View Full Version : Solutions for blocked ASIS markers during gait analysis? -responses

Tom Collins
04-28-2008, 07:23 PM
Thank you for the large number of responses to my query. The full responses are given below but I have also included a brief summary at the start. The responses roughly fit into four categories and so I have arranged them according to these.

Original request

We are having problems with tracking ASIS markers during 3D-kinematic gait analysis of some patients; we wondered if any of you had come across a similar problem and had found any solutions?: For clinical gait analysis we use a variation of the Helen Hayes marker set (Kadaba et al. 1990, J Orthopaedic Research 8(3):383-392) with 8 Qualisys infra-red cameras, distributed approximately in an oval around the measurement volume, above head height. For some patients seen in the Gait Lab, particularly with hemiplegia, one arm is held relatively immobile beside the body during walking. Depending on the particular position of the arm it sometimes blocks line-of-sight for cameras recording the position of the ASIS marker nearest to the arm, with the result that there are considerable gaps in the trajectory data for this marker. We have tested using various alternative markers, such as over the iliac crest, or using 2PSIS markers and 1ASIS, but results have notable differences when compared to the standard marker set results, which means we don't have confidence in what is being shown. One possible solution could be to strap the patient's arm in an elevated position but presumably this would alter their gait pattern, and the change would be hard to quantify. Have you encountered similar problems with some patients?If so, what solutions have you found?

Tom Collins, Gait Laboratory, Queen Mary's Hospital, London, UK

Summary or responses

Reconstruct ASIS position from alternative markers
In their study, presented at two international conferences (see refs. below), McClelland et al. placed two extra markers lateral to the pelvis on a plane with the ASISs and sacrum. When these were used to recreate ASIS data the differences in resulting joint angles were small. Other responses suggest placing markers on the iliac crests or just two PSIS markers and the remaining ASIS marker (depending in part on resolution of the motion capture system). There seem to be variations in the actual computation for reconstruction, depending on software and assumptions about marker behaviour. There appears to be variation in using reconstruction based on a static trial (with no ASIS data for dynamic trials) or only filling in gaps of dynamic trials. The reconstruction approach relies on the assumption that the pelvis is rigid, so soft tissue artefact does not introduce errors to the ASIS reconstruction that are too great.

Use a pelvis cluster
A cluster is attached over the sacral region and related to the ASIS and PSIS positions during quiet standing. During movement the anatomical frame is reconstructed from the tracked cluster markers. The responses below anecdotally report satisfactory results. Various studies have discussed the advantages/disadvantages of using clusters (e.g. Vogt et al. Gait&Posture 2003:18;178-184, Percy et al. J. Biomechanics 1987:20(10);943-949, Manal et al Gait&Posture 2000:11;38-45, Benedetti et al. Clinical Biomechanics 1998:13;204-215). This approach may be particularly useful for movements in which both ASIS markers are occluded, e.g. squatting or other activities requiring large hip flexion. Variations appear in the method used for attaching the pelvis cluster. As with reconstruction from individual markers the cluster approach relies on the assumption of pelvis rigidity.

Change camera setup
Concentrate cameras on the pelvic region. Place cameras at different elevations, particularly around waist/hip level or lower down facing up. This obviously depends on numbers of cameras available as adequate coverage is still needed for the rest of the markers.

Constrain patient’s arm
Ask the patient to walk with crossed arms or with the arm in a sling. Two studies have been referenced below; these conclude that notable changes can occur in body kinematics when an arm is constrained. This suggests that arm constraint should be avoided if not normally used by the patient. (Interestingly, a sling may improve gait characteristics for some patients.)

The full responses are copied below.

Reconstruct ASIS position from alternative markers

I have been using the Vicon system to assess patients with knee replacements whilst performing a range of activities. Like you, I initially had problems with the occlusion of the ASIS markers, so we attached 2 extra markers somewhere on the pelvis and calculated the position of the real ASIS markers in relation to the additional and sacral markers so that we could recreate the ASIS marker when it was missing. We found that during walking the joint angles produced using any of the models was always less than 1.5 degrees different to plug-in-gait (most angles were less than 1 degree different) I have submitted a manuscript for publication about this but you can read an abstract from the ESMAC conference in Gait and Posture - McClelland, J A., Grant, C., Webster, K. E. and Feller, J. A. (2007) Gait and Posture, 26, S81.

Jodie McClelland, PhD Candidate, Musculoskeletal Research Centre, School of Physiotherapy, La Trobe University
Bundoora, Australia j.mcclelland@latrobe.edu.au

McClelland JA et al 2007. Missing ASIS markers: possible alternatives. In: Biomechanics of the lower limb in health, disease and rehabilitation, University of Salford, Salford 3rd-5th September 2007:102-103

If you can get reliable data from other markers, why don't you use those markers to reconstruct the position of the ASIS markers? The specifics of how to do this would vary with your software resources and abilities, but conceptually you'd just record a static trial with 5 markers (say, sacrum, iliac, crests, and ASIS), then remove the ASIS markers andreconstruct their position during gait trials from the positions of the other three visible markers. Brian Schulz, Ph.D. Biomechanist / Career Development Awardee, VISN 8 Patient Safety Research Center (118M)James A. Haley Veterans' Hospital, USA Brian.Schulz@va.gov

Try adding extra markers along the iliac crest (optional), as well as the PSIS's (necessary), thus creating a unified transverse pelvic plane. If all of the markers are present at any instant in the data acquisition, you will have the x-y-z coordinate relationship between all of these markers. Knowing this, as long as you don't mind using the assumption that the markers all move along the same plane, you can create an algorithm that will replace the marker when it disappears since it can only be in a certain place relative to the opposing ASIS and PSIS's. I know this is all relative to the resolution of your motion capture system, skin movement, etc. etc. but this is a method I use with our Vicon Motion analysis software when I have certain trunk markers disappear for a short duration. They were actually able to provide me with this patch. Sam Perlmutter, Neuroscience PhD Candidate, Physical Therapy and Human Movement Sciences, Northwestern University, USA sperlmutter@northwestern.edu

If you can define the position of the ASIS with a marker during a static trial, and then have at least three other markers on the pelvis (say with two SACR markers and the other, visible ASIS) you can do a "replace 4" macro. Define each marker with respect to the other three (Vicon has a canned routine for this, but you can write it yourself in Matlab, excel (complicated), visual basic or whatever you are using. I think that is may be best to define each marker in relation (XYZ) to a virtual marker exactly halfway between each other pair. This way if there is marker motion its more likely to cancel out and less likely to have a systematic error propagate through to the unseen, replaced marker.

Michael Orendurff , Division Director, Movement Science Laboratory, Texas Scottish Rite Hospital for Children, USA morendurff@hotmail.com

If you are able to record the ASIS markers at least in a static trial, you could define the model and use additional markers (placed anywhere on the pelvis) to reconstruct the position of the ASIS markers from the positions of the additional markers if you want these positions. A good idea could be using one additional marker on the sacrum, as the back is very likely to be free to be seen (no belly, no arms).
Gaspar Morey Klapsing gmorey@inescop.es

Probably one which you have tried, but if you initially know the relationship of 4 markers on the pelvis segment (2PSIS & 2ASIS) you can calculate the position of a missing marker from the other three. Jon Spriggs, SPORTEX, University of Birmingham, UK j.spriggs@bham.ac.uk

We normally use 5 markers: 2ASIS, 2PSIS and sacrum, and we indifferently use the only visible markers. In effect, by using a redundant number of markers, it is possible to estimate a marker position during the gaps as well as to do an anatomical calibration. How did you use the PSIS markers? Did you correctly modify the function used to compute the pelvis reference system and the hip joint centres?
Esteban Pavan, Movement Biomechanics and Motor Control Lab, Polytechnic of Milan pavan@biomed.polimi.it

Have you tried using a digitising pointer or a static calibration file to locate the offset of the occluded marker relative to three other markers? (For example the other ASIS and the PSIS marker plus a third marker that could be anywhere on the pelvis - as long as it is not co-linear with the other two markers). After you created the 'virtual' ASIS marker you could process the data the same way you would for any other way, which should ensure consistency. Visual3D is a software that allows post-processing the data to re-create the missing marker that way:http://www.c-motion.com/ The example "Creating a landmark from 3 points and a reference location" probably describes what you would want to do: http://c-motion.com/help/Visual3D_Modeling/Landmarks/Landmark_Examples.htm
Nils Betzler, PhD Student, School Of Life Sciences, Napier University, Edinburgh, UK N.Betzler@napier.ac.uk

Based on the Helen Hayes marker set, ASIS markers are used to define the locations of the hip joints. If they are obstructed, you can add extra markers to the pelvis (ie, iliac crest, PSIS, etc), which are NOT directly used to define the hip joint centers. But they provide a reference frame to locate the required ASIS markers (based on a static trial) when they are obstructed. In some motion capture systems, this can be done through a definition of a "virtual" marker.
This virtual tracking has been quite useful, especially in patients with walkers or with extra adipose tissue. Arnel Aguinaldo, MA, ATC, Director, Center for Human Performance, San Diego, USA aaguinaldo@sdchp.comAssistant Professor, Biomechanics, Department of Exercise and Nutritional Sciences, San Diego State University

