An early unit of our Masters Programme in Clinical Gait Analysis has focused on how best to capture clinical video. One of the things we talked about was parallax and I was surprised at how little awareness there was of this as an issue. Parallax describes a number of phenomena associated with how what you see changes with your line of sight. Wikipedia is filled with examples from astrophysics. The video below shows some fairly gentle examples of how the relationship of foreground and background objects changes as your point of view moves.

In clinical gait analysis the main effect we are concerned with is that how we perceive a given angle depends on the position we are viewing it from. Thus if the knee is flexed to 90° in the sagittal plane and we are looking from a position perpendicular to that plane then we will perceive the angle to be 90° but if we are looking from either a little in front or a little behind then we will perceive the knee to be more extended. More than that, if the thigh is internally rotated so that the knee is not flexing in the true sagittal plane then we will under-estimate knee flexion even if we are square on to the person. We perhaps understand this best in relation to the coronal plane where most of us know that if the knee is flexed then we will see what appears to be varus or valgus if either we are not looking straight on at the person or if the thigh is internally (or externally) rotated.

To help understand and also to quantify effects I’ve prepared the little Flash animation below. Imagine there is an object in the centre of your gait lab composed of two red rods set exactly at right angles. As you first see the animation the camera is rotating around the object and you can see how the view seen by the camera changes. You can see that even though the object in the lab is entirely stationary the angle read off the computer screen varies from 0° to 90°. If you click on the pause button the animation will stop and you’ll then be able to drag the camera to whatever position you want. Play around with this and see how the perceived angle changes with the camera angle.

What you should find is that the effects are relatively small in the sagittal plane. If you have the camera within +/-13° to the true perpendicular then you will perceive an error of less than 1°. You need to be more than 24° off true to be more than 5° out. Positioning the camera to get a true coronal plane view, however, is much more critical. Being just 5° out in camera alignment will result in an erroneous reading of either 10° of knee varus or valgus (depending on which side of the walkway the camera is viewing from). For those concerned with symmetry of gait remember you get a double whammy as you’ll see an apparent 10° of valgus on one side and of varus on the other side. I must admit than in constructing the animation the effects in the coronal plane are quite a lot bigger than I was expecting. (For the geeks the animation is based on an isometric projection – perspective effects with a camera close to the person could exaggerate the effects even further.)

Of course this varus-valgus thing is the cause of all those huge coronal plane knee artefacts that we all get in swing when we are doing gait analysis. If your don’t identify the coronal plane of the thigh properly from the markers you place for your static calibration you are effectively causing your gait analysis system to look at the knee from the wrong angle and all that knee flexion that occurs in mid-swing will appear as varus or valgus on your knee traces. If you see this happening a lot in the gait data then its worth reviewing how you place these markers.

For those  of you not at ESMAC Eric Desailly presented quite a nice paper (page 33) showing that although this effect is a significant factor contributing to those dodgy looking knee trace it may not be the only one.

It’s nice to receive the first review of my book (in the Journal of Biomechanics) that I’ve been able to read easily (the only other one I’m aware of is in French and was really a little beyond what I’d learned in school over thirty years ago).

It’s interesting that Steve has picked up on the issue of the omission of sample paperwork which he feels would have been useful. My original intention had been to include a whole suite of procedures and protocols which could have been copied and pasted into service level documentation folders across the world. The first thing that prevented me was very straight forward; I had a page limit to work to from the publishers.

The other thing that made me change my mind, however, was the growing realisation that the process of writing service level documentation is actually more important than the result. Sitting down and writing such documents is an extremely useful exercise and doing it properly ensures that you have thought through the issues for yourself. I’d suggest that it’s much better for gait analysis service teams to produce such documents for themselves than to rely on what some “expert” has written for a different service operating in a different context.

Take a Referral Form for example. In some ways this is simply a piece of paper that tries to force the referring clinician to tell you what you need to know about the patient so that you can do your job properly (don’t get your hopes up – I don’t know of any gait analysis service in the world that actually gets useful information from referrers however well-structured the form).  On the other hand though, it reflects the relationship you want to have with your referrers and may include an implicit or explicit specification of the patients you feel qualified to assess. You can just pinch a copy of someone else’s form, but it is much more valuable to work through these issues for yourself and develop your own form that reflects the characteristics of your service. Thinking about what you do is at least as important as doing it.

