Thermography is a noninvasive diagnostic imaging technique that involves the recording of cutaneous thermal patterns generated by the emission of surface heat; the patterns are reported in the form of a color map. Surface heat measured from the skin is directly related to the local dermal microcirculation, which is directly controlled by the sympathetic portion of the autonomic nervous system. Only the local dermal microcirculation is responsible for the surface heat generated. Heat is not conducted from deeper portions of the body to the surface and thus does not create changes in the surface temperature. The clinical basis for thermography is the correlation of temperature recordings with various conditions of disease or injury as they relate to autonomic function.1,2 Thermography can be used as a diagnostic screening tool, as an adjunct test to enhance interpretation of physical examination findings, to guide therapeutic management, and to assess long-term response to treatment.1–5
Early thermographic imaging systems were developed in the 1940s, introduced to the medical community in 1959, and used to assess arthritic joints in humans. Obtaining those thermographic images required several minutes, and the temperature could not be quantified.6 Over the years, the equipment and technology have become more sophisticated. In the 1960s, color images were introduced, and in 1965, thermal imaging was used in equine medicine.7 In the 1970s, computers were used to provide color images and to perform image analysis and data storage. Current thermographic imaging systems use focal plane array detectors (with high-speed images and spatial resolution6) and image recognition software for objective analysis.8 These technologic advances may make thermography a more acceptable modality for a variety of medical applications in human and veterinary medicine.
Thermography has been used in human medicine for assessment of breast cancer8,9; cutaneous evaluation of burn patients8; rheumatologic assessments6; evaluation of vascular disorders,10 including Raynaud's phenomenon,11,12 scrotal varicocele,13 pulpal blood flow of teeth,14,15 pneumothorax,16 inflammatory conditions,17 radiculopathies,18 and intervertebral disk disease19; assessment of Chiari malformation and syringomyelia20,21; and assessment of joint disease.22,23 It has also been used as an adjunct for polygraph testing.24 In horses, thermography has been used in lameness evaluations3 and more specifically to detect ligamentous, osseous, muscular, articular, and neurologic injuries.3–5,25–27
Because of the technologic advances and lack of sedation needed for imaging, thermography has potential use as a screening test for a number of conditions in veterinary patients. Normal thermographic patterns in dogs have not been determined, nor have the limitations of imaging through an intact coat been studied. Additionally, the effect of clipping a dog's coat on surface skin temperature and thermographic patterns has not been determined. The coat has been implicated as a cause of difficulties for interpretation of thermographic data in horses,7,25,28,29 but in 1 study,28 clipping did not change the overall thermographic pattern on the limbs of horses. Therefore, the purpose of the study reported here was to describe a thermographic imaging protocol, identify normal thermographic patterns for various ROIs of the limbs of healthy dogs, and determine the effect that clipping the coat has on the thermographic pattern and limb temperature of healthy dogs.
Materials and Methods
Animals—Ten Labrador Retrievers from a training-breeding programa were used. All dogs were sexually intact females, and the median age was 13 months (range, 11 to 15 months). All dogs were evaluated via a CBC count, serum biochemical analysis, radiography of all limbs, and orthopedic evaluation conducted by a board-certified surgeon (DJM), and all were found to be clinically normal. All of the laboratory and radiographic evaluations and physical examinations were performed 2 days prior to thermal imaging to acclimate the dogs and minimize possible effects on the sympathetic nervous system that could change thermal patterns. During the 3 days of testing and imaging, the dogs were confined to temperature-controlled runs (21°C) and allowed only limited activity.
Thermal imaging—Thermal images were obtained with a stand-mounted infrared camerab with a focal plane array amorphous silicone microbolometer. The camera was connected to a laptop computer for real-time data analysis.
Images were obtained of each dog, with views determined by the authors, in the same room with ambient temperature controlled at 21°C by a centralized climate control system. It was not necessary to further acclimate the dogs to the imaging room temperature because the dogs were housed at the same temperature. Two technicians wore latex gloves and used the head and tail when positioning the dogs for imaging to minimize thermal artifacts from manual contact. The dogs were placed in front of a uniform interior wall to minimize background artifact that may have been created by temperature differences in exterior walls. The camera was positioned approximately 1.5 to 4.6 m from the dogs, depending on the ROI. The protocol for imaging included cranial and caudal views of the body, full left and right lateral body views, and full-limb views of all limbs (Figures 1 and 2). Additionally, images of each limb were obtained to provide views of each joint region (Figure 3).
