Research Article: FluoroCellTrack: An algorithm for automated analysis of high-throughput droplet microfluidic data

Date Published: May 1, 2019

Publisher: Public Library of Science

Author(s): Manibarathi Vaithiyanathan, Nora Safa, Adam T. Melvin, Arum Han.

http://doi.org/10.1371/journal.pone.0215337

Abstract

High-throughput droplet microfluidic devices with fluorescence detection systems provide several advantages over conventional end-point cytometric techniques due to their ability to isolate single cells and investigate complex intracellular dynamics. While there have been significant advances in the field of experimental droplet microfluidics, the development of complementary software tools has lagged. Existing quantification tools have limitations including interdependent hardware platforms or challenges analyzing a wide range of high-throughput droplet microfluidic data using a single algorithm. To address these issues, an all-in-one Python algorithm called FluoroCellTrack was developed and its wide-range utility was tested on three different applications including quantification of cellular response to drugs, droplet tracking, and intracellular fluorescence. The algorithm imports all images collected using bright field and fluorescence microscopy and analyzes them to extract useful information. Two parallel steps are performed where droplets are detected using a mathematical Circular Hough Transform (CHT) while single cells (or other contours) are detected by a series of steps defining respective color boundaries involving edge detection, dilation, and erosion. These feature detection steps are strengthened by segmentation and radius/area thresholding for precise detection and removal of false positives. Individually detected droplet and contour center maps are overlaid to obtain encapsulation information for further analyses. FluoroCellTrack demonstrates an average of a ~92–99% similarity with manual analysis and exhibits a significant reduction in analysis time of 30 min to analyze an entire cohort compared to 20 h required for manual quantification.

Partial Text

Development of fluorescence and image-based single cell technologies has enabled systematic investigation of cellular heterogeneity in a wide range of diseased tissues and cellular populations [1, 2]. While conventional single cell analytical tools like flow cