Quantifying Soft Tissue Artefact for the Humerus and Scapula

Summary

Although extensively utilized to estimate bone kinematics, skin-marker motion capture is plagued by errors arising from soft-tissue artefact (STA). The error caused by STA is substantial and “puts at risk the validity of a significant body of research in the basic, clinical, and applied sciences”. This project quantifies and visualizes STA for the humerus and scapula in 20 healthy subjects. The generated dataset and visualizations will serve as a guide for designing and validating STA suppression algorithms.

Code and Demos

STA Visualization	Demo	JavaScript
The code in this repository creates an interactive 3D JavaScript application - based on three.js - for visualizing STA of the humerus and scapula. It is currently in the proof-of-concept phase, and updates will be forthcoming at the end of December 2020. Feel free to try out the live demo of the code (please be patient, the 3D bone models may take 5-20 seconds to download).

Current Status

• I have stopped working on this project because my research on quantifying axial rotation was expanded onto two manuscripts - one of which has been published and the other one is undergoing the peer-review process. The manuscript for this project will be written by another graduate student in our lab.
• Biplane fluoroscopy and skin marker data from all 20 subjects has been captured and organized in a Subversion repository.
• Data analysis is currently at ~80% complete.
• JavaScript application for visualizing STA has been completed.

Background

Biomechanics is the study of human (and animal) movement from a mechanical perspective. The principles of mechanics can be applied at all biological levels (e.g. joint, tissue, fiber, cell, chromosome) but my research focuses on joint biomechanics, and specifically joint kinematics. Kinematics is the study of the motion of objects without considering the forces that cause them to move. In a typical study of kinematics, the researcher will record the motion of individual body segments (e.g. foot, hand, thigh, arm, head) while the subject under study performs a prescribed motion (e.g. arm elevation, forward elevation, combing hair, throwing a baseball). Kinematic recordings provide a foundation for a variety of scientific inquiries including: prosthetic design and testing, computational models, clinical diagnosis of pathological conditions, and rehabilitation techniques.

Most studies rely on skin marker motion capture to record body segment kinematics. In this technique multiple markers (at least 3) are affixed to the skin at key anatomical landmarks on each segment of interest. The subject is then instructed to perform a sequence of prescribed motions while being recorded by a set of cameras. At least two cameras are necessary to infer the 3D position of the individual markers, but to increase accuracy and reliability many more cameras are utilized (I have noticed that most studies utilize at least 8). Modern motion capture systems can locate markers with at least 1 mm of accuracy. Skin marker motion capture is easy to implement, relatively inexpensive, and is non-invasive. Unfortunately, it has a major drawback: it can be highly inaccurate.

So far, I have left out one key detail that explains the source of this inaccuracy: in a kinematic study we are interested in recording the motion of the bone underlying the segment of interest. As the subject moves, because of skin gliding, muscle contraction and inertial effects, the motion of the markers on the surface of the skin will be decoupled from the motion of the bone. We label this source of inaccuracy soft tissue artefact (STA). The effect of STA can be substantial, especially for rotations of the bone around its long axis. No widely accepted method for reducing the effect of STA on bone kinematics exists. One key impediment to devising such a method is the lack of datasets that quantify STA. The paucity of datasets characterizing STA inhibits our ability to mathematically model it and to devise an algorithm for attenuating its effects on bone kinematics. The reason why few studies measure STA is simple: it's difficult to measure.

To measure STA, the true motion of the bone and the motion of skin markers must be recorded simultaneously. Measuring the true motion of the bone is challenging, that is why skin marker derived kinematics are utilized as a surrogate measure. Three main techniques exist for measuring the true motion of the bone: radiostereometric analysis (RSA), bone pins, and model-based tracking.

• In RSA, small metal beads are implanted into the bone of interest then biplane fluoroscopy (two x-ray systems positioned at an angle to each other) is utilized to record a subject's motion.
• The bone pins technique involves drilling pins, which have markers attached to them, directly into the bone of interest. The same camera system utilized to track skin markers records the markers attached to the bone pins while the subject performs a prescribed motion.
• In model-based tracking, a 3D model is created from a computed tomography (CT) - or magnetic resonance imaging (MRI) - scan of the bone of interest. The subject's motion is recorded via biplane fluoroscopy, and - utilizing specialized software - the 3D model is spatially aligned to simultaneously match both videos obtained from biplane fluoroscopy. Unfortunately, the 3D alignment process is only semi-automatic, requiring a significant amount of human intervention.

The obvious and significant disadvantage of RSA and bones pins is their invasive nature. Although model-based tracking is non-invasive, it requires significant post-processing (human time capital) and can expose a patient to radiation if a CT scan is utilized.

In this project, I will measure STA by piggybacking on an investigation designed to study reverse shoulder replacement. As a control, this investigation measured the motion of 20 healthy subjects using both model-based tracking and skin markers. This dataset provides the perfect setting for measuring and visualizing STA.