Comparison of infrared thermal imaging with two canine pain assessment tools in dogs undergoing treatment for chronic back pain.

: Historically, the evaluation and assessment of the clinical response to treatment for canine back pain is subjective and relies on owner and clinician assessment of pain. This study evaluated the use of sequential infrared thermal images as a measure of the response of canine patients with back pain to a prescribed series of photobiomodulation therapy (PBMT) treatments. Qualifying participants had histories of pain and dysfunction associated with spinal osteoarthritis or intervertebral disk disease, or of non-specific uni- or bilateral back pain along the paravertebral epaxial muscles. Each patient was initially thermally imaged prior to PBMT treatment and then received multiple PBMT treatments delivered to the appropriate spinal area on days 1, 2, 3, and 4. Participants were reimaged on day 7. Thermal images provided an objective measure of superficial temperature changes over the area of PBMT treatment of each patient after the PBMT regimen. The temperature correlated with statistically significant changes in Colorado State University Canine Chronic Pain Scale scoring (CPS) and owner assessment using the Canine Brief Pain Inventory (CBPI), which includes a Pain Severity Score (PSS) and Pain Interference Score (PIS). The correlation of objective thermal imaging data with more subjective outcome measures suggests thermal imaging may be a valuable additional tool in monitoring therapy outcome.


Introduction
Non-specific back pain is a common condition seen in pet canines. Intervertebral disc disease is generally presumed to be the most common culprit, and one epidemiologic study suggests the prevalence of the disease in dogs <12 years of age to be 3.5% [1]. Epaxial muscle stress and repetitive injury are also causes of back pain seen in police and working dogs [2], and the diagnosis of degenerative lumbosacral stenosis is prevalent in medium to large breed dogs, especially German Shepherd and Retriever should only be compared to their baseline or previous images, and the presence or lack of symmetry should be noted. Unexpected areas of increased or decreased temperature, changes in temperature over time from one imaging session to the next, and lack of thermal asymmetry indicate an alteration from normal.
In dogs, initial and subsequent pain assessments are helpful to qualify and quantify the extent of pain, as well as to monitor response to therapeutics. [39]. There are multiple metrology tools that have been developed for the assessment of chronic canine pain, and the Colorado State University Canine Chronic Pain Scale (CPS) is commonly used by veterinary staff to evaluate patients. While this tool enjoys frequent use in clinical practice and has been used as an assessment tool when evaluating therapy after back surgery [40], it has no published validation study. Assessment by the owner/caregiver also provides valuable insight, and the standard metrology tool utilized for canines is the Canine Brief Pain Inventory (CBPI). The CBPI evaluates the magnitude of pain via the Pain Severity Score (PSS), and the overall impact of that pain by measuring the Pain Interference Score (PIS). The CBPI has been validated for dogs where a ≥1 change for PSS and ≥2 change for PIS is considered significant [41].
In the first study of which we are aware comparing the results of thermal imaging to monitor treatment response with an objective measure and two clinic metrics, treatment response was monitored in police working dogs with bilateral hip osteoarthritis [42]. This study showed a correlation of the thermal imaging results with results of weight distribution (stance analysis) and clinical metrology instruments, including CBPI, that assess pain and function.
Photobiomodulation induces autocrine signaling within cells, resulting in modulation of cell physiology and function and paracrine cell-to-cell signaling resulting in modulation of tissue physiology and function. Photoreceptor molecules within the cell absorb photonic energy, resulting in increased release of adenosine triphosphate, reactive oxygen species, and nitric oxide. These bioactive substances incite a biochemical cascade of events that lead to increased circulation, reduced pain, modulation of inflammation, and acceleration of healing.
Our hypothesis was that there would be a correlation between patient response to PBMT, pre-and post-treatment thermal images, and patient pain assessment metric tools.

Qualifying Participants
Study participants were a convenience sample composed of patients of a rehabilitation center that met the inclusion criteria of a history of back pain, short to medium haircoat, body weight of 15-55 kg., and no current dermatological disorders. Patients were led into a room with an ambient temperature of 20° C followed by 15 minutes of inactivity to allow for equilibration to room temperature. Patients were positioned in a standing position for thermal imaging. Animal handlers were not allowed to touch the dorsum of the patient's back during the equilibration time or during imaging.
Hair was not clipped from the patient since hair removal can affect the surface temperature of the area being imaged for as long as 60 minutes after removal [59].
A dorsal thermal image of the back was taken. A ghosted outline of the patient in the thermal imaging device software allowed repeatability when framing the image. In each thermal image, an anatomical area from T6-7 to S1-3, including all paravertebral epaxial musculature, was defined as the region of interest (ROI). This ROI also defined the treatment area for PBMT in each patient. The thermal imaging system used has a 640 X 512 high-resolution medical-grade detector with a <.02°C sensitivity and accuracy of +/-1°C. Temperature data were collected from 327,680 points and compiled using calibrated veterinary-specific software (Digatherm IR 640, Infrared Cameras Inc, Beaumont, TX, United States).
PBMT was administered at a fluence of 20 joules/cm 2 , using continuously delivered used for all patients, the treatment time for patients varied as a function of the total area (cm 2 ) being treated. The delivery handpiece was kept in-contact with the hair and skin and moved uniformly over the ROI. All thermal imaging and PBMT treatments were performed by the same person on each patient throughout the study.
The PBMT treatment parameters and application protocol were similar to those previously reported for treatment of osteoarthritis and neuromusculoskeletal disorders in dogs [57][58]60]. Information about device specifications, application method, and treatment parameters previously established as important [61] is in Table 1. On days 2, 3, and 4 PBMT was administered using the same protocol and delivery as the day 1 treatment.

