Seminars on 3D gait reliability avaible on youtube

[Allan Carman]

I have posted two 1hr seminars on YouTube that address 3D gait methods and their reliability. They were informal seminars at the School of Sport and Recreation, AUT University, Auckland,
and available on YouTube under the channel 'SPRINZ seminars' and titled
'Allan Carman 3DMA reliability' part 1 and part 2
These will be of wider interest to those working in 3DMA and particularly relevant to the recent post on kinematic cross-talk, as they explore current method, factors affecting reliability, normal gait kinematics and a modern approach to 3DMA.

Allan Carman

[Bernard]

Re: Seminars on 3D gait reliability avaible on youtube

Dear Allan,

Thanks for your video. Regarding your post hoc treatment of kinematic cross talk, I was abit less clear. Do you have a paper that you can refer me too? Or a methods paper? Do you have an opinion on the use of PCA to 'correct the axis'?

Regards,
Bernard

[Allan Carman]

Re: Seminars on 3D gait reliability avaible on youtube

I am not in favour of PCA for axes alignment and minimizing cross-talk

1. The axes about which the z,y’,z’’ rotations are made are not orthogonal.
2. These respective rotations are not independent
3. In gait there are moderate correlations between the three rotations for a given joint, perhaps indicating some functional dependency between movements such as weight bearing, knee flexion, knee internal rotation and knee adduction.

For the gait cycle correlations between flex/ext and Abd/Add; Flex/ext and Ext/Int; Abd/Add and Ext/Int for the hip are about 0.7, 0.7, 0.9, for the knee about 0.5, 0.5, 0.6 and for the ankle 0.8, 0.5, 0.3.

1. Nonlinear error (cross talk) results from misalignment of the proximal and distal segments axes with respect to the true axes of rotation. Introducing erroneous rotations. A segment axis misalignment affects calculated joint rotations at both the proximal and distal joints.
2. The PCA method treats the three calculated rotations as if they are independent variables. As in an orthogonal three axes coordinate system.
3. The PCA method performs a linear transformation of the data to in effect find the principle axes that describing the distribution of data (where rotations x,y’,z’’ are treated as x,y,z coordinates in a point mass or inertia distribution) about an imaginary orthogonal axis system. The correlations between the transformed flex/ext (x) and abd/add (y’) knee joint rotations in gait are reduced to near zero (Baudet etal. 2014) when expressed in the imaginary axis system.
4. The PCA method also treats the knee rotations in isolation without considering the ankle or hip

In summary the PCA method does not address the source of the error (the axis misalignment), the non-linear nature of joint rotations and their error, considers the knee joint in isolation, and by minimizing correlations between joint rotations removes real knee abd/add and Ext/Int rotations and correlations (even if small) that exist between the three rotations of the same joint.

[Bernard]

Re: Seminars on 3D gait reliability avaible on youtube

Many thanks Allan for the great insight. Would you have a reference (or a paper, so I can use the methods) to the method of adjusting cross talk you employed in the presentation?

Regards,
Bernard

[Allan Carman]

Re: Seminars on 3D gait reliability avaible on youtube

There is no published paper for the method but I am working on one. I viewed the post-hoc correction as a fine tuning of axes alignment and should not relied upon alone for aligning axes. Every effort should be spent on the initial estimation and the first refinement of axes alignment in the static pose and trial movement (The first two step presented in the seminar) prior to calculating segment rotations from the movement trials. Which is then followed by post hoc correction of calculated rotations (the third step or refinement).

Cheers
Allan