Why should a thermographer be so concerned with the physi-
cal principles, operation, and care of a thermal imager? Because with any complicated combination of microscopic analog devices, mechanical mechanisms, and compensatory digital software, a great deal can go wrong. Sometimes the problem is readily discerned. The imager may freeze up or reboot frequently. The device may fail to turn on in the first place, or the image may become streaked or speckled with noise. At other times, however, significant problems are far more subtle and may even go unnoticed because the imager seems fully operational, generating clear and complete images. The primary focus of this chapter is on those subtle yet critical errors that can be overlooked unless they are checked at regular intervals. Thermal imagers measure microbolometer leaflet resistance, not temperature. Thus, they must be calibrated at the factory using a series of known blackbody temperatures and sensing the resultant resistance values, as shown in Figure 4.1. During calibration, a refer- ence table is produced containing the digitized temperature/resis- tance measurements entered into the calibration software. Every microbolometer pixel within the imager�s FPA chip has a slightly different temperature offset and gain. To produce an accurate ther- mogram, these different gains and offsets must be corrected to a normalized value. The manufacturer�s calibration software accom- plishes this via non-uniformity correction (NUC), which uses multi-point correction of both thermal gain and offset of each pixel. The pixel calibration is stored in the imager�s memory as a numeric table that relates resistance values to blackbody temperatures. Later, when the imaging system takes a measurement, it utilizes the cali- bration data to calculate the sensed surface temperature. Up to 1% of the pixels in a microbolometer array may pro- duce outputs far outside of a correctable range. Ideally, the NUC process identifies these faulty pixels and replaces them using a nearest-neighbor averaging algorithm. Faulty pixels are difficult to detect during routine imager operation due to the NUC pro- cess. Fortunately, these faulty pixels are not clinically significant as long as the bad pixels are scattered throughout the detector matrix and not adjacent to one another. As pixel offset and gain may change with time, the imager needs to be recalibrated periodically, typically every two years. This chapter is concerned with checking the imager�s calibration to assure these offset and gain values remain within specified limits over all regions of the image. The following three measurements are necessary when manu- ally verifying the calibration of a thermal imager: 1. Thermal Gain. An actual 1.0�C (1.8�F) temperature differ- ence between two locations on a target should be detected and displayed as a 1.0�C (1.8�F) temperature difference by a properly calibrated thermal imager; i.e., the Thermal Gain should be 1.0. This 1:1 relationship should remain constant whether the measurement is near room tem- perature (Troom, ~20�C/68�F), body temperature (Tbody, ~37�C/98.6�F), or at any temperature in between. If the thermal gain is in error, a 1.0�C (1.8�F) temperature dif- ference on a target may be displayed by the imager as only 0.2�C (0.36�F) or some other wrong value, making the quantitative measurements erroneous, yet the qualitative image may still appear completely normal by adjusting the color span.2 As the quantitative interpretation of human thermograms relies heavily on differential temperature measurements, assuring a Thermal Gain between 0.95 and 1.05 (preferably 1.0) across the physiologic temperature range is of paramount importance. 2. Temperature Offset. This refers to the temperature dif- ference between the target surface temperature detected by the infrared imager and the actual temperature of that surface. Temperature offset and thermal gain are related: if the imager�s offset value is stable across the physiologic temperature range, the thermal gain will be 1.0 over those temperatures. Stable offset values up to �2�C (3.6�F) are typical for professional microbolometer imagers. 3. Temperature Flatness. �Flatness� measures the consis- tency of a constant temperature object as sensed at different regions of the image, usually measured at the center and four corners of the image at Troom. The imager should be fully warmed up before this test so that the imaging chip has reached a stable temperature. The flatness of the image at Tbody can also be measured by using a Blackbody Calibrator.