Day 7
The patient was given a physical examination followed by 15 minutes of inactivity to equilibrate to the ambient room temperature. A dorsal thermal image was taken of the patient's back, using the same framing as previous images.

Data Collection
On days 1 and 7, data were collected from each participant. This data included the thermographic measurement of the minimum, maximum, and average temperature in each ROI, a CPS, an owner assessment of improvement or no improvement using a "better", "worse", or "same," an owner generated CBPI worksheet and an owner evaluation of the quality of life (QOL) using a scale of "poor", "fair", "good", "very good", or "excellent". For each patient, the changes in temperatures in the region of interest (minimum, average, and maximum) (ΔTmin, ΔTave, ΔTmax) for days 1 and 7 were calculated. These calculations were performed by the thermal imaging device software.
A timeline of the sequence of initial patient evaluation, owner assessments, photobiomodulation therapy treatments, and thermal imaging is in Figure 1.

Statistical Analysis
Descriptive statistics were calculated. Normally distributed continuous variables were expressed as mean and standard deviation and non-normal distributed variables were expressed as median and range. To determine the correlation between two independent variables (ROI ΔT and CBPI and ROI ΔT and CPS) for each subject who received photobiomodulation, paired t-tests and the Mann-Whitney U-test were used to compare the change in CPS, PSS and PIS scores with the change in ROI thermal temperatures from day 1 to day 7. Two-tailed assessments were used and P values <.05 were considered significant. Because the outcome measurements were independent, Benjamini-Hochberg correction was also calculated for P values, where the false discovery rate was set at 15%. All analyses were performed using a statistical program (IBM SPSS Statistics for Windows, Version 25.0, IBM Corp, Armonk, New York). Twelve dogs meeting the inclusion criteria participated in the trial. All patients completed day 1 and day 7 pain assessment. All twelve dogs had a history of generalized lower back pain.

Patients
Four of the twelve had a diagnosis of chronic back pain combined with multi-joint osteoarthritis. One dog had a known adrenal tumor. Two dogs had a history of ataxia and weakness. Although the laser therapist was not blinded to which animals had multiple joint osteoarthritis, there was no attempt to provide PBMT to anywhere on the patient other than the specified ROI (in this case the thoracolumbar epaxial muscles.) All owners completed the CBPI on days 1 and 7. All dogs completed thermography studies specifically as outlined. Baseline characteristics of all 12 dogs are presented in Table 2.

Qualitative Analysis of Thermal Images and Correlation to Pain Scores
For each dog, the thermal images were paired -with days 1 and 7 presented on the same page. Three veterinarians, experienced with thermal imaging, and blinded to patient signalment, history, and pain scores, evaluated the twelve pair of images qualitatively and recorded improvement or no improvement on a scale of "better", "worse", or "same".
These results were compared to the owner assessment of QOL on days 1 and 7. Any improvement of QOL was tallied as "better", any worsening of QOL was reported as "worse" and no change in QOL was reported as "same".

Ability of Pain Scores and ROI Temperatures to Detect Response to PBMT
From day 1 to day 7, ten of the twelve dogs' clinician pain assessment scores improved as measured by the CPS. The remaining two dogs' CPS scores stayed the same.
The owner assessments performed on days 1 and 7 were more variable, with eight owners reporting improvement in PSS (six were considered clinically significant with improvements in scores ≥1) , and ten owners reporting improvement in PIS (eight were clinically significant with score improvement ≥2). Minimum, maximum and average pain scores on days 1 and 7 are depicted in Figure 4.  Table 3. Table 3. Comparison of response to treatment outcomes using the owner-completed CBPI, the clinician-completed CPS, and the change in ROI ΔT. The graph in Figure 5 demonstrates the minimum, maximum and average ΔTmax results on days 1 and 7. Our objective was to determine if changes in ROI temperatures correlated with changes in pain scores, therefore we used two statistical comparison tests to evaluate the results. Paired T-tests showed a significant correlation between all three ΔT measurements and owner assessment of PIS (P= .001, .006, .009) as well as between the PSS score and the CPS score versus ΔTmax (P= .018). Upon application of the Benjanimi-Hochberg correction factor for two independent variables, we were still able to conclude significance of these correlations, outlined in Table 4.  The Mann-Whitney U-test showed a statistically significant correlation between all three ΔT measurements and owner assessment of PIS (P= .001, .008, .01). Both the PSS and the CPS showed statistically significant correlation only with ΔTmax (P=.03, .02). Table 5 summarizes the U-test results. The relationship between improvement in PSS, PIS and improvement in ROI ΔTmax is presented in Figure 6.  ; when pain scores improved, ΔTmax reduced, and conversely, where pain scores worsened, ΔTmax increased.

