|Abstract:||According to the latest World Health Organization (WHO) statistics, breast cancer is the most common type of cancer among women worldwide. The WHO has further emphasized that early diagnosis and treatment are key in mitigating the burden of disease. In spite of this assessment, the standard histopathology of breast cancer still relies on manual microscopic examination of stained tissue. Being qualitative and manual in nature, this standard diagnostic procedure can suffer from inter-observer variation and low-throughput. In addition, stain variation between different samples and different laboratories creates problems for supervised image analysis methods for automated diagnosis. A quantitative, label-free and automatable microscopic modality for breast cancer diagnosis is, thus, needed to address these shortcomings in the standard method.
Furthermore, prognostic biomarkers are important tools used by clinicians in order assess the disease course in patients. Being correlated with outcomes, these markers allow pathologists to determine aggressiveness of disease and tailor treatment accordingly. However, the current set of biomarkers for breast cancer are ineffective in predicting outcomes in all patients and there is a need for additional markers of prognosis to better account for variation among individuals. Microscopic and imaging tools for extracting new, quantitative biomarkers during breast histopathology are, thus, also desirable.
Although a number of new quantitative imaging modalities for diagnostic and prognostic evaluations have been proposed, a key challenge remains compatibility with the existing workflow for easier clinical translation. Quantitative methods that minimally affect the clinical pipeline already in place are expected to have a greater impact than those that require significant new infrastructure.
During my graduate work I have approached these problems in modern breast histopathology by using quantitative phase imaging (QPI). QPI is a label-free microscopy technique where image contrast is generated by measuring the optical path-length difference (OPD) across the specimen. OPD refers to the product of the refractive index and thickness at a point in the field of view. Since this measurement relies on a physical property of tissue and is label-free, it provides an objective and potentially automatable basis for tissue assessment. We employ a QPI technique called Spatial Light Interference Microscopy (SLIM) for investigations carried out during this thesis research.
The specific aims of my thesis research are:
1. Label-free quantitative evaluation of breast biopsies using SLIM: In this work, we show by imaging a tissue microarray (TMA) that our QPI based method can separate benign and malignant cases by relying on tissue OPD based features. By employing image processing and statistical learning, we demonstrate a label-free quantitative diagnosis scheme that can provide an objective basis for tissue assessment. A quantitative method like this can also, potentially, be automated, reducing case-load for pathologists by automatically flagging problematic cases that require further investigation.
2. Quantifying tumor adjacent collagen structure in breast tissue using SLIM: Recent evidence shows that the structure of tumor adjacent collagen fibers influences tumor progression. In particular, collagen fiber alignment and orientation can facilitate epithelial invasion to surrounding tissue. We demonstrate that SLIM can be used to detect this prognostic marker that in the past had been detected using Second Harmonic Generation Microscopy (SHGM). Our SLIM based method improves on the SHGM based method in terms of throughput and the fact that cellular information can be obtained, in addition to collagen fiber structure, in a single image.
3. Quantitative histopathology on stained tissue biopsies: The instruments and image analysis tools developed in Aims 1 and 2 are designed for unstained tissue biopsies. Since standard tissue histopathology inevitably requires staining, we aim to demonstrate that we can extend these tools to stained tissue biopsies. In this way, the standard diagnostic workflow will be minimally disrupted. In addition, from a single shot, both an OPD map and stained tissue bright field image will be obtainable for evaluation. We demonstrate that QPI images of stained tissue can be used to solve diagnostic and prognostic problems in breast tissue assessment, using quantitative markers.