We experience similar problems in our lab (8 Qualisys cameras, distributed at different heights around the measurement volume). We have team handball players performing match-like sidestep cutting with a ball. The complex movement and the addition of a ball held in front of the player contribute to problems with trajectory gaps for the ASIS markers. We are using a marker set as shown in the picture (heel and elbow markers not shown). Our solution has been to use 4 pelvis markers, 2 ASIS and 2 PSIS. In the data analysis (Matlab), we perform a interpolation based on the other three markers when one marker is missing. This is done when more than 10 frames are missing. We assume the pelvis marker set to be rigid and calculate the position of the missing marker based on the relationship between the markers in the last frame before the gap. When making gaps in a complete trajectory and comparing the rigid body interpolation with the real tracking, we get acceptable results. For the rest of the data analysis (kinetics and kinematics), the two PSIS markers are merged to one (using the midpoint between them). We are interested in feedback to this solution, especially regarding possible pitfalls. Eirik Kristianslund, Medical student Eirik.Kristianslund@nih.no Tron Krosshaug, PhD, Department of Sports Medicine & Oslo Sports Trauma Research Center, Norwegian School of Sport Sciences, Oslo

Use a pelvis cluster

We have similar problems when doing motion analysis during drop-landing and stop jump tasks. When the subjects come to a deep squat, the ASIS markers always get blocked. We asked Vicon engineers to come up with a solution, and they suggested using a modified Plug-in Gait model. During static capture, we applied a marker cluster plate with four markers on it to the sacrum, besides the usual Plug-in Gait marker placement. Running a special program within the Vicon Nexus software allowed the Nexus to identify the relative positions of the marker cluster plate to the ASISs and PSISs markers. During dynamic motion capture, we removed the ASISs and PSISs markers, while keeping the marker cluster plate in place. Since the Nexus already knew where the ASISs and PSISs markers were relative to the marker cluster plate, it just created virtual ASISs and PSISs markers, and did the usual Plug-in Gait kinematic and kinetic calculation for all the joints. However, we haven't done detailed analysis of this method. Although during one gait analysis session, the values looked about the same for modified Plug-in Gait model and original Plug-in Gait model.
Hung-Chun "Gordon" Huang, Graduate Student Researcher, Neuromuscular Research Laboratory, UPMC Center for Sports Medicine, Pittsburgh, USA huh8@pitt.edu

We ran into a similar issue with select subject populations, and we came up with a method that accurately tracks the motion of the pelvis but does not have the problem of blocked markers. We now use a cluster we constructed of 4 markers which we carefully attach to the subject's sacrum using medical bandage. The cluster is a section of reinforced foam, onto which we've placed plastic tubing in a cross pattern. On the four ends of the tubing, we've placed a 14mm marker. Our calibration procedure for each subject then includes four additional pointing trials. We use a pointer(with markers placed at certain distances along its length) to point to the 4 anatomical landmarks(LASI, RASI, LPSI, RPSI). We then have code that uses a reference coordinate system on the cluster to bring the pelvic pointer markers into the subsequent dynamic trials.
This methodology(when properly utilized) seems to very accurately replicate the motion of the actual pelvic markers.

Jason R. Franz, Research Engineer, Gait Laboratory, Physical Medicine and Rehabilitation, University of Virginia jrf5n@virginia.edu

Perhaps the use of a pointer (with virtual markers) and a cluster of markers can aid in this capture type. Please, see the example below: http://www.c-motion.com/products/pointer.htm We still didn't use this solution - away with limitations, we used a sling in the arm that hides the marker. This problem is also common in patients with cerebral palsy (crouch gait) and that use Canadian crutches. Wagner de Godoy, AACD – Brazil wagner.godoy@bol.com.br

I haven't tried this technique before, but could you simply attach a lightweight stick at the ASIS with markers at each end of the stick that are outside of the range of arm movement and calculate the position of the ASIS based on these two markers?

Alena Grabowski, PhD, Postdoctoral Associate, Biomechatronics Laboratory, MIT Media Lab, USA alena.grabowski@gmail.com

You could try using a marker cluster rather than single markers. You only need the ASIS in a static neutral trial , after that you can remove them but only if you have say a tight elastic Velcro belt around the pelvis with several markers on it - at the same time . ( the assumption here is that you are modeling the entire pelvis as one rigid body) Then the ASIS points become virtual landmarks . This rigid body cluster approach has many advantages compared to always relying on using the single marker set. Kevin McQuade, Univ of Wa., Seattle, USA kmcquade@u.washington.edu

