Abstract: The aim of the study was to demonstrate the ability of high-resolution ultrasonography (HRU) to image crusted canine wounds and compare 2 methods of evaluating wound area. The dorsal thoracolumbar areas of 10 beagle dogs were clipped, and 2 rows of 6 full-thickness punch 5mm biopsy samples were taken from either side of the midline under general anesthesia. Wounds were allowed to heal by secondary intention, while treatment gel containing an aliamide mast cell modulator, adelmidrol, or gel base was applied twice daily to randomly selected treatment and control groups. Concentric 8mm punch biopsy samples of the healing wounds were taken on Days 1, 2, 4, 8, and 14 for histopathological assessment. Wounds were imaged until the second biopsy with 20MHz HRU daily for 28 days. High-resolution ultrasonography proved to be quick and easy to use on canine skin and generated images of the dermis beneath crusts and healing epidermis suitable for analysis. Wound morphometry was performed on the digital HRU images (diameter and depth of wound; number of pixels in the wound area). The measurements were used for the calculation of wound area. A Bland-Altman plot was constructed. Least products regression was used to test for the presence of fixed and proportional bias between the 2 methods. There was an increase in the difference between the 2 methods as the measured area increased. Visual appraisal of the Bland-Altman plot suggested the presence of proportional bias, and this was confirmed using least products regression. Repeated measures analysis of variance identified a significant change in wound area over time for both methods but identified a significant difference between treated and control animals for the pixel method only. Both methods appear suitable for assessment of wound area over time, but the pixel method appears to be more accurate.

Disclosure: Funding for this work was provided by a grant scheme from The Royal Veterinary College, University of London, Hatfield, United Kingdom, and by Innovet Italia s.r.l., Rubano, Italy

T

he evaluation of wound healing has proved to be a challenge in the clinical arena both in animals and in humans. Goldman and Salcido reviewed methods for assessment of wound area and volume.¹ Wound tracing and planimetry can provide consistent estimates of wound surface area, while digital photography and image analysis can provide estimates of wound area and allow assessment of changes in color at the surface of the wound and in the surrounding tissues. Wound volume assessment provides a greater picture of the healing process. It requires measurement of wound depth and calculation of volume based on assumptions of wound cavity shape, unless the volume is determined by filling the wound with a substance, such as alginate,¹ which provides a mold that can be measured, or by filling the wound with a known volume of saline. Specialized laser beam technology has also been developed to allow 3-dimensional imaging and calculation of wound volume,² but this depends on access of the light beam to all regions of the wound that are to be measured. However, volume assessment by these techniques requires interference with the healing tissue, and none of these methods can be applied when an eschar (crust) has formed or when the epidermis has regenerated sufficiently to cover the healing wound.³
       Ultrasonography has been developed for the assessment of wound healing, because it can provide quantitative data on the healing process deep within the tissue without being invasive.³ Initially, low-frequency transducers allowed only macroscopic imaging, but in the 1980s, experiments with ultrasound frequency higher than 15MHz showed that subsurface planes of living tissue could be visualized at microscopic resolution in a 2-dimensional image.⁴ The development of even higher frequency ultrasound equipment, typically higher than 20MHz, allowed visualization of the living epidermis, dermis, hypodermis, and deep fascia at a microscopic level. As a result, the term ultrasound biomicroscopy was introduced.⁴ A strong correlation between features identified on high-resolution ultrasound biomicroscopy (HRU) and features visualized using histology has been identified.⁵ High-resolution ultrasound biomicroscopy has been used to monitor healing in acute⁶ and chronic wounds,⁷ visualize skin structures,⁸ measure skin thickness,^9,10 differentiate dermal burns,¹¹ and document and assess postradiation skin reactions in breast cancer.¹²
       High-resolution ultrasound biomicroscopy has not been evaluated in the assessment of wounds in the dog. The purpose of this report is to demonstrate its ability to image crusted cutaneous wounds in canine skin and to compare 2 methods of measurement of wound area using HRU images.

