Need for Standardisation of Measurements in Ther- mal Imaging K. Ammer Ludwig Boltzmann Forschungsstelle für Physikalische Diagnostik, Hanuschkrankenhaus, Heinrich Collinstrae 30, A-1140 Vienna, Austria School of Computing, University of Glamorgan, Pontypridd, Wales, CF37 1DL, United Kingdom Abstract A protocol for capturing a series of images from the whole body of a healthy subject has been es-tablished. A total of 24 views of the body were specified and within these views, a total of 87 regions of interest (ROI) were defined. The repeatability of some standard views by different investigators and the inter- and intra-rater reliability of temperature readings from selected regions of interest has been investigated. The highest variation in position- ing was found in the hands and feet. The face varied in a very narrow range. Individual dimensions of these body regions contribute to the variation of positioning. In the case of dorsal hands the distance between both little fin- gers may be longer than the distance from the wrist to the tip of the middle finger. Such a condition prevents the precise positioning in a defined manner. Similar conditions may occur in the views Upper Back, and Anterior Knees. According to the results of this investigation the rules for positioning and image capture of dorsal hands, upper back and anterior knees have been modified. Inter-rater reliability coefficient alpha and ICC of the ROI Lower Arm, and the hourglass shaped ROI at the anterior knee confirmed excellent repeatability of ROI placement. The influence of the angle of view on tempera- ture readings from an identical object is also discussed. Reference values for the surface temperature of body regions based on images captured according to our protocol will reflect mainly the individual temperature varia- tion. 1. Introduction Existing literature lacks information on reference values of temperature distribution on the human body surface, although some data are available on the symmetry of temperature di- stribution [1, 2]. In radiography, standardized positions of the body for image capture have been included in the protocol for quality assurance for a long time [3]. Reproducibility of joint positioning of the knee and of measurements related to standard radiographs have been reported [4]. The American Academy of Thermology has published Technical Guidelines for thermal imaging, but information on body positioning to capture images of defined body re- gions was incomplete [5]. Positions of placement of regions of interest (ROI) are rarely de- scribed and information on the reproducibility of temperature readings from thermal images is scarce [6]. Standardized positions of the body for image capture and definition of reproducible for temperature meas-urements can reduce systematic errors. Images cap-tured and analyzed in a clearly defined protocol will mainly show the individual variations in temperature distribu- tion, and deviations from these standard images would be indicative of a physical dysfunc- tion. 2. Materials and Methods We have established a protocol for capturing a series of images to cover the whole body surface of a healthy subject. The EuroQol protocol defines a healthy subject as someone who had no problems with mobility, no difficulty in caring for himself, no restrictions in per- forming normal activities, experienced no pain or dis-comfort and was not suffering from anxiety or depression [7]. A total of 24 views of the body were specified and within these views, a total of 87 regions of interest (ROI) were defined. The consistency of the 8 standard views: Face, Dorsal Neck, Upper Back, Anterior Left Arm, Dorsal Hands, Both Knees Anterior, Lateral Right Leg and Plantar Feet was evaluated. The distance, measured in pixels, from the upper or the lower edge of the image to anatomical landmarks was used for evaluation. The cross-section tool of the software package CTHERM [8] was used for the determination of distances.