Illustration of standard positioning of a dog for a full-body thermographic study with an infrared camera via cranial, caudal, and lateral views. Cr = Cranial. Ca = Caudal. LFL = Left forelimb. RFL = Right forelimb. LHL = Left hind limb. RHL = Right hind limb.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Illustration of standard positioning of a dog for full-limb thermography. See Figure 1 for key.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Illustration of standard positioning of a dog for thermography of various ROIs in the forelimb and hind limb. See Figure 1 for key.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
After images were obtained of the dogs with intact coats, the coat was clipped on the forelimbs and hind limbs only with a No. 40 clipper blade in a surgical preparation pattern. Images of each dog were then obtained by use of the same limb-imaging protocol 15 minutes, 60 minutes, and 24 hours after clipping.
Each image was saved within the software programc for further evaluation and review. The program was preset for a temperature range of 8°C with a 16-shade color map. The color map indicated warmer temperatures as white and red and cooler temperatures as blue and black. Each ROI was focused by the operator of the camera, and an image was saved by the operator of the computer. After the image was saved, the temperature scale was adjusted so that the color range of the image was balanced and diagnostically useful. The program saved the final image with the specific 8°C temperature range for each specific image.
Image temperature analysis—Each thermal image was assessed for ranges of temperatures within a given ROI. The region of the shoulder was defined as mid-scapula to the distal third of the humerus and the cranial aspect of the humeral head to the level of the fourth rib. The region of the elbow was defined as the distal third of the humerus to the proximal third of the antebrachium and the cranial aspect of the antebrachium to the caudal edge of the olecranon. The region of the carpus was defined as the distal third of the antebrachium to the digits and the cranial aspect of the distal portion of the antebrachium to the caudal aspect of the accessory carpal pad. The region of the hip was defined as the dorsal aspect of the ilium to the midfemur and the cranial aspect of the ilium to the base of the tail. The region of the stifle was defined as the distal third of the femur to the proximal third of the tibia and the cranial aspect of the patella to the caudal aspect of the gastrocnemius muscle. The region of the tarsus was defined as the distal third of the tibia to the digits and the cranial aspect of the distal portion of the tibia to the caudal aspect of the calcaneus. By use of the program,c the maximum, mean, and minimum temperatures within each ROI were calculated. All temperatures were recorded and analyzed statistically. Temperature patterns were evaluated and analyzed by use of image-processing softwared for recognition of patterns.
Statistical analysis—To determine variability of temperatures of each ROI among the 10 dogs at each of the 4 time points (intact coat and 15 minutes, 60 minutes, and 24 hours after clipping), the mean, SD, and CV were used. To determine whether, for a given ROI, the means varied among the 4 time points, a mixed-model repeated-measures ANOVA was used. Pairwise multiple comparisons among time points were performed by use of the Tukey-Kramer adjustment method. Paired t tests were used to compare corresponding ROIs of the left and right sides to determine whether mean temperatures were the same in both limbs for the 30 site-time combinations. To address the problem of spurious false-positive results when performing numerous (ie, 120) statistical tests, the Benjamini-Hochberg method for controlling the false discovery rate was used to compute adjusted P values. An adjusted value of P < 0.05 was considered significant.
Image pattern analysis—An image-processing software programc was used to investigate and analyze images. The program included a collection of computer vision and image-processing routines. The library routines were distributed in a common-object–model, dynamically linked library and could be used in custom-user applications independent of the program.c The program was used to explore algorithm development before custom applications were created.
For this study, 3 custom programsd were written by use of the C# language. A Windows-based program was written to automate the process of creating the mask images that were used to select an ROI from input images. For each input image, a corresponding mask image was created to indicate the ROI for that image. Another program was written to automate the process of running the classification experiments. Results of feature extractions were cached because of the time required to calculate each of the features. The output from the program was a file that contained a table with the requested data. This program could also be used to select and load a group of experiments from a directory, modify some of the experimental variables, and reanalyze all the experiments by use of the new variables. These results could then be compared with previous results to determine the effects of the modifications.
