A Technique for the Use of 3D Surface Imaging to Study Wound Dressing Fixation

The performance of dressings significantly affects wound healing and quality of life for patients.¹ Despite extensive collective nursing care experience, uncertainty remains about the optimum choice of many parameters that affect dressing performance, such as the shape, extensibility, and fixing position of the dressings. The context of this study was an assessment of the potential for stereo-photogrammetry to provide quantitative information to support clinicians and dressing designers. Specifically, this was a pilot study to investigate the influence of skin deformation on the fixing of dressings used in the treatment of a variety of acute and chronic wounds, such as leg ulcers, pressure ulcers, and post-operative wounds. The reference skin surfaces chosen were the upper torso, neck, and legs. The intention was not to conduct a comprehensive, statistically significant trial but to investigate the utility of a novel technique on a range of subjects with varying skin properties. Digital surface photogrammetry techniques were developed to obtain shape information for various sites on the neck and legs of young, middle-aged, and elderly subjects. Subjects entered into the study included volunteers with a range of ages from the Department of Medical Physics and selected patients from the Vascular Unit at the Bradford Teaching Hospitals NHS Foundation Trust (Bradford, Yorkshire, UK). First, the subject was marked using adhesive labels or marker pens to identify a grid of points on the skin around the neck, knee, or ankle. Images were captured using a DSP 400 3D-imaging system manufactured by 3dMD Ltd. (Atlanta, Ga, USA). Positional differences of the grid points were computed when the subject was relaxed and when he or she made typical movements to articulate the joint in the region of interest. Studies were conducted with and without dressings affixed. The data from these positional changes can be used to compute parameters that indicate changes in the shape of the study area. Two major parameters were thought to be significant: skin stretch and skin shear. In this article, the use of skin stretch is examined. Software (3D MedPlot, developed by the Department of Medical Physics, Bradford Teaching Hospitals NHS Trust) was developed to enable rapid, semi-automated assessment, analysis, and display of skin deformation under joint movement. The results were displayed by color-coding the grid to highlight the regions of skin that compressed or stretched. Absolute and relative changes were tabulated alongside the display. Individual arcs in the grid can be selected for closer study. Preliminary studies have highlighted the complex and sometimes counter-intuitive response of skin to joint articulation. In the neck study, it was observed that while skin in the central part of the neck was stretching, there was a region of skin compression over the adjacent shoulder in a direction orthogonal to the direction of stretch. In the ankle study, it was observed that while skin deformed uniformly without a dressing, there was significant wrinkling of the skin with a dressing. These observations suggest some hypotheses as to why the fixation of dressings may fail in practice. Equipment A multicamera system was used to capture the data that define the shape of a surface. This surface data is essentially a list of coordinates (x, y, z) that uniquely identify the positions of points in 3D space relative to an arbitrary datum point (0, 0, 0). The coordinates are preferably expressed in recognized units of length, such as millimeters. The collection of points is known as the point-cloud. A mesh or “wire-frame” image can be drawn by joining pairs of adjacent points using straight lines (Figures 1A and B). A typical surface acquired to represent the vicinity of a wound or joint will have between 5,000 and 20,000 points, depending on the curvatures involved. The separation between the points will be 1–3 mm. For practical reasons, the amount of processing and the amount of data that are stored are limited so the separation between the points, the number of points, and the area of the surface are linked. For the same number of points, smaller separation (ie, improved resolution of surface detail) implies a smaller surface area in the image. There are several technologies available for acquiring surface data. The predominant methods are based on scanning the subject using lasers, or “structured light,” or the use of stereo-photography. In laser scanning, a beam of light is projected onto a surface and viewed obliquely. The beam is scanned across the subject,^2–4 where a flat surface yields an image of a straight line and a curved surface distorts the line. The shape of the surface can be calculated from the distortion using a triangulation technique. A similar technique uses structured light (ie, multiple beams with unique color patterns) to capture a surface without scanning.⁵ This technique has been used to measure wound dimensions for venous ulcers.^6,7 Stereoscopic photography can be implemented in various ways. In one technique, 2 images are simultaneously taken from a pair of cameras rigidly fixed in relation to each other. This was the basis for early 3D photographs. In another technique, 2 or more images are taken using one camera placed in different positions. The camera positions must be accurately recorded when using this technique. In both techniques, the surface shape data can be calculated using triangulation. Although 2 images are sufficient to locate a single point in space, several images are needed to “see” around solid objects or inside “holes” in practice. Stereo-photogrammetry is the technique of using stereoscopic images to obtain measurements of solid objects. Surveyors, archaeologists, architects, and engineers use the technique to record the dimensions and shapes of existing structures and objects. It is also widely used for medical applications.