Label-free screening and enumeration of tumor cells in blood using digital holographic microscopy and machine learning



Journal Title

Journal ISSN

Volume Title



Circulating tumor cells or CTCs isolated from the peripheral blood of cancer patients have shown promise as clinical biomarkers enabling monitoring of various treatment regimens. Current methods of isolation rely either on specific cell proteins or differences in physical characteristics between CTCs and blood cells. Although these methods are able to enrich CTCs, the presence of background WBCs mandates antibody-based immunostaining to differentially identify CTCs which limits their promising translational utility. Typically, downstream assays such as cytogenetics and in vivo expansion require a sufficiently large number of CTCs for analysis. Since these are killed during the staining process, it is not possible to know beforehand, which patient samples are suitable for CTC characterization. Therefore, it becomes important to develop screening tools that can achieve live CTC enumeration from cancer patient blood samples. Such a technology will enable effective decision making on the type of downstream assay to be pursued depending on the detected CTC burden. The main goal of my dissertation is to develop a robust platform for the enumeration of live tumor cells in blood. Importantly, the technique should demonstrate high throughput, which is needed to translate it to detect CTCs from patient blood samples. This is achieved by interrogating cells in flow, using a 3D imaging technique, digital holographic microscopy (DHM). To achieve reliable enumeration, a deep learning (DL), image-based classification model is optimized and trained to identify cancer cells from mixed samples containing a background of WBCs. A major advantage of this label-free enumeration platform is that it can be easily interfaced with any upstream enrichment method enabling this promising technology to be easily translated across different laboratories.

Embargo status: Restricted until 09/2023. To request the author grant access, click on the PDF link to the left.



Cancer Screening, Digital Holography, Deep Learning