MECHANICAL CHARACTERIZATION OF NORMAL AND CANCEROUS BREAST TISSUE SPECIMENS USING ATOMIC FORCE MICROSCOPY

Breast cancer is one of the most common malignancies among women worldwide. Conventional breast cancer diagnostic methods involve needle-core biopsy procedures, followed by careful histopathological inspection of the tissue specimen by a pathologist to identify the presence of cancerous lesions. However, such inspections are primarily qualitative and depend on the subjective impressions of observers. The goal of this research is to develop approaches for obtaining quantitative mechanical signatures that can accurately characterize malignancy in pathological breast tissue. The hypothesis of this research is that by using contact-mode Atomic Force Microscopy (AFM), it is possible to obtain differentiable measures of stiffness of normal and cancerous tissue specimens. This dissertation summarizes research carried out in addressing key experimental and computational challenges in performing mechanical characterization on breast tissue. Firstly, breast tissue specimens studied were 600 um in diameter, about six times larger than the range of travel of conventional AFM X-Y stages used for imaging applications. To scan tissue properties across large ranges, a semi automated image-guided positioning system was developed that can be used to perform AFM probe-tissue alignment across distances greater than 100 um at multiple magnifications. Initial tissue characterization results indicate that epithelial tissue in cancer specimens display increased deformability compared to epithelial tissue in normal specimens. Additionally, it was also observed that the tissue response depends on the patient from whom the specimens were acquired. Another key challenge addressed in this dissertation is accurate data analysis of raw AFM data for characterization purposes. Two sources of uncertainty typically influence data analysis of AFM force curves: the AFM probe's spring constant and the contact point of an AFM force curve. An error-in-variable based Bayesian Changepoint algorithm was developed to quantify estimation errors in the tissue's elastic properties due to these two error sources. Next, a parametric finite element modeling based approach was proposed in order to account for spatial heterogeneity in the tissue response. By using an exponential hyperelastic material model, it was shown that it is possible to obtain more accurate material properties of tissue specimens as opposed to existing analytical contact models. The experimental and computational strategies proposed in this dissertation could have a significant impact on high-throughput quantitative studies of biomaterials, which could elucidate various disease mechanisms that are phenotyped by their mechanical signatures