A Template Based Method for Normalizing Thermal Images of the Human Body

Abstract

The medical thermal images provide information about human body physiology. The images are analysed using regions of interest (ROI), which in human body are characterized of having a complex shape, differing slightly within subjects. Those differences such as size and position over time between different examinations affect an accurate analysis. A standardised method is needed to address comparison or average of several images. The proposed method is based in a geometrical template using triangles and barycentric coordinates. Despite an accuracy of 98% within the studied images, further research is needed in the automatic discovery of control points of the ROIs.

Infrared Thermal Mapping, Analysis and Interpretation in Biomedicine

“…Similarly, because a regular ROI such as a rectangle, square, circle or elipse, poorly outlines certain anatomical regions [13]. Vardasca et al [40] designed an automated ROI fitting method to address the issues associated with obtaining a representative temperature value from a user-drawn regular ROI from thermal images of limbs. These examples suggest that automation of thermal images are typically for "bespoke" applications.…”