Computational Methods for Rare Cell Detection in Streak Image Flow Cytometry
Loading...
Links to Files
Permanent Link
Author/Creator
Author/Creator ORCID
Date
2018-01-01
Type of Work
Department
Computer Science and Electrical Engineering
Program
Computer Science
Citation of Original Publication
Rights
Distribution Rights granted to UMBC by the author.
Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.
This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.
This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
Abstract
Detection and analysis of Circulating Tumor Cells (CTCs) have shown promising cancer clinical and research applications. A major challenge for the detection and analysis of CTCs is their rare nature. CTCs are found at very low concentrations in blood. Therefore, large volumes of sample are needed for meaningful enumeration, which especially impedes the analysis of CTCs using standard flow cytometry (due to its low throughput). This issue is addressed by the recent development of a high throughput imaging cytometer equipped with a wide field flow-cell. This wide-field flow cytometer adapts a technique known as "streak photography” where exposure times and flow velocities are set such that the cells are imaged as short "streaks”. Streak cytometry technology enables analysis of cells in very low concentrations (0.1 cell /ml) in large volumes (10 ml) and rapidly (1 minute). However, dynamic imaging conditions in streak cytometry introduce more challenges to current automated cell counting methods, especially with low cost, low resolution webcams or smartphone cameras affordable for use in point of care and global health settings. The lack of automatic enumeration methods for streak imaging limits clinical utility of wide field streak cytometry. In this dissertations we propose to combine traditional geometrical and intensity distribution (GID) features with visual words plus a novel image classification method based in relational features that characterize an object in relation with other objects in frames in a video file. This new cell identification/quantification method, Relational Streak Algorithm (RSA), consists of three parts: (1) finding streaks with a binary mask that contains potentially all the cells in a frame, (2) identifying candidate cells using GID features, and (3) filtering out spurious cells and identifying true cells with machine learning approaches for image classification by GID features, visual words, and relational features. We incorporate the relational features in a selective permeable filter that can either discard cells, allow cells to proceed through the filtering layers, or defer the cells to the most sensitive classifier (in this case the visual words classifier) for final classification. We evaluated the RSA using samples with nominal concentrations of 1 cell per mL and 1 cell per 10 mL (consistent with acceptable numbers of CTCs considered to show clinical significance). The RSA performed well with both concentrations. In the 1 cell per mL dataset, the algorithm achieved 88% sensitivity with an F1 score of 91%. In the 1 cell per 10 mL dataset, the algorithm achieved sensitivity of 84% with an F1 score of 75%, outperforming earlier versions of the algorithm and current tools for cell tracking (CellTrack and MTrack2) used as comparisons. These findings demonstrate superiority of the new analytical capabilities of streak-based cytometry when coupled with the RSA for automated cell detection and counting. This cell counting capability enables automated low-cost streak imaging flow cytometry detector for clinical and research use and offers the possibility of expansion of cell-based clinical diagnostics to resource-poor settings.