be

normalized by height, not leg length.

>I prefer leg-length, and I assume I must have used 55% height for this, but

I confess I can't recall. I don't think it would make much difference to the

results.

4. The spreadsheet formula used to estimate Gastroc EMG has a small

error: the wrong intercept c is used (at time 0, intercept from time 2

is used, etc). The error does not significantly change the predicted EMG.

>Thanks... there may be more!

5. First page of spreadsheet contains some "hidden" rows which are used

to calculate values on rows labelled "Fail". These seem to be

alternative values for slope m, computed when the correlation r is less

than 0.8. I'm not sure what these are used for.

>Yes, I was concerned that there might be times when a linear regression

would be inadequate. I defined this as less than 0.8 and that's the ones

marked with a "Fail". If you examine them closely you'll see that this

almost always occurs when the absolute value of the data is small (i.e.

close to zero), so it's not surprising that the regression fails. There are

one or two 'real' failures, but surprisingly few considering the amount of

data involved.

I presented this idea at the 2002 gait meeting: Kirtley C, Cisper J, Sukal T

& Wall A (2002) Normalization for Autointerpretation of Gait Analysis Data

Gait and Clinical Movement Analysis Society (GCMAS), Chattanooga, Tennessee,

April 17-20. I have also written it up as a paper but decided not to submit

it for many of the reasons you have pointed out :-)

It would be really nice if someone could repeat the work with their data. It

might then be used as a normative dataset for comparisons with patient data.

Chris

On 1/16/08, William Rose wrote:

>

> Chris and Gili,

>

> Thank you, Chris, for making the spreadsheet with normalized gait data

> available. The data you have compiled is very nice. I particularly

> like the way you have allowed us to enter a value for velocity (cell in

> red) and instantaneously see how all the graphs change. It helps my

> understanding and will be useful for teaching. I have a couple of

> comments/questions.

>

> 1. Stansfield (2003) and Stansfield (2006) papers in Gait & Posture did

> not include time-resolved data for slope m, intercept c, and correlation

> r, all of which you have nicely presented at 2%-of-the-gait-cycle

> increments. Did the authors provide those data to you by private

> communication, or were they published elsewhere?

> 2. The data used to compute the regression coefficients were collected

> from children of average age 9.6 yrs, range 7-12. In the spreadsheet,

> the regression data are used to predict gait parameters as functions of

> normalized velocity for an adult. The authors warn against such an

> "extrapolation". Since the normalized velocities of adults and children

> are similar, maybe it is OK, but is there evidence to support it?

> 3. In the spreadsheet, velocity is normalized by leg length, but

> Stansfield et al (2006) normalize velocity by height. Stansfield et al

> (2003) compare normalization of gait data by height and by leg length

> and conclude that neither is perfect and that the choice between them is

> difficult; in that paper they do not report regressions of gait

> parameters vs velocity. If the regression parameters m and c in the

> spreadsheet are based on Stansfield's 2006 paper, in which velocity is

> normalized by height, then the velocity in the spreadsheet should be

> normalized by height, not leg length.

> 4. The spreadsheet formula used to estimate Gastroc EMG has a small

> error: the wrong intercept c is used (at time 0, intercept from time 2

> is used, etc). The error does not significantly change the predicted EMG.

> 5. First page of spreadsheet contains some "hidden" rows which are used

> to calculate values on rows labelled "Fail". These seem to be

> alternative values for slope m, computed when the correlation r is less

> than 0.8. I'm not sure what these are used for.

>

> Thank you again for making these data available, particularly in such a

> nice usable format.

>

> Bill Rose

>

>

> Chris Kirtley wrote:

> > Dear Gili/others interested,

> >

> > I made a spreadsheet some time ago that has everything you need I think.

> It

> > calculates all the various data according to your selected walking

> speed,

> > based on At Hof's non-dimensional normalisation. I used Ben Stansfield's

> > results published in Gait & Posture.