Materials and Methods

       The study formed part of a wound treatment evaluation that compared healing of wounds treated either with an active gel containing the aliamide, adelmidrol (a downregulator of mast cell degranulation), or the gel base alone. Treatment effects will be published separately. The authors obtained ethical and regulatory approval from the Royal Veterinary College Ethics Committee and United Kingdom Home Office. Ten healthy beagle dogs (8 males and 2 females) of various ages (ranging from 2 to 9 years) were used. The animals were housed in individual kennels with daily access to outdoor runs and were monitored regularly; they were fed Chappie Complete dry diet (Waltham, Leicester, United Kingdom). The dorsal thoracolumbar areas of the dogs were clipped and prepared aseptically. Six pairs of full-thickness skin wounds were created in the clipped area under general anaesthesia using 5mm biopsy punches. General anaesthesia was achieved by intravenous administration of propofol (Rapinovet, Schering Plough, Welwyn Garden City, United Kingdom) in a bolus of 6mg/kg for induction, with maintenance at a dose of 0.1–0.4mg/kg/min; the dogs were intubated during the anesthetic period. The wounds were placed 6cm on either side of the dorsal midline and were approximately 3cm apart in each row, extending from just caudal to the scapulae to the mid-lumbar region. For each animal, wounds on 1 side were randomly allocated to receive the active wound healing gel, while those on the other side received the gel base only. The gels were applied twice daily. Wounds were not bandaged but were allowed to form crusts and heal by secondary intention. Crusts were not removed, and dogs wore Elizabethan collars to prevent self-trauma. Clavulanate-potentiated amoxicillin (Synulox®, Pfizer, New York, NY) was administered twice daily to all dogs by mouth until wound healing was complete. Single, 8mm punch biopsy samples centered on the healing wounds were taken from sites chosen at random on left and right sides on Days 1, 2, 4, 8, and 14 for histopathological assessment (reported elsewhere); these wounds were then sutured and not evaluated further. The 2 wounds that were not biopsied for a second time were examined throughout the study.
       The 20MHz portable ultrasound scanner used (Longport Digital Scanner [LDS1], Longport International Ltd., Silchester, United Kingdom) is fitted with a polyvinylidene difluoride transducer incorporated into a probe filled with distilled water and scanned using a digital stepping motor. The ultrasonic beam is propagated through an aperture covered with a disposable rubber membrane; a new membrane was used for each wound. The digitized scans were stored on the associated hard drive and were visualized using a color palette. Scans were taken through the center of the wound bed and the adjacent intact skin. Wounds were imaged daily, longitudinally (parallel to the spine) and transversely (perpendicular to the spine), for 28 days by applying the transducer to the wound area using light pressure and the wound gel (active formulation or base) as transmission medium. All wounds were imaged until the second biopsy procedure was completed.
       Wound morphometry was performed on the digital HRU images. The diameters at the top (level of epidermis), bottom, and middle (mid-point between visible top and bottom areas) and length (distanced from top to bottom) of the wound were measured on both the longitudinal and transverse images (width-length method) (Figure 1A). Wound area was calculated as image length multiplied by the average of the 3 width measurements. The circumference of each HRU wound image was also traced, and the wound area was calculated by the software based on the number of pixels enclosed within the circumference (pixel method) (Figure 1B).
       Owing to equipment malfunction, no data were analyzed for the second through sixth days of the study after the creation of the wounds.

Figure 1

Statistical Analysis

       Bland-Altman plots were constructed to compare the difference in wound area against the average wound area as measured by the 2 approaches (pixel versus width-length).¹³ Least products regression was used to test for the presence of fixed and proportional bias between the 2 methods of measurement.¹⁴ Fixed bias refers to 1 method giving values that are higher or lower than the other by a constant amount, while proportional bias means that 1 method gives values that are higher or lower than those of the other by an amount that is proportional to the level of the measured variable.¹⁴ Repeated measures analysis of variance (ANOVA) was used to assess whether there was a significant difference in wound area over time and between control and treated animals for both the pixel and width-length methods. Repeated measures ANOVA was also used to assess whether wound area differed significantly between transverse and longitudinal images. All statistical analyses were conducted using the software SAS 9.1 for Windows (SAS Institute Inc., Cary, NC) and SPSS 12 for Windows (SPSS Inc., Chicago, Ill).

Results

       High-resolution ultrasound provided images throughout the skin allowing differentiation of the wound and surrounding tissue as wound healing occurred (Figure 2) and the formation of granulation tissue within the healing wound (Figure 3). Even when crusting obscured the exact position of small healing wounds, they could be readily visualized by HRU. At the end of the study, dermal healing was still incomplete.

