Hannah Vincent, “Developing In Silico Model of Breast Tissue to Assess Biomechanics of Tumor Microenvironment”
Mentor: Mahsa Dabagh, Biomedical Engineering, Engineering & Applied Science (College of)
Poster #43
The microenvironment endemic to breast tumors is not well understood. Due to the unique anatomy of each case, forces and interactions between noncancerous breast tissue components and malignant tissue are not easily quantified. As such, there exists a need for in-silico modeling built from patient specific data, to more accurately model the biomechanics of breast tissue and tumor microenvironments. The aim of this study is to develop in-silico models of patient-specific breast tissue (malignant and benign), which will then be applied to assess tumor malignancy by quantifying the forces and interactions within the breast tumor tissues. For this study, both standard and high-definition Fourier Transform Imaging (FTIR) were used to image and classify one-hundred-and-one breast tissue samples (tissue arrays from US BIOMAX) into four pathologies: “Normal Breast Tissue”, “Hyperplasia”, “Invasive Carcinoma of No Special Type”, and “Invasive Lobular Carcinoma”. Sub-pathology types “Atypical Hyperplasia” and “Hyperplasia with Saccular Dilation” were included as “Hyperplasia”. An FTIR imaging microscope was used to record high-definition data, to which numerical noise reduction methods were applied to improve signal-to-noise ratio (SNR). Standard definition data was recorded on the same breast tissue samples. Each sample image contained up to six distinct tissue components; malignant epithelium, noncancerous epithelium, dense stroma, loose stroma, reactive stroma, and other (e.g., large areas of inflammatory cells). The images were received in Mat file format and processed in Matlab to extract separate jpegs of each tissue type present in the sample. For example, a sample did not contain malignant tissue extracted only five images. In total, 101 FTIR mat files were processed. The extracted images were then individually imported to MIMICS, a medical imaging software. The images were segmented to remove metadata and create a two-dimensional representation of each tissue type, which was exported in STL format for further analysis in LAMMPS.