Statistical signal processing for quantitative assessment of pulsed terahertz imaging of human breast tumors

This paper presents utilization of pulsed terahertz imaging to detect and assess the margins of excised human breast tumors. The freshly excised bulk tissue and the block of the same tissue fixed in formalin and embedded in paraffin (FFPE) are scanned using the reflection imaging module. The results show that while the THz images of block tissue demonstrate strong contrast between cancerous and fibroglandular (normal) regions, the contrast is less strong in the images of freshly excised bulk tissue. This could be due to remains of blood and presence of air bubbles in the specimen. To validate the THz images, we develop signal processing and statistical data analysis techniques based on image morphing and Bayesian modeling. The results are quantified in the form of receiver operating characteristic (ROC) curves.