Figure 3

Figure 2

       A Bland-Altman plot was constructed to compare the difference in wound area against the average wound area as measured by the 2 approaches (Figure 4A). In general, the pixel method recorded higher wound areas than the width-length method with a mean difference of -0.7mm2 (SD 5.80) and 95% limits of agreement of 10.90mm² and -12.30mm². Reduced major axis regression confirmed the presence of both fixed bias [a’ = -15.13 (-16.38, -13.87); confidence intervals not spanning 1], and proportional bias [b’ = 0.85 (0.80, 0.91); confidence intervals not spanning 0].

Figure 4A

       Repeated measures ANOVA identified a significant decrease in wound area over the 27 days (p<0.001) for both the width-length and pixel methods. However, a significant difference between control and treated wounds was only detected when the pixel method was used to calculate wound area (p<0.001) and not when the width-length method was used (p=0.163). In general, the pixel method recorded higher wound areas than the width-length method for both control and treated animals (Figures 4B and 4C). For both methods, wound area varied over the experimental period but displayed a decrease over the 27 days. The pixel method identified an increase in wound area between Days 8 and 18, while the width-length method showed the wound area to decrease during this period.
       Repeated measures ANOVA identified a significant difference in wound area between the transverse and longitudinal wounds for the width-length method (p=0.032) but not for the pixel method (p=0.149).

Figure 4B

Figure 4C

       Mean width of the top, middle, and bottom diameters all differed significantly from each other (p<0.001). On average, the wound was widest at the mid-point (6.76 +/- 1.75mm) and narrowest at the bottom (5.63 +/- 1.77mm). Mean diameter at the top of the wound was 6.17 +/- 1.70mm. Although top, middle, and bottom diameters were very similar initially, from Day 15 onward, the bottom diameter appeared to decrease more rapidly than the top measurement (Figure 5) due to the deposition of granulation tissue in the healing wound.

Figure 5

Discussion

       This study demonstrated the ability of HRU to penetrate through eschar and to illustrate the dimensions of healing wounds within the canine dermis. The method allowed clear differentiation of the wound cavity from the surrounding structures. As in human skin, it permitted the identification of collagen laid down in granulation tissue during remodeling as zones of increased reflection in the healing dermis.³ Even small wounds could be readily found, imaged, and measured accurately. Scanning did not appear to damage the healing wounds. The HRU apparatus was easy to set up and operate, and the time required to scan was usually less than a minute for each wound.
       Both visual appraisal of the Bland-Altman plot and least products regression confirmed the presence of proportional bias. In other words, the difference in area between the methods of wound area measurement was greatest on large wounds and smallest on small wounds. This finding may be due to the fact that smaller wounds tended to be more clearly defined than large wounds and, therefore, easier to measure accurately. The mean difference between the 2 methods was 0.7mm³ with the 95% limits of agreement indicating that the pixel method might overestimate wound area by as much as 10.9mm³ or underestimate it by up to 12.3mm³.
       The repeated measures ANOVA detected a significant change in wound volume over time for both methods, yet only the pixel method identified a significant difference between treated and control animals. This could be attributed to the fact that the pixel method appeared to be more accurate than the width-length method and was, therefore, more likely to detect small differences, such as those that may exist between groups of animals.
       Given the irregular shapes of the wounds, calculation of wound volume based on assumptions of wound shape was likely to introduce error. However, scanning wounds in 2 directions perpendicular to each another produces useful data on the shape of each wound. Algorithms that can use the information generated by multiple scans to estimate wound volume need to be created. This is likely to enable wounds and other lesions within the skin to be assessed with increasing accuracy. This may be of special interest in the assessment of deep chronic wounds. Evidence has shown that HRU can differentiate between healing and nonhealing wounds.⁴ A recent study concluded that HRU scanning permits the quantitative assessment of structural changes deep within the wound and that temporal changes in the width of the wound base can be used as an indication of the progress of repair.³
       Bearing in mind the limitations of the study, both methods of wound area calculation appeared to be useful indicators of wound healing over time. The 2 methods should not be used interchangeably due to the difference in the measured area found using both methods. Possibly due to its greater accuracy, the pixel method appeared to be more sensitive than the width-length method and was more likely to detect small differences, such as those that may exist between groups of animals. The pixel method would appear to be the preferred method of calculating wound area.

Acknowledgements

       The authors thank Longport International for loaning the HRU scanner used in this study and Mr. Paul Wilson for the technical help provided.