The primary tools used included analysis features and pattern classification. The features considered included thinness, aspect ratio, rst-invariants, spectral (Fourier) transformations, histograms, and texture features. Histogram features included mean, SD, skew, energy, and entropy, whereas texture features included energy, inertia, correlation, inverse difference, and entropy. Pattern classification tools included various standardization methods, distance and similarity measurements, and 3 classification methods.
For image pattern analysis, 3 sets of analyses were performed for a total of 229 experiments. Success was defined as the percentage of the class being evaluated (ie, lateral body, cranial body, forelimb, and hind limb) found by the algorithm that reflected the correct class.
Results
For each ROI within a category (ie, intact coat and 15 minutes, 60 minutes, and 24 hours after clipping), mean temperatures among the 10 dogs were similar, which correlated with the CVs determined for each ROI. The CV ranged from approximately 1% to 3% for the mean temperatures.
Values obtained for the intact coat differed significantly from each of the 3 time points after clipping for each of the limb regions (Table 1). In some instances (left and right carpus and left and right tarsus), the value for the 24-hour time point differed from the value for the 60-minute time point. No differences in temperature were detected between 15- and 60-minute values or between most of the 60-minute and 24-hour values; a significant difference was detected between 15-minute and 24-hour values for the left shoulder and elbow. No differences between 60-minute and 24-hour values were detected for the right shoulder and elbow, left and right hip, and right stifle. None of the left side versus right side differences were significant.
Mean ± SD temperature (°C) for various ROIs in 10 healthy dogs obtained with an intact coat and 15 minutes, 60 minutes, and 24 hours after the coat was clipped.
For grouping of thermal patterns for similar regions of each dog (ie, right and left forelimbs), the success rate was 40%. This low result may have been attributable to a higher correlation among the same ROI of all dogs than within a particular dog or because of small sample size. When images were allocated into 4 classes (lateral body, cranial body, entire forelimb, and entire hind limb), the mean success rate was 90%, which indicated the successful separation and identification of ROIs.
For the effect of clipping, mean success rate was 75% for use of image pattern analysis for recognition of similar thermographic patterns in the forelimbs and hind limbs. Success rate for the hind limbs was 90% and for the forelimbs was 60%. Consistent patterns were evident for each dog; for left side, compared with right side; and among dogs (Figures 4 to 7).
Thermographic image of the right hind limb of a healthy dog. The inguinal region (blue arrow) and saphenous vein (white arrow) are indicated.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Caudal thermographic view of a healthy dog. The medial saphenous vein (white arrow) and anal region (black arrow) are indicated.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Cranial thermographic view of a healthy dog. The periocular region and muzzle (black arrows) and cephalic vein (white arrow) are indicated.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Lateral thermographic image of the forelimbs of a healthy dog. The omobrachial vein (blue arrows) is indicated on the left forelimb, and the accessory cephalic vein (white arrow) is indicated on the right forelimb.
Citation: American Journal of Veterinary Research 68, 10; 10.2460/ajvr.68.10.1064
Discussion
The first objective of the study reported here was to describe a thermographic imaging protocol. Views for the dogs were based on established views determined for horses.1 The temperature of the room was based on the ambient temperature commonly found in medical facilities and used in various human and equine studies and was maintained constant for the duration of imaging for optimum results.1–3,30,31
Another objective of the study was to identify normal thermographic patterns for various ROIs in dogs. Analysis of results indicated that there were similar thermographic patterns and mean temperatures among dogs in each ROI and between left versus right sides. The adjusted P value for all ROIs was < 0.05, and the image pattern analysis for ROI was 90%, which confirmed the similarity of thermographic patterns and temperatures among dogs. As revealed in research of equine and human thermographic patterns,1,2,4,5 dogs also appeared to have distinct vascular patterns associated with normal vasculature in the limbs.
