J Manipulative Physiol Ther 2004 (Mar); 27 (3): 155-159 ~ FULL TEXT
Edward F Owens Jr, DC, John F Hart, DC, Joseph J Donofrio, DC,
Jason Haralambous, DC, Eric Mierzejewski, DC
Palmer Center for Chiropractic Research,
741 Brady Street, Davenport, IA 52803, USA.
BACKGROUND: Paraspinal thermography is used by chiropractors as an aid in assessing the presence of vertebral subluxation. Few reliability studies have been carried out, with mixed results. Digital infrared scanning equipment is now available with location tracking that may enhance reproducibility. Digitized scans enable a computer-aided interpretation of thermographic patterns.
OBJECTIVE: To assess the ability of examiners to reproduce thermal patterns. STUDY DESIGN: Repeated measures with 2 examiners assessing the same patient on 2 occasions. Thirty asymptomatic students served as subjects.
METHODS: A TyTron C-3000 handheld thermographic scanner interfaced to a Microsoft Windows compatible personal computer was used for all recordings. Each examiner recorded 2 scans on each patient. It took an average of 3 minutes to complete all 4 scans. Data were exported to a spreadsheet for initial analysis, then SPSS was used for calculation of intraclass correlation coefficients (ICC). Since the starting and stopping points of scans were not always the same, care was taken to align scans visually, using well-distinguished peaks on the charts as guides. Scans were cropped to remove artifacts that might have occurred at the beginning and end of the scans. Intraexaminer and interexaminer ICCs were calculated.
RESULTS: Skin temperatures ranged from 35.4 degrees C to 30.0 degrees C over all scans. The average temperatures changed little from the first to the last scans, indicating that subjects' overall skin temperatures were stable during the scanning procedure. Intraexaminer ICCs ranged from 0.953 to 0.984. The left and right channel data show slightly higher congruence than the Delta channel. The interexaminer reliability coefficients ranged from 0.918 to 0.975. Again, the Delta channel shows slightly less reliability, although the ICCs were quite high for all channels.
CONCLUSIONS: Intraexaminer and interexaminer reliability of paraspinal thermal scans using the TyTron C-3000 were found to be very high, with ICC values between 0.91 and 0.98. Changes seen in thermal scans when properly done are most likely due to actual physiological changes rather than equipment error.
From the Full-Text Article:
Thermography has been used since the early days of chiropractic as an assessment for the neurological component of the vertebral subluxation, and there have been many systems of measurement and interpretation developed. There also are many claims for the clinical usefulness of the methods. However, we could find no reports in the peer-reviewed literature on the validity of thermography as a tool for subluxation analysis.
The particular method of interpretation we intend to use for thermographic analysis is the “pattern analysis” developed by B.J. Palmer and others. [3, 4] Temperature pattern analysis is 1 component of the subluxation assessment system in use at Sherman College. An article by Owens and Pennacchio  describes the system more fully.
In pattern analysis, successive thermal plots are compared, looking for certain constant features of the temperature profile. When enough constant features are found, the patient is considered “in pattern” and most likely in a subluxated state. Hence, it is crucial to know whether changes are due to errors in data collection or actual changes in the patient's skin temperature profile. This work is a step toward developing an evidence base that might satisfy that need. The first step is reliability testing of the equipment as clinicians use it.
We have attempted to remove the operator as much as possible from the interpretation of the similarity of thermal scans. DeBoer et al  were the first to suggest the use of the ICC in assessing reliability of thermal scans, such as those used in this study. Generally, a diagnostic test produces only 1 or 2 measures that are fairly easy to compare for congruence. Digital thermal scanning, on the other hand, produces a set of 300 or more data points per scan that makes comparison more difficult. The scan can be thought of as a waveform with amplitude and slope features. The ICC is useful because it provides one index of reliability that takes into account both aspects of the waveform and also can test the congruence between several different operators.
In this study, we took care to align and crop data sets to provide a good visual “fit” of the scans. The rationale used is that the temperature peaks represent real hot or cold spots on the patient's back that do not move between scans. The starting and stopping points of the scan were not marked, however, and did vary from scan to scan. The peaks on the temperature plots were used as comparison points to “register” the scans to each other. In future studies, reflective or insulating markers might be affixed to the patient's back to provide more definitive locations for registration.
The aligning and cropping step of data analysis was labor intensive, somewhat subjective, and perhaps prone to operator error. Alignment was produced by sliding files so that their plots began at different points, until the most obvious first peaks in the chart lined up. Using the waveform analogy again, the phase of the plot was shifted, but the amplitude was not changed. The reliability coefficients calculated were quite high, all above 0.90. Using this method, we found the reliability of thermal scanning with the TyTron to be better than that reported by Plaugher et al  but in the range of the work of DeBoer et al.  This work, especially regarding the statistical analysis of reliability, is most similar to DeBoer et al, [2, 12] . perhaps accounting for the similarity of findings. The results of Plaugher et al  were more different from ours, perhaps because they used a temperature sensor that contacts the skin and also used a different approach, based on the “break” analysis, to determine reliability. Keating et al  and Boline et al  had conflicting results in their reliability studies, presumably using the same equipment and methods, but their methods are not well enough described to judge the similarity to the current study.
We have provided data on absolute temperatures as well as side-to-side differentials. Future studies might benefit from an automated process using a software application to search for the best fit by sliding the scans in increments.
One unexpected finding was the need to align the left and right side temperature scans separately. In two thirds of the files, aligning the left channel produced good alignment of the right channel as well, but in the rest, further alignment was needed. Apparently, there was some technical error in one third of our scans that caused either the left or right channel data collection to lag behind. Such an artifact could be due to holding the scanning head at an improper angle, where 1 probe leads the other. Care should be taken during scans to avoid twisting the scanning gun if precision results are desired.
Shifting the left and right channels with respect to each other has a dramatic effect on the Delta channel, which is the calculated side-to-side difference between the channels. In our study, we recalculated the Delta channel after the best alignment was achieved. Care should be taken in interpreting changes that are seen in the Delta channel as calculated by the TyTron C-3000 software. The changes could be due to changes in actual patient temperature patterns, but they could also be produced by tilting the scanner head during the scan. Changes seen in the individual left or right channel temperature profiles would not be prone to this error.
Digital thermal scanning with the TyTron C-3000 appears to be reliable enough for further clinical testing. Thermal scans are stable in the short-term (3 to 10 minutes), but we do not yet know how they fluctuate over longer time periods or in response to care. Other work has been done in this lab to look at changes that occur in thermal profiles over a 31-minute period.  The eventual goal is to develop a system of paraspinal skin temperature assessment and interpretation that will serve as a valid and reliable tool to detect the presence of neurological effects indicative of vertebral subluxation.
Intraexaminer and interexaminer reliability of paraspinal thermal scans using the TyTron C-3000 were found to be very high, with ICC values between 0.918 and 0.984. Intraexaminer reliability is slightly higher than interexaminer reliability. Changes seen in thermal scans when properly done are most likely due to actual physiological changes rather than equipment error.