There is a similar issue with normative reference data (see YouTube video of my presentation last year to GCMAS on the subject). The main reason for collecting normative reference data is to learn from the experience. Putting 30 able bodied volunteers through the lab, reflecting on the mean traces and variability around it and comparing this with data from other service is a learning and quality assurance exercise that all services should go through (and should repeat every so often). Pinching someone else’s data simply so you can have those nice grey bands in the background of your own graphs is missing the point entirely. It still amazes me how many services do this.

As this post’s title suggests, this is purely my personal opinion. Maybe specimen forms that could have been adapted would have been useful. Maybe I should have made these available but in .pdf format so people would at least have to do the word processing for themselves! The page limit argument is a bit ridiculous in a world of electronic supplements. What do you think? Maybe if enough  people demand them through comments here I’ll work on it for the next edition (if people buy enough copies to make Mac Keith want to publish one!)

Apologies for not writing a more substantive post but I’ve been working hard over the last few weeks to get our new MSc in Clinical Gait Analysis by distance-learning running.  We had our first virtual classroom on Wednesday evening which was quite a milestone for me. Inevitably there were teething troubles particularly with admissions and registrations process but the final result is that we’ve ended up were we wanted to be with a good sized cohort of students embarking on the programme together. It’s large enough for a feeling of community to develop, small enough for things to feel manageable in this first year. Rather remarkably the seven students represent 4 continents which brings challenges for timetabling the virtual classrooms! We actually had applications from students on the two other continents but they came too late for us to get them on board on time.

The programme is designed for staff already working in gait analysis services. Staff currently employed in laboratories tend to take on either a “clinical” or “technical” role and one of the aims is to encourage and support individuals to develop competencies in all aspects of the role of a clinical gait analyst. It’s part of the CMAster project and next year VU Amsterdam and KU Leuven will be offering different residential options with potential for students to split their studies between any two of the three universities.

Our programme that spreads over three years of part-time study. The first two years are taught and the third gives the students a chance to undertake their master’s project. The taught programme is divided into four modules: Measuring walking, Healthy walking, Walking with pathology and Clinical interpretation. The first two modules are taught primarily by experiential learning requiring students to perform lots of gait analysis to consolidate their understanding of how to capture data and the reasons that healthy people walk the way they do. We’ve also managed to steer clear of any exams with assessment through a range of more relevant tasks.

All in all I’m pretty happy for the programme and looking forwards to it. Perhaps more importantly for this blog it is really starting to raise some issues for me that would make good topics for posts. All I need is the time to write them. So here’s looking forward to the new academic year and more on-line interaction.

So what went wrong? The determinants of gait are clearly wrong from a technical perspective and it is doubtful whether they contribute anything clinically. Yet they continue to be taught across the world as a basis for the understanding of human walking. If you doubt this then Google up “determinants of gaits” and see just how many sets of slides you get from educators at a range of different institutions.

I think the problem is in the persuasive simplicity of the models that Inman and Eberhart proposed. The pictures of simple articulated structures with neat ball joints and tidy pin joints are just so attractive. There’s  a reason for this which is that, generally, they are very good models, there’s absolutely nothing wrong with them. The problem is in the data that is used to drive the models. The hip joint is a ball and socket and could move as suggested by Inman and Eberhart it’s just that it doesn’t (during normal human walking). In the language of computational biomechanics it is not the model that is wrong but the simulation.

The fundamental problem is that no matter how convincing any simulation looks it is essentially an hypothesis that needs to be tested with experimental measurement.  One of the bizarre aspects of the determinants of gait story is that the group at Berkeley where probably the only people in the world at the time who actually had the data available to test their hypotheses. For some reason they chose not to (being fair, this was 60 years ago, when performing such an analysis would have been much more difficult than today). The result can also been seen as a failure of peer review publication – the paper should have been recognised as an untested hypothesis and rejected (again this is judging the world of sixty years ago by contemporary standards).

The take home message has implications beyond the study of walking. I suspect that simple and attractive models that have never been adequately tested are in use throughout biomechanics. I’d be interested to know if any readers can suggest widely accepted models that could do with being tested as rigorously as the Determinants of Gait have been over the last fifteen years or so. The Root  model (1977) that is the basis of modern podiatry is one example that is the subject of much current investigation although I’m not sure that it has ever been accepted particularly well outside the field of podiatry.

.

Root ML, Weed JH, and Orien WP, Normal and Abnormal Function of the Foot – Clinical Biomechanics Volume II, Los Angeles: Clinical Biomechanics Corp; 1977