Discussion
Infrared thermography has long been used in various industries to measure minute changes in surface temperature otherwise invisible to the naked eye. Veterinary medicine offers a unique opportunity to explore thermography as a screening tool to aid the clinician in evaluating non-verbal patients. The use of thermography in veterinary medicine was been reported for the evaluation of lameness in horses [62][63][64], companion animal clinical applications [14,[25][26][27][29][30][31][32][33][34][35][36][37], and structural screening in working dogs [38,42]. When combined with palpation, thermal imaging has been shown to be a useful tool in differentiating painful cats from non-painful cats [22]. The objective of this study was to evaluate correlation between patient response to PBMT with pre-and posttreatment thermal images and patient pain assessment metric tools.
Research studies that evaluate pharmacologic or non-pharmacologic pain treatment modalities often use pain scores along with an objective measurement to evaluate success or failure of treatment. For example, force plate gait analysis, combined with owner pain assessment, has been used to evaluate the efficacy of carprofen in dogs [65]. While pain scoring of non-verbal patients is difficult, its use is critical in providing adequate and successful pain management for the patient [66,67]. Subjective pain scores, such as numerical rating or visual analogue scores, have been shown to have limited correlation with objective force plate data in dogs evaluated after knee surgery [68]. It is widely accepted that there is a significant caregiver placebo effect which can reach close to 40% in pet owners when compared to an objective outcome measurement and be even higher when veterinarians or veterinary staff perform pain assessment scoring [69]. Therefore, especially for chronic pain conditions, it is generally accepted practice to utilize a combination of pet owner scoring, objective measurement data, and clinician pain assessment to determine success or failure of a treatment protocol.
In this study, our objective measurement was temperature normalization over the assigned treatment area region of interest. Our results suggested that when we use an owner-reported, validated, canine chronic pain scoring tool, (CBPI), there was a statistically significant correlation between changes of temperature in the region of interest and the pain interference scores, while the pain severity scores were only significant when evaluating ΔTmax.
We expected that ΔTave measurements would be the most useful, however, in this sample population, ΔTmax correlated best with CBPI scores. When collecting temperature data, all data points in the ROI are used to calculate the minimum, maximum, and average temperatures. The number of temperature data points at the highest temperatures, ΔTmax, either increases or decreases as the circulation increases or decreases within that ROI. In this study, ΔTmax decreased in patients that responded to PBMT with a reduction in inflammation and a corresponding decrease in circulation.
In a previous publication evaluating the use of thermal images in monitoring response to therapy, maximum temperatures in ROIs showed greater correlation significance than average temperatures when compared to other metrics [42]. The authors of that study theorized that the maximum temperature may better reflect the changes in circulation in the underlying tissue than average temperature, and that using maximum temperature data in a ROI may eliminate variables in ROI average temperature calculation due to non-affected tissue inclusion in the ROI.
Not surprisingly, the use of the most basic assessment tool, the owner QOL score, was the least accurate outcome measurement. Inherent bias in reporting is a continual problem in evaluation of treatment success [70] and our study reinforced the issue. It is interesting to consider whether sharing the Day 7 images with the pet owners would have changed their opinion of treatment outcome.
Pain is multifactorial, complex, and in all species of animals, can be influenced by biological, psychological, and social factors [71,72]. The successful use of any modality to treat pain can be difficult to assess as pain scoring and observational evaluation alone are often quite subjective. In this study we evaluated objective thermal imaging data as a therapy outcome measure, comparing it to more subjective pain and quality of life scoring.
Comparing an objective measure of therapy outcome to subjective measures of outcome can only validate the objective measure as being as accurate as the subjective measure. We acknowledge the limitations of this study, namely the small sample size, and the lack of a control subset of patients. Because the thermal imaging evaluation was not used to compare response between patients, we do not feel that the differences in patient size, coat color and length, and variety in disease pathologies posed significant variability to the data, as each patient behaved independently in their response to the prescribed therapy and that response was recorded via the collected thermal imaging surface temperature data. Informed Consent and Ethical Review: In the United States where this study was conducted, there are no regulations or guide-lines that specifically apply to private practices that wish to conduct research [73]. This study was performed in a private clinical practice, using client-owned animals,

Conclusions
where a valid veterinary-client-patient-relationship existed. The therapy treatments and data collection described (PBMT, medical thermal imaging, and the use of pain score questionnaires) are treatments and interventions that are within the normal scope of routine veterinary practice and considered standard of care. Informed client consent was obtained from each owner. A copy of the