Thermographic analysis of the abdominal region of pregnant women

Abstract

The use of thermography in the biomedical field is becoming increasingly popular and is providing promising results in numerous medical fields such as complications from diabetes [1, 2, 3] and breast cancer [4, 5, 6]. This dissertation is one of the earliest investigations which study whether thermography could also be applied to obstetrics. The contributions of this study include: 1. Dynamic thermal imaging, which was for the first time used to investigate both temporal and spatial temperature patterns for the abdominal region of pregnant and non-pregnant women. The few studies on obstetric thermography only acquire static images so the temperature results are subjective to the participants’ conditions at that instant. 2. The development of a video registration technique, which can be applied to the homogenous texture of human skin, without the use of any markers thus making the system completely non-invasive. In order to obtain reliable temperatures, the images in the thermal video had to be geometrically aligned. Two image alignment techniques were considered: a sequential image registration via triangulation, and optical flow. Principal component analysis (PCA) image reconstruction of the registered images was also implemented to remove any residual movement still present in the registered images. 3. The investigation of the validity of the standard acclimatisation period of 20 minutes, which is used by all thermographic studies in the literature. Most of the thermal patterns for non-pregnant women still exhibited temperature variations at 60 minutes, whereas most thermal patterns for pregnant women achieved a steady state value after one hour. 4. The use of PCA for temperature data analysis, which had never been considered for medical thermographic analysis. Both pregnant and non-pregnant participants experienced a gradual temperature decay, or rise in some cases, which is generally present across the whole abdomen. The first three principal components were found to be consistent in all participants and thus could be associated with normal processes of the human body. This investigation has shown that the use of dynamic thermal imaging combined with suitable registration and data analysis techniques can yield valuable information on passive and active physiological processes for obstetrical applications.