Browsing by Subject "microfluidics"

Abnormal tissue growth can result from a variety of factors, including hereditary and carcinogen-induced mechanisms. Overall abnormal tissue development can be described with a step-wise process that is initiated by a single cell mutation which eventually evolves to metastatic invasive cancer. Between the appearance of single mutated cells to its development into in situ cancer tumors, there is a wide time window for early detection and intervention, which can significantly alter the course of the disease and reduce morbidity and mortality. The gold standard for many different abnormal tissue(sickle cell disease, breast cancer etc.) detection is histopathological evaluation. However, histological diagnosis methods can be time-consuming, labor-intensive and expensive. As an alternative, quantitative phase imaging (QPI) is a label-free microscopy technique that provides height information of thin, transparent objects. Without the need of extraneous dyes, QPI offers nanoscale sensitivity to cell morphology and structural changes. Previous studies have shown that a QPI-based imaging modality coupled with microfluidics and machine learning, termed holographic cytometry (HC), is an effective tool for identifying unique biophysical traits of red blood cells(RBCs). In this dissertation, initial works begin with investigating HC’s abilities in identifying abnormal cells in simple homogenous samples such as breast cancer cell lines and proceed to expand the tool’s use for heterogenous samples. Through further hardware development, a novel portable high throughput interferometric chamber (InCh) system, which can screen for abnormal cells in point-of-care settings, is presented in chapter 5. The findings on the implementation of HC to distinguish carcinogen-exposed cells from normal cells and cancer cells are presented in chapter 3. This has potential application for environmental monitoring and cancer detection by analysis of cytology samples acquired via brushing or fine needle aspiration. By leveraging the vast amount of cell imaging data that is obtained with HC, are able to build single-cell-analysis-based biophysical phenotype profiles of the examined cell lines. Multiple physical characteristics of these cells show observable distinct traits between the three cell types. Logistic regression analysis provides insight on which traits are more useful for classification. Additionally, demonstrate that deep learning is a powerful tool that can potentially identify phenotypic differences from reconstructed single-cell images. Different from lab-grown cell lines, human cell samples are a heterogeneous mixture consisting of both healthy, moderately healthy and unhealthy. In the case of sickle cell disease(SCD), only a portion of RBCs are sickle-shaped while majority of a blood sample may still be disc-shaped. To address this issue, a selective search algorithm, based on previous biophysical profiling methods with HC, has been developed to single out any abnormal RBCs within a biopsy sample. By identifying the defining features of sickled red blood cells (RBCs), the algorithm constructs a comprehensive profile that distinguishes between diseased RBCs in sickle cell disease (SCD) and normal RBCs. This profiling enables differentiation between SCD patients and healthy individuals, facilitating better diagnosis and treatment planning. Despite the many benefits of applying the HC system for cytological sample evaluation, it has, to this day, been a difficult modality to implement in clinical settings due to its extreme susceptibility to environmental vibrations, high maintenance, and high build costs. A new design of high throughput QPI has been constructed in this work to counter the issues with the current HC system. Customized microfluidic structures act as beamsplitter, enabling the design of a near common-path interferometric imaging system and eliminating much of the optical components required in the previous HC system design. This new system demands fewer optical components, requires less alignment, and is resistant to environmental vibrations. By combining this newly developed hardware with the software algorithms from the previous two experiments in this thesis, new possibilities for point-of-care clinical applications are unlocked.