> >

> > http://www.univie.ac.at/cga/data/SpeedAllSummary.xls

> >

> > Chris

> >

> >

> > On 1/15/08, Gili Tishler wrote:

> >

> >> I am a senior student in Mechanical Engineering from Ben Gurion of the

> Negev

> >> University in Israel. My senior project concerns Biomechanics

> >> and requires normative gait cycle data of an adult (around the age of

> 20).

> >> While searching for kinematics (joint angles) and kinetics (joints

> moments),

> >> I found the Biomech website and thought maybe it can help me obtain the

> gait

> >> data I need. The best form to obtain this data is in spreadsheet format

> (

> >> e.g. Excel) or in text file, if you have it. If not, maybe you could

> >> suggest other people / websites that will have this data.

> >>

> >> Thank you in advance,

> >> Have a good day

> >> Gili Tishler

>

--

Dr. Chris Kirtley MB ChB, PhD

608 Dockside

44 Ferry St.

Kangaroo Point

Queensland 4169

Australia

Tel. (07) 3891 6644 x 1608

Fax 3891 6900

West End Family Medical Centre

Wednesdays & Fridays (07) 3844 4111

Clinical Gait Analysis: http://www.univie.ac.at/cga

Book:

http://www.amazon.co.uk/exec/obidos/ASIN/0443100098/203-6674734-4427132

Your questions...

1. Stansfield (2003) and Stansfield (2006) papers in Gait & Posture did

not include time-resolved data for slope m, intercept c, and correlation

r, all of which you have nicely presented at 2%-of-the-gait-cycle

increments. Did the authors provide those data to you by private

communication, or were they published elsewhere?

> No - I simply manually digitised their graphs (very tedious!) and then a

student (Anji Wall)calculated all the regression coefficients with a

MATLAB/Excel program. At Hof provided his actual EMG data (it's on the CGA

site) so this was much simpler to crunch.

2. The data used to compute the regression coefficients were collected

from children of average age 9.6 yrs, range 7-12. In the spreadsheet,

the regression data are used to predict gait parameters as functions of

normalized velocity for an adult. The authors warn against such an

"extrapolation". Since the normalized velocities of adults and children

are similar, maybe it is OK, but is there evidence to support it?

>Yes, you're right. But until someone else publishes their data it's all

we've got. I doubt there will be many differences, but of course there may

be. Perhaps more importantly, the Edinburgh group use an unusual marker

attachment which results in a flexion offset on th eknee angle and other

discrepancies with other labs' normative data. I really did this as an

exercise to encourage people to generate the same spreadsheet for their own

dataset.

3. In the spreadsheet, velocity is normalized by leg length, but

Stansfield et al (2006) normalize velocity by height. Stansfield et al

(2003) compare normalization of gait data by height and by leg length

and conclude that neither is perfect and that the choice between them is

difficult; in that paper they do not report regressions of gait

parameters vs velocity. If the regression parameters m and c in the

spreadsheet are based on Stansfield's 2006 paper, in which velocity is

normalized by height, then the velocity in the spreadsheet should be

normalized by height, not leg length.

>I prefer leg-length, and I assume I must have used 55% height for this, but

I confess I can't recall. I don't think it would make much difference to the

results.

4. The spreadsheet formula used to estimate Gastroc EMG has a small

error: the wrong intercept c is used (at time 0, intercept from time 2

is used, etc). The error does not significantly change the predicted EMG.

>Thanks... there may be more!

5. First page of spreadsheet contains some "hidden" rows which are used

to calculate values on rows labelled "Fail". These seem to be

alternative values for slope m, computed when the correlation r is less

than 0.8. I'm not sure what these are used for.

>Yes, I was concerned that there might be times when a linear regression

would be inadequate. I defined this as less than 0.8 and that's the ones

marked with a "Fail". If you examine them closely you'll see that this

almost always occurs when the absolute value of the data is small (i.e.

close to zero), so it's not surprising that the regression fails. There are

one or two 'real' failures, but surprisingly few considering the amount of

data involved.

