Poster | 6th Internet World Congress for Biomedical Sciences |
Constantino F. Mugarra Gonzalez(1), Isaac Llorens i Eixea(2), Juan Gutierrez Aguado(3), Eva M. Perez Vidal(4)
(1)(2)(3)(4)Facultat de Fisica. Universitat de Valencia - Burjassot. Spain
[Health Informatics] |
[Medical Electronics & Engineering] |
[Orthopedics & Traumatology] |
[Physical Therapeutics & Rehabilitation] |
Functional studies of human organism require not only static measures, but also an analysis of dynamic behaviour.
Taking as an example vertebral column, the study of its functionality needs something more than static images coming from typical RX. These images are projections on a plane of a three-dimensional structure and as a consequence of this process it can be lost valuable information of the spinal column.
In order to obtain spinal geometry and its functional state, describing interactions between involved anatomical areas, it is necessary a powerful data acquisition system. Such a system would allow obtaining descriptors of position and mobility in spines.
We propose a stereo vision system capable of reconstructing 3-D geometry of vertebral column in personal computers. This is an economically assumable solution because it consists of non-expensive specific hardware.
The main features of the proposed system are:
Being a non-invasive instrument of analysis and measure, without interfering the movements of the person under study, requiring a short time of patient preparation.
Needing short time for acquisition, analysis and process of data.
Obtaining accurate results, with commercial hardware and low computational burden to warrantee a low global cost.
The method proposed has no need of infrared beams or reflectors, hence reducing the complexity of the involved devices. Infrared light sensors are also very sensitive to changes on the reflection angle, leading to increasing problems when measuring dynamic images.
Visible light simplifies the patient preparation, and allows using non-invasive marks (self-adhesive white small labels). Marks are put over selected vertebras on the patient´s back (1), and light used is centred in violet for better image thresholding.
The system consists of two monochrome high-resolution TV cameras (non-interlaced Sony XC55) and a image acquisition card (Matrox Meteor II). Cameras are positioned in such a way that each one covers the whole patient´s back.
As a drawback, cameras need to be calibrated, a slow procedure that it is only needed once. If cameras are fixed it is not necessary to recalibrate them. The method chosen to calibrate is the one proposed by Tsai (2).
When these two images for stereo vision are taken they are thesholded and shrank to obtain the points that better represent the selected vertebras. From this moment on, working on images are not necessary because the image has been reduced to the relevant points, each one representing a vertebral position in each camera projection plane. Correspondence between pairs of points will allow calculating the vertebra positions in real world. In order to reconstruct a smooth vertebral column profile, the points obtained are represented by means of a cubic B-Spline curve. This curve has some desirable properties:
Everything is integrated in a 3D digital environment in a personal computer, allowing medical professionals to develop functional studies and static measures related to possible deviations in vertebral columns. Becoming a useful method to discriminate if patient needs a more exhaustive study by means of invasive techniques.
The following image describe the main steps of a 3-D column reconstruction real case. fig 1.
In order to evaluate the goodness of our method to obtain the 3-D position of the reference points placed on the patient´s back, a previous study was realised to verify the measures quality within the volume of space where the patient would be placed.
This was achieved placing a reference bar with two points separated 500 mm. Three images were taken with three different bar orientations (horizontal, vertical and oblique) within the volume of interest. The real world coordinates were calculated for this points and the results are summarized in the table tab. 1.
The obtained distances for the different positions of the bar are
Horizontal Point: d=499.2437 mm
Vertical Point: d=497.0056 mm
Oblique Point: d=495.1592 mm
The obtained results are very close to real distances. Thus, the experimental estimated error is less than 1%.
The proposed method to study column diseases is easy, quick and accurate enough.
It is non-invasive, so it implies to reduce the radiological doses on people, since the traditional methods of diagnosis are based mainly on RX.
The 3-D column vertebral data obtained with this technique, allows an evaluation of possible pathologies in the spinal column, becoming a useful method to discriminate if patient needs a more exhaustive study by means of invasive techniques.
[Health Informatics] |
[Medical Electronics & Engineering] |
[Orthopedics & Traumatology] |
[Physical Therapeutics & Rehabilitation] |