Improving Quantitative Infrared Imaging for Medical Diagnostic Applications

Infrared (IR) thermography is a non-ionizing and non-invasive imaging modality that allows the measurement of the spatial and temporal variations of the infrared radiation emitted by the human body. The emitted radiation and the skin surface temperature that can be derived from the emitted radiation data carry a wealth of information about different processes within the human body. To advance the quantitative use of IR thermography in medical diagnostics, this dissertation investigates several issues critical to the demands imposed by clinical applications. We developed a computational thermal model of the human skin with multiple layers and a near-surface lesion to understand the thermal behavior of skin tissue in dynamic infrared imaging. With the aid of this model, various cooling methods and conditions suitable for the clinical application of dynamic IR imaging are critically evaluated. The analysis of skin cooling provides a quantitative basis for the selection and optimization of cooling conditions in the clinical practice of dynamic IR imaging. To improve the quantitative accuracy for the analysis of dynamic IR imaging, we proposed a motion tracking approach using a template-based algorithm. The motion tracking approach is capable of following the involuntary motion of the subject in the IR image sequence, thereby allowing us to track the temperature evolution for a particular region on the skin. In addition, to compensate for the measurement artifacts induced by the surface curvature in IR thermography, a correction formula was developed based on the emissivity model and phantom experiments. The correction formula was integrated into a 3D imaging procedure based on a system combining Kinect and IR cameras. We demonstrated the feasibility of mapping 2D IR images onto the 3D surface of the human body. The accuracy of temperature measurement was improved by applying the correction method. Finally, we designed a variety of quantitative approaches to analyze the clinical data acquired from patient studies of pigmented lesions and hemangiomas. These approaches allow us to evaluate the thermal signatures of lesions with different characteristics, measured in both static and dynamic IR imaging. The collection of methodologies described in this dissertation, leading to improved ease of use and accuracy, can contribute to the broader implementation of quantitative IR thermography in medical diagnostics