Skip to main content
eScholarship
Open Access Publications from the University of California

UC San Diego

UC San Diego Electronic Theses and Dissertations bannerUC San Diego

Automated detection and classification of circulating cancer cells via high-throughput microscopy

Abstract

Mammography is the gold standard for detecting breast cancer, and there is active research to detect cancer at its earliest stages to improve prognosis. Tissue-specific biomarkers have been studied to identify micrometastases in bone marrow and/or peripheral blood. Many studies, however, report conflicting results for detection sensitivity and specificity, forestalling translation into clinical use. Automated image cytometry has many potential advantages, and this dissertation explores the feasibility of identifying circulating tumor cells (CTCs) via high- throughput microscopy and automated cell-by-cell classification. We used a fully-automated microscope that is functionally equivalent to a Q3DM/Beckman Coulter EIDAQ /IC 100 image cytometer, including optics for high-content image acquisition of multiple fluorophores and robotics for stage movement and autofocus. Feature data (fluorometric as well as morphometric measurements) were collected from \textit}in vitro} spiked models of breast cancer in blood. Prestaining with CellTracker creates a gold standard for assessing cancer origin. Not all breast cancer cells stain for cytokeratins and EpCAM (two markers of epithelial origin). Genetic instability and heterogeneity of cancers makes finding an absolute set of differential expression markers difficult, if not impossible, to reliably identify all CTCs. We hypothesize that CTCs may be detected by morphology instead of biomarkers. In fact, pathologists often look for morphological changes and abnormalities when analyzing tissue biopsies to diagnose cancer. Clumped cells are a challenge for analysis. In flow cytometry, clumps are simply discarded. Useful information may still be recovered in image cytometry, and image processing techniques were evaluated to improve clump segmentation and cancer detection. Gating on nuclear size (area) and shape (wiggle) allows a tractable subset of suspicious cells for a pathologist's review rather than requiring inspection of all cells. However, CTCs also have morphological heterogeneity, so supervised classification via automated neural networks was developed to leverage the wealth of feature information available. Preliminary in vivo animal studies looked at what correlation exists between the number of CTCs and disease progression. Routine detection of CTCs would enable physicians to observe the efficacy of treatment and modulate therapy

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View