Estimating muscle attachment contours by transforming geometrical bone models

https://doi.org/10.1016/j.jbiomech.2003.08.005Get rights and content

Abstract

For individualization of a biomechanical model, it is necessary to estimate the muscle attachments of the person to whom it is to be adapted. One of the methods to estimate muscle attachments is to use model transformations to transform a model with known muscle attachments to the bones of a person. We hypothesize that the location and shape of muscle attachment sites correlate with the shape of the bones they are attached to. If this hypothesis holds, it is possible to predict the location of muscle attachments when the shape of the bones is known. To validate this hypothesis, geometric models of three sets of shoulder bones were built. These models consist of 3-D surface models of the scapula, clavicle, and humerus, with the muscle attachment contours connected to them. By means of geometric transformations, the models were transformed, so the muscle attachments of the different data sets could be compared. Using these techniques, 50 per cent of the muscle attachment contours could be predicted with high accuracy. The muscle attachment contours that could not be predicted were all influenced by measurement errors. For 30 per cent of the muscle attachment contours, it was not possible to distinguish the interindividual differences from the inaccuracies of the method used. From this study, we concluded that most muscle attachment contours can be predicted by means of geometric models of the bones.

Introduction

In biomechanical research, models are used, among others, to simulate the movements of parts of the human body. For the shoulder, this started with complete shoulder specimens, where part of the muscles replaced were by threads (Fick and Weber, 1877; Mollier, 1899). These techniques are still used (Wuelker et al., 1995), but more and more computer models become available (Högfors et al. (1987), Högfors et al. (1991); Karlsson and Peterson, 1992; van der Helm, 1994a; Niemi, 1996; Maurel and Thalman, 1998). These computer models are often based on cadaver measurements (Högfors et al., 1987; Veeger et al., 1991; van der Helm and Veenbaas, 1991; Karlsson and Peterson, 1992; Klein Breteler et al., 1999). Medical imaging techniques, such as CT and MR imaging, supply data that can be used for biomechanical models (Koolstra and van Eijden, 1992; Kalra, 1995). In 1995 data from The Visible Human Project became available, providing a large and detailed data set used by a large group of researchers (National Library of Medicine (NLM), 1995). A biomechanical shoulder model was made based on these data (Maurel and Thalman, 1998). Questions that can be asked for all these methods of obtaining data are: What are the quality and the reproducibility of these measurements? Is it justifiable to use specific models, based on data of only one subject, in a general manner?

In our laboratory, a biomechanical model of the human shoulder has been developed (van der Helm, 1994a). This model, also known as the Dutch Shoulder Model, is based on the measurements of one out of a group of seven cadavers (van der Helm et al., 1992). In the dissection experiment 14 shoulders were dissected, and only one set was used to build the biomechanical shoulder model. The data from the other shoulders were stored for later use.

This shoulder model is used to analyse the kinematic (Pronk, 1991) and dynamic behaviour of the shoulder in order to improve the diagnosis, treatment, and prevention of shoulder disorders (van der Helm, 1994b). The model needs the positions and orientations of the shoulder bones as input, as well as the external forces. Using an optimization scheme, the model predicts the muscle forces for the different muscles.

Quantitative validation of the Dutch Shoulder Model is very difficult: measuring muscle forces of the shoulder in vivo is not possible. Moreover the relation between EMG and muscle force is only partly known (Niemi, 1996; Groot, 1998). From qualitative validation of the model by means of EMG, it was concluded that the model can predict muscle activities for a number of situations: kinematic and dynamic situations (van der Helm, 1994b); timing of muscle activities during fast goal-directed arm movements (Happee and van der Helm, 1995); wheelchair propulsion (van der Helm and Veeger, 1996); principal actions (Groot, 1998).

To get a better agreement between the simulation results of the model and the measured data of a specific person, we must base the model on the geometry of that specific person. One of the methods to obtain geometric model parameters for a specific person is to use medical images. The problem is that it is not possible to determine the muscle architecture by means of MRI, and that the whole process from MRI to model parameters is very complicated and time consuming. In this paper, we propose a method of obtaining muscle attachment contours by transforming geometric surface models of the bones and muscle attachments of a generic shoulder model to geometric surface models of the bones of a specific person.

The specific bone models are obtained from MR images or CT images. They are transformed by means of translation, rotation, and scaling as well as by deformation. Because the muscle attachment contours of the specific shoulder model are transformed together with the surface model of the bone, adapted muscle attachment contours result from transformation. If this method works for muscle attachment contours, it can also be used for other geometrical model parameters, like joint rotation centres. The adapted shoulder model, built by means of these transformed muscle attachment contours, has the muscle architecture of the generic model and the geometry of the measured subject.

In this paper, the following questions are answered:

  • Is it possible to predict the muscle attachment sites of a specific shoulder model using only the geometry of the bones of a specific subject?

  • What is the quality of these predicted muscle attachments?

Section snippets

Methods

In Fig. 1 the transformation process is explained; the generic shoulder model is based on model parameters from cadaver data 1. By means of a geometric model of the bones of cadaver data 1 and a geometric model of the bones of an individual subject, the model parameters are transformed into the subject. From these model parameters, an individualized shoulder model is built.

To validate this method, we used data of three cadavers. From these data, geometrical models of the bones are built, and

Results

After the transformation of the 3-D model of the bones, the measured muscle attachment contours should be located on the surface of these bones. By checking the distance between the mapped muscle attachment contours and their corresponding projection contours on the surface of the bone models, the accuracy of the measurements can be checked. We calculate for each muscle attachment contour the mean distance from the muscle attachment contour points to the surface of the corresponding bone

Discussion

In this paper, we looked at the differences between muscle attachment data from three different cadavers, obtained during one dissection experiment. In most cases, researchers have difficulties enough to obtain one data set, so they have no other measurements they did themselves with which they can compare their data. When they compare their data with measurements from other research groups, they only do so in a qualitative way, because the differences between the different groups in obtaining

Conclusions

By checking the distance between the muscle attachment contours and their corresponding bone surfaces, we have shown that 85 per cent of the muscle attachments were measured correctly. We have also shown that it is possible to predict muscle attachment locations on the basis of the geometry of the bones of a specific subject. To establish this, we had to transform 3-D surface models of these bones into each other. For 45 per cent of the muscle attachments, the prediction of the muscle

References (29)

Cited by (0)

View full text