I presented this idea at the 2002 gait meeting: Kirtley C, Cisper J, Sukal T

& Wall A (2002) Normalization for Autointerpretation of Gait Analysis Data

Gait and Clinical Movement Analysis Society (GCMAS), Chattanooga, Tennessee,

April 17-20. I have also written it up as a paper but decided not to submit

it for many of the reasons you have pointed out :-)

It would be really nice if someone could repeat the work with their data. It

might then be used as a normative dataset for comparisons with patient data.

Chris

On 1/16/08, William Rose wrote:

>

> Chris and Gili,

>

> Thank you, Chris, for making the spreadsheet with normalized gait data

> available. The data you have compiled is very nice. I particularly

> like the way you have allowed us to enter a value for velocity (cell in

> red) and instantaneously see how all the graphs change. It helps my

> understanding and will be useful for teaching. I have a couple of

> comments/questions.

>

> 1. Stansfield (2003) and Stansfield (2006) papers in Gait & Posture did

> not include time-resolved data for slope m, intercept c, and correlation

> r, all of which you have nicely presented at 2%-of-the-gait-cycle

> increments. Did the authors provide those data to you by private

> communication, or were they published elsewhere?

> 2. The data used to compute the regression coefficients were collected

> from children of average age 9.6 yrs, range 7-12. In the spreadsheet,

> the regression data are used to predict gait parameters as functions of

> normalized velocity for an adult. The authors warn against such an

> "extrapolation". Since the normalized velocities of adults and children

> are similar, maybe it is OK, but is there evidence to support it?

> 3. In the spreadsheet, velocity is normalized by leg length, but

> Stansfield et al (2006) normalize velocity by height. Stansfield et al

> (2003) compare normalization of gait data by height and by leg length

> and conclude that neither is perfect and that the choice between them is

> difficult; in that paper they do not report regressions of gait

> parameters vs velocity. If the regression parameters m and c in the

> spreadsheet are based on Stansfield's 2006 paper, in which velocity is

> normalized by height, then the velocity in the spreadsheet should be

> normalized by height, not leg length.

> 4. The spreadsheet formula used to estimate Gastroc EMG has a small

> error: the wrong intercept c is used (at time 0, intercept from time 2

> is used, etc). The error does not significantly change the predicted EMG.

> 5. First page of spreadsheet contains some "hidden" rows which are used

> to calculate values on rows labelled "Fail". These seem to be

> alternative values for slope m, computed when the correlation r is less

> than 0.8. I'm not sure what these are used for.

>

> Thank you again for making these data available, particularly in such a

> nice usable format.

>

> Bill Rose

>

>

> Chris Kirtley wrote:

> > Dear Gili/others interested,

> >

> > I made a spreadsheet some time ago that has everything you need I think.

> It

> > calculates all the various data according to your selected walking

> speed,

> > based on At Hof's non-dimensional normalisation. I used Ben Stansfield's

> > results published in Gait & Posture.

> >

> > http://www.univie.ac.at/cga/data/SpeedAllSummary.xls

> >

> > Chris

> >

> >

> > On 1/15/08, Gili Tishler wrote:

> >

> >> I am a senior student in Mechanical Engineering from Ben Gurion of the

> Negev

> >> University in Israel. My senior project concerns Biomechanics

> >> and requires normative gait cycle data of an adult (around the age of

> 20).

> >> While searching for kinematics (joint angles) and kinetics (joints

> moments),

> >> I found the Biomech website and thought maybe it can help me obtain the

> gait

> >> data I need. The best form to obtain this data is in spreadsheet format

> (

> >> e.g. Excel) or in text file, if you have it. If not, maybe you could

> >> suggest other people / websites that will have this data.

> >>

> >> Thank you in advance,

> >> Have a good day

> >> Gili Tishler

>

--

Dr. Chris Kirtley MB ChB, PhD

608 Dockside

44 Ferry St.

Kangaroo Point

Queensland 4169

Australia

Tel. (07) 3891 6644 x 1608

Fax 3891 6900

West End Family Medical Centre

Wednesdays & Fridays (07) 3844 4111

Clinical Gait Analysis: http://www.univie.ac.at/cga

Book:

http://www.amazon.co.uk/exec/obidos/ASIN/0443100098/203-6674734-4427132