Another objective of the study was to determine whether there was symmetry in the thermographic patterns and mean temperatures between the left and right side of each dog. Results indicated there was such symmetry. In equine studies,1,4,30 bilateral symmetry of thermographic patterns is described as being similar. Variations were caused by variations in tissue vascularization and metabolism among horses.4 In the study reported here, dogs also had similar thermographic patterns. These findings suggested that when evaluating patients with unilateral disease for altered thermographic patterns, the contralateral ROI may be used as a control image for comparison purposes.
A third objective of the study was to determine the effect of clipping the coat on the thermographic pattern and skin temperature. In some equine studies,7,29 it has been necessary to clip the coat to obtain reliable thermal images, whereas results of other studies25,28 have not supported this finding. In 2 studies,25,28 investigators determined that clipping the coat of horses was not necessary to achieve a reliable thermal image; however, the clipped limbs were warmer than the unclipped limbs. In the study reported here, there was a significant difference in mean temperature between the regions when there was an intact coat and the regions after the coat was clipped at all time points, but the thermographic patterns remained similar. There was a significant difference between the mean temperatures of most ROIs at 15 and 60 minutes and 24 hours after clipping, when compared with the mean temperature for the unclipped state. In most instances, the difference between mean temperatures at 60 minutes and 24 hours was not significantly different in the various ROIs. Therefore, when clipping of the coat over an ROI is performed, waiting a minimum of 60 minutes is necessary for stable temperature readings; however, thermal imaging patterns can be consistently determined as early as 15 minutes after clipping.
Differences were detected in the temperatures of the ROIs in the distal portions of the limbs from 60 minutes to 24 hours after clipping. Specifically, there was a significantly lower mean temperature for the distal portions of the limbs at 60 minutes, compared with the mean temperature at 24 hours. This may be attributable to the greater surface area–to–tissue density ratio in the distal portion of the limb, compared with more muscular regions of the limb, which could make the distal portion of the limb more vulnerable to influences of environmental temperature, especially when the insulating effect of the coat has been eliminated. These findings suggested that when an ROI located in the distal portion of the limb is imaged, it is necessary to wait 60 minutes to determine accurate mean temperatures; however, the effect of delaying imaging more than 24 hours on mean temperature values is unknown.
Image pattern analysis revealed that for the effect of clipping the coat, there was a mean success rate of 75% for use of image pattern analysis for recognition of similar thermographic patterns in the forelimbs and hind limbs. Success for the hind limb analysis was high (90%) and indicated that there was consistency of the thermal imaging patterns. The lower success rate for the forelimb analysis (60%) indicated that the images may not have been consistent. This difficulty may be resolved by further standardization in the image-acquisition process. The hind limb analysis indicated similarity of patterns for the clipped and unclipped state. These findings suggested that clipping the coat in healthy dogs increased the measured skin temperature but did not change the thermal pattern. It is unknown whether clipping affects the thermographic image patterns in dogs with abnormalities.
ABBREVIATIONS
| ROI | Region of interest |
| CV | Coefficient of variation |
Guide Dog Foundation for the Blind, Smithtown, NY.
Med 2000 IRIS, Meditherm Inc, Beaufort, NC.
CVIPtools, Computer Vision and Image Processing Laboratory, Department of Electrical and Computer Engineering, School of Engineering, Southern Illinois University, Edwardsville, Ill. Available at: www.ee.siue.edu/CVIPtools. Accessed on Nov 11, 2006.
Developed by Patrick S. Solt, Department of Electrical and Computer Engineering, School of Engineering, Southern Illinois University, Edwardsville, Ill.
References
- 2.
Meditherm manual of clinical thermology. Beaufort, NC: Meditherm Inc, 2004;4–6.
- 4.↑
Purohit RC, McCoy MD. Thermography in the diagnosis of inflammatory processes in the horse. Am J Vet Res 1980;41:1167–1174.
- 5.
Vaden MF, Purohit RC, McCoy MD, et al. Thermography: a technique for subclinical diagnosis of osteoarthritis. Am J Vet Res 1980;41:1175–1179.
- 6.↑
Ring EFJ. The historical development of thermal imaging in medicine. Rheumatology 2004;43:800–802.
- 8.↑
Head JF, Elliot RL. Infrared imaging: making progress in fulfilling its medical promise. IEEE Eng Med Biol Mag 2002;21:80–85.
- 9.