^8,9 The use of photography has the added benefit of “texture maps” that can enhance the appearance of the 3D image in addition to the numerical data that defines the shape of the surface. Essentially, the texture maps are plane photographs that are overlaid onto the representation of the surface and displayed on a computer monitor (Figure 2). However, it should be carefully noted that although the texture maps record a high level of detail—the same as a normal digital image (and limited only by the number of pixels in the light-sensitive component in the camera)—the resolution of the surface is limited by the spacing of points in the point-cloud, which is several orders of magnitude worse than the texture maps. In this study, the data were acquired using an integrated medical imaging system, DSP 400 (Figure 3). The DSP 400 produces 3D surface models from a number of 2-dimensional images that contain information that uniquely identifies points on the surface to be captured. This is achieved by projecting an infrared “speckle pattern” over the subject. The speckle is barely visible to the human eye and gives no discomfort to the patient. Texture images are simultaneously captured using white light. The DSP 400 uses 6 monochrome cameras to capture both surface information and a black-and-white texture map (ie, a black-and-white stereo-photograph). The cameras are arranged in an “H” shape with infrared (IR) cameras at the corners and the monochrome visible light cameras on the central cross bar. The images for all 6 cameras are taken simultaneously using IR and white light projectors. During the processing stage, adjacent pairs of the IR images are used to create stereographic images. These images are then merged together to create a single 3D surface. At the same time, the visible-light images are combined to make a single stereo-monochrome image that can be overlaid onto the surface obtained from the IR cameras. Methods The subjects were marked using adhesive labels or marker pens to identify grids of points on the skin in one of the regions of interest (ie, around the neck, knee, or ankle). These landmarks were spaced about 1.5–2.0 cm apart in an approximately rectangular grid-like pattern (Figures 2 and 4). Some points were uniquely identified for reference purposes. Images were captured using the DSP 400 imaging system. For each subject, a sequence of images was acquired, capturing the positions of the landmarks in various attitudes. For example, neck regions were photographed with the head looking straight ahead, looking down, looking up, looking to each side, and looking both up and down to each side for a total of 9 images. The number of images and the range of positions were determined by the region of interest and the suppleness of the subject. Sequences of images were taken with and without dressings affixed. The “neutral” or “reference” position varied according to the region of interest: • For studies of the neck, the head faced straight ahead • For studies of the knee, the knee subtended approximately 90° between thigh and calf with the weight supported by the foot • For ankle studies, the foot was approximately perpendicular to the lower leg with the weight of the leg supported on the heel. The 3D images were initially viewed in the 3D Patient software package supplied by 3dMD Ltd. The landmarks for each image were manually selected using 3D Patient and the mouse. The positions were exported and saved in text files that could then be imported into 3D MedPlot. Changes in grid point positions were computed with reference to the neutral position. Stretch and shear were calculated from the changes in position in relation to the properties of the lines drawn between the landmarks in a rectangular grid. Figure 4 shows a texture-mapped 3D surface file with a set of landmarks connected in a grid formation. Skin stretch was defined as the percentage change in geodesic surface distance between 2 adjacent landmarks. Skin shear was defined as the change in angle between 2 adjoining lines expressed in degrees. Software^10,11Viewing and manipulating the 3D data. Given the complex nature of the data being manipulated, a specialized user interface and 3D viewing environment was developed. The software allowed for easy translation and independent rotation around each axis. The ability to restrict manipulations to one axis allowed for more precise manipulation either under mouse control or through the keyboard entry of numerical rotations and displacements. Each object displayed on the screen could be scaled and manipulated individually or as part of a group of objects to which the same manipulation matrix was applied. A number of tools were implemented to facilitate the selection, processing, and display of nodes and lines individually and collectively. Measurements. The 2 parameters of interest were skin stretch and skin sheer. Each of these is assessed through comparison of landmark grids from successive images. Polygon n refers to the nth polygon in the list of polygons assembled from adjacent points in the point-cloud. Although the software has the potential to handle any polygon, this study used quadrilaterals exclusively. Skin stretch. Skin stretch is defined as percentage change in geodesic surface distance between 2 landmarks fixed to the skin: 1n = √ {(x₁–x₂)² + (y₁–y₂)² + (z₁–z₂)²} where (x, y, z) denote the 3D coordinates of the path intercept with the edge of each surface polygon and ln is the distance across polygon n. The total path length L is defined as L = Σ 1n and the difference in path length between successive images is defined as d = L₂–L_ref where d is the difference in path