I used an OPTOTRAK system for some studies in graduate school, so it might not be the same but I thought I'd send my thoughts to you. We used the ISB standard for finding the hip joint with the ASIS, PSIS, and long axis of the femur (the actual reference escapes me). At any rate, we placed a rigid body on the PSIS and sacrum and digitized the PSIS and ASIS. That way, as long as the three markers on the sacrum were visible, we could get the imaginary points. Of course that makes a couple of assumptions, mainly that the sacrum moves with the pelvis during gait.Another option might be to place three markers on the hand and track its motion during gait. There might be a pattern of the hand in relation to the markers on the hip, that could then be used to fill in the gaps. Not ideal, but might give you something.Kevin McIntire, Postdoctoral Scholar, Stanford University, USA skevimc@gmail.com

Change camera setup

I do come across similar problems previously, here're some remedies that I took:1) Adjust the positions of the cameras so that there are more cameras concentrated towards the pelvic region of the subject, rather evenly spacing out the cameras around the subject. 2) Try positioning the cameras at different elevations; for instance, there must be a couple of cameras aiming from the top towards the pelvic region and another couple from the bottom towards the pelvic region.3) Use a larger size marker for the ASIS4) If the above suggestions can be implemented, chances are that gaps will be greatly reduced but not necessarily completely eliminated. Filling of minor gaps will be required based on the knowledge of the inter-ASIS distance, ASIS to Sacrum distance and the general profile of the ASIS marker trajectory from multiple trials.
Raye, Yeow Chen Hua g0501329@nus.edu.sg

I have encountered ASIS problems, not specifically in patients but in general, and the only solution I know that doesn't interfere with the subject or their gait pattern is changing the way you set up your capture volume. If possible, it will probably help you to take a couple of the cameras and bring them down to waist level in the front of your capture volume. In fact, if you are only concerned with lower-body analysis, you can bring them all down some and even move some in close while others stay back for a wider view.

Tiffany Zachry, Moss Rehabilitation Research Institute, Philadelphia, USA zachryt@einstein.edu

Back in the "old" days all 3D labs mounted their cameras at about adult hip level - mainly to deal with the problem that you describe. As the Motion Capture companies has added the ability to support more cameras, the fashion has been to mount the cameras higher. This makes the lab prettier, more techie looking and, requires that the lab buy more cameras to obtain good data. The higher camera positions provide more of a birds-eye view which causes your problems around the hip, especially if the subject uses a walker .. although on the plus side, the higher camera views do help with resolving the markers placed on the feet. You have two possible solutions, either lower some of the cameras that you have, or buy more cameras and place the new cameras about three feet off the floor at the front of the walkway so that they can see the markers as the subject moves. Edmund Cramp, Motion Lab Systems, Inc. Baton Rouge, USA eac@motion-labs.com

Have you tried placing one or preferably two cameras near ground level facing upward? This might capture the markers throughout the trials.

Bhushan Borotikar biomech_forum@yahoo.com

Constrain patient’s arm

We have the same problem in our gait lab & we had to tell the patients to cross their arms (although we use 6 Qualysis infra-red cameras). Maybe active markers can solve the problem, but we didn't buy them and we don't know how they would be in reality Mohammad Reza Farazdaghi mfarazdagh@gmail.com

Perhaps the use of a pointer (with virtual markers) and a cluster of markers can aid in this capture type. Please, see the example below: http://www.c-motion.com/products/pointer.htm We still didn't use this solution - away with limitations, we used a sling in the arm that hides the marker. This problem is also common in patients with cerebral palsy (crouch gait) and that use Canadian crutches. Wagner de Godoy, AACD – Brazil wagner.godoy@bol.com.br

Yavuzer G, Ergin S. Effect of an arm sling on gait pattern in patients with hemiplegia. Arch Phys Med Rehabil 2002;83:960-3.
>From abstract: The able-bodied group did not show any difference in gait parameters while using the sling. However, in patients with hemiplegia wearing a sling, walking speed and stance period of the paretic side increased, double support time of the paretic side decreased, excursion of the center of gravity (COG) decreased, and weight bearing of the paretic side increased Conclusions: An arm sling improved gait, especially during gait training sessions of patients with hemiplegia who have impaired body image and excessive motion of the COG.

Ford, M.P., R.C. Wagenaar, and K.M. Newell, Arm constraint and walking in healthy adults. Gait & Posture 2007:26;135-141.From abstract: The aim was to examine the functional adaptations when constraining one arm in healthy adults during treadmill walking. A comparison between walking with one arm constrained and normal walking revealed decreased, transverse pelvic, thoracic, and trunk rotation, however there were slight increases in non-constrained arm movement amplitude. Reduced arm movement amplitude did result in altered frequency and phase relations between the arm and leg. Persons with upper extremity movement dysfunction may walk slower due to atypical coordination between upper and lower body movement at higher walking velocities.
__________________________________________________ _______________
Play the Andrex Hello Softie Game & win great prizes