Gautherie M, Haehnel P, Walter J, et al. Thermovascular changes associated with in situ and minimal breast cancers. J Reprod Med 1987;32:833–842.
- 10.↑
Verheye S, DeMeyer GRY, Krams R, et al. Intravascular thermography: immediate functional and morphological vascular findings. Eur Heart J 2004;25:158–165.
- 11.
Merla A, Di Donato L, Di Luzia S, et al. Infrared functional imaging applied to Raynaud's phenomenon. IEEE Eng Med Biol Mag 2002;21:73–79.
- 12.
Clark S, Dunn G, Moore T, et al. Comparison of thermography and laser Doppler imaging in the assessment of Raynaud's phenomenon. Microvasc Res 2003;66:73–76.
- 13.↑
Tucker AT. Infrared thermographic assessment of the human scrotum. Fertil Steril 2000;74:802–803.
- 14.
Kells BE, Kennedy JG, Biagioni PA, et al. Computerized infrared thermographic imaging and pulpal blood flow: part 2. Rewarming of healthy human teeth following a controlled cold stimulus. Int Endod J 2000;33:448–462.
- 15.
Kells BE, Kennedy JG, Biagioni PA, et al. Computerized infrared thermographic imaging and pulpal blood flow: part 1. A protocol for thermal imaging of human teeth. Int Endod J 2000;33:442–447.
- 16.↑
Rich PB, Dulabon GR, Douillet CD, et al. Infrared thermography: a rapid, portable, and accurate technique to detect experimental pneumothorax. J Surg Res 2004;120:163–170.
- 17.↑
Venta I, Hyrkas T, Paakkari I, et al. Thermographic imaging of postoperative inflammation modified by anti-inflammatory pretreatment. J Oral Maxillofac Surg 2001;59:145–150.
- 18.↑
So YT, Aminoff MJ, Olney RK. The role of thermography in the evaluation of lumbosacral radiculopathy. Neurology 1989;39:1154–1158.
- 19.↑
Doman I, Illes T. Thermal analysis of the human intervertebral disc. J Biochem Biophys Methods 2004;61:207–214.
- 20.
Koyanagi I, Iwasaki Y, Isu T, et al. Thermographic findings of syringomyelia. No Shinkei Geka 1988;16:1149–1154.
- 21.
Fujimoto A, Matsumura A, Nakamura K, et al. Chiari malformation type I associated with familial spastic paraplegia: report of a surgically treated case. Childs Nerv Syst 2005;21:336–338.
- 22.
Gratt BM, Sickles EA, Wexler CE. Thermographic characterization of osteoarthrosis of the temporomandibular joint. J Orofacial Pain 1993;7:345–353.
- 23.
Gratt BM, Sickles EA, Wexler CE, et al. Thermographic characterization of internal derangement of the temporomandibular joint. J Orofacial Pain 1994;8:197–206.
- 24.↑
Pavlidis I, Levine J. Thermal image analysis for polygraph testing. IEEE Eng Med Biol Mag 2002;21:56–63.
- 25.
Tunley BV, Henson FMD. Reliability and repeatability of thermographic examination and the normal thermographic image of the thoracolumbar region in the horse. Equine Vet J 2004;36:306–312.
- 26.
Eddy AL, Van Hoogmoed LM, Snyder JR. The role of thermography in the management of equine lameness. Vet J 2001;162:172–181.
- 27.
Purohit RC, McCoy MD, Bergfeld WA. Thermographic diagnosis of Horner's syndrome in the horse. Am J Vet Res 1980;41:1180–1182.
- 28.↑
Turner TA, Fessler JF, Lamp M, et al. Thermographic evaluation of horses with podotrochlosis. Am J Vet Res 1983;44:535–539.
- 29.
Clark JA, Cena K. The potential of infrared thermography in veterinary diagnosis. Vet Rec 1977;100:402–404.
- 30.
Palmer SE. Effect of ambient temperature upon the surface temperature of the equine limb. Am J Vet Res 1983;44:1098–1101.
- 31.
Simon EL, Gaughan EM, Epp T, et al. Influence of exercise on thermographically determined surface temperatures of thoracic and pelvic limbs in horses. J Am Vet Med Assoc 2006;229:1940–1944.