length, L₂ is the path length under change, and L_ref is the reference path length. Once landmark data relating to the same set of landmarks over a range of movements were imported, the geodesic distance between each adjacent point on each grid was calculated. By comparing each grid to a user-selected reference grid, the percentage change in distance between landmarks was calculated and assigned a color that was used to shade the grid line on screen. The percentage change color-scale could be normalized to 100% or to the maximum change in distance in that data set. Color-coding the mesh in this way gives a striking visual representation of the movement of skin as joints flex. In the images in Figure 5, lines that stretch are color-coded shades of yellow through red, while lines that compress are color-coded blue through green. Results This was an initial pilot study to define techniques and develop software. It is the intention of the group to extend the technique and to look specifically at the effect of stretch and sheer on the skin focusing on areas at greatest risk of pressure ulceration. The study suggests several avenues of further research. Ten patients (mean age 77 years; minimum 53, maximum 91) and 2 volunteers (age 27 and 53 years)—7 men and 5 women—were imaged using the protocol to allow assessment of skin deformation. Initial imaging showed the shoulder, knee, and ankle to be of specific interest, as these regions exhibit marked skin deformation under joint manipulation and are reported by clinical team members to be sites of common dressing fixation failure. All 3 of these areas were imaged on both volunteers to give an indication of the type and magnitude of deformation that might be expected and to allow for experimentation in developing the protocol and skin marking techniques. Neck study. Figure 5 shows 4 mesh samples taken from successive images of a grid in the right neck and shoulder region of a volunteer as the head was turned from a relaxed central position to full left in equal steps. Figure 4 shows the position and orientation of the landmark mesh on the volunteer. The yellow and red lines running along the long axis of mesh 4 indicate skin stretch as would be intuitively expected. The blue lines running along the short axis in the lower half of the mesh indicate a compressive or bunching effect on skin as it is drawn up and over the shoulder during this movement. Table 1 shows the maximum deformations seen within each mesh. The behavior of healthy skin was thus shown to be somewhat counter-intuitive at times. Knee study. Figure 6 shows the anterior aspect of a volunteer’s left knee. The mesh on the left shows the knee in a relaxed straightened position while the mesh on the right shows it bent to approximately 80°. As expected, lines of high extension can be seen along the long axis of the mesh. However, equal extension can be seen along the short axis of the mesh as skin stretches in a radial pattern from the base of the patella. The skin motion is, in its simplest form, radial from the base of the patella where it meets the femur. The radial motion is symmetrical in the early stage of articulation but becomes complex as the skin stretches over the fibula head and internal tuberosity. The motion becomes more complex as muscular thickening proximal to the joint causes extensive surface shape change toward the end of articulation. The knee poses some of the most complex issues not least in that it demonstrates some of the highest localized stretch and compression of the joints studied here. Also, substantial changes exist in the pattern of skin deformation that are dependent upon anatomical variation as well as physical parameters, such as load bearing. Perhaps one of the more interesting changes is illustrated in Figures 7A and 7B, which show the deformations seen when a relaxed straight leg has a load applied by placing the foot on the floor and applying body weight equally to both legs. The upward movement of the patella occurs alongside uniform compression of skin over the patella with the highest compression occurring along the edge of the vastus medialis (color-coded green in Figure 7B). Compression is also seen along the edge of the vastus lateralis with an associated high stretch region over the bulge of this muscle. Skin stretch is seen along the distal edge of the patella as it is drawn upward with further areas of compression seen laterally across the edge of the patella ligament. Figure 8 shows the stretch image of a relaxed unloaded knee bent to approximately 90°. It is intuitive to assume that the skin over the patella would undergo uniform stretch and that at the side of the knee there would be a region of transition between stretch and compression around the mid plane. The lateral compression over the patella ligament is perhaps less intuitive and the product of the movement of the ligament and the change in the shape of the muscles of the lower leg. The changes of skin distribution around the knee as illustrated in Figures 7 and 8 show the complex patterns of stress and strain that must be accommodated by dressings and highlight the importance of determining suitable fixation points. Ankle study. Figure 9 shows the external aspect of the right ankle of a volunteer. The image is taken with the foot in full extension, and the associated skin stretch expected along the length of the calf can be seen. Posterior to the outer malleolus is a large area in which compressive deformations can be seen. In Figure 9, this region is under smooth compression, which does not cause skin wrinkling. A scaffold applied to the skin may prevent smooth compression and result in wrinkling of the skin. The localized stress caused by the small radii of curvature are thought to be a cause of dressing failure in this region as well as the cause of increased stress and trauma to the skin. Figure 10 shows color images (taken with a standard digital camera) to illustrate the effect. The images show the same region of the ankle, with the foot extended by the same amount in each case. In Figure 10A, the skin compresses smoothly with no wrinkling; in Figure 10B, the compressive effect induces wrinkling. Patient data. The study encountered many problems in imaging patients. It proved difficult to find an acceptable method of marking the skin of some patients with fragile skin—in some cases patients experienced pain with the marking method. No action was taken that might have incurred a risk of infection. Further, many patients found flexing their joints to be difficult. Nevertheless, many images were acquired and processed. One general observation is that the amount of local deformation (stretch/compression) was significantly less in the patient group than in the volunteers. Table 5 shows the results for an instance where the flex angle was the same for one of the patients and one of the volunteers. This pair of experiments is interesting because the gross degree of movement was the same, but the apparent skin deformations are quantifiably different. The authors suggest that the larger deformation of the more elastic skin of the volunteer is concentrated in the ankle, whereas the lesser deformation of the patient’s skin is distributed over a greater area, extending in some cases up to the knee. If this is found to be generally applicable, it is a phenomenon that may need to be considered when developing guidelines for the fixation of dressings. These changes may reflect the effect of lipodermatosclerosis and tissue edema, recognized complications of chronic venous insufficiency, on skin mobility.^12,13 Discussion and Conclusion The purpose of the study was to investigate the potential of stereo-photogrammetry to provide quantitative information to support clinical and industrial work in the design and use of wound dressings. In this respect, the technique has been shown to offer a powerful, new approach to understanding the interaction between a dressing and the region around a wound. It clearly offers a method of measurement that complements study of the physical properties of tissue.¹⁴ Some observations were expected, such as the observation that the skin of younger, healthy subjects is elastic and moves in such a way that smooth extension and compression occurs around joints as they flex, whereas in the elderly subjects, especially those with edema and/or ulceration, there is little discernible movement of the skin. However, some observations were unexpected, such as the way skin on the neck and shoulder extends in one direction while at the same time compressing in the orthogonal direction or the way in which skin behind the foot appears to compress smoothly but wrinkles when constrained by a dressing. It is also clear that fixed dressings alter the patterns of stretch/compression and shear. Quantitative observations of this phenomenon affect the design of dressings and the way in which they are applied and fixed to the patient. Hopefully, further study will help explain why some dressings result in blistering of the skin adjacent to the wound. It could be said that the insights offered by the use of the technique could have been obtained in different ways. However, the technique does give quantitative evidence where in the past the indications have been qualitative and often anecdotal. It also provides concrete evidence, in the form of accurate 3D images, that can be used for issues relating to consent and clinical governance. The study suggests avenues of further research and has provided many clues for future development of the technique. For example, it is clear that the extent of skin deformation in most patients is significantly less than in the younger volunteers. Furthermore, it proved difficult to find a method of marking the skin of some patients that is acceptable on both safety and cosmetic grounds. Therefore, in the interest of improved resolution and patient compliance, it is important that in future studies a technique is developed that uses the natural pattern and texture of the skin. This has the potential to dramatically increase the volume of data available, and it will be challenging to display the results in a meaningful way. Several technical developments are also expected, such as new indices of elasticity and novel ways of presenting information about the performance and behavior of skin near a wound. Based on the experience the authors gained using the methods described, it is also expected that natural skin marks will be essential in studies that include regions of the skin that are ulcerated. Imaging patients’ skin while they receive compression therapy is not possible using this technique. The authors accept that compression can cause skin damage, though when compression causes damage, the damage tends to occur over bony prominences, such as the malleoli or the tibial cress, not in relation to the wound or dressing. Damage in this area is usually due to failure to manage exudate, skin maceration, infection, or product sensitivity. To investigate the effect of compression in this situation, a transparent compression system would need to be developed to allow imaging of the skin surface. A future project will examine whether the lines of relaxed skin tension used by surgeons for optimal incisions (eg, Langer’s lines) can be imaged using the aforementioned technique to further investigate the effect of disease, age, and scarring on skin movement patterns. An obvious longer term project involves the design of dressings that not only accommodate the essential curvature of body parts but also tolerate the change in shape occasioned by normal movements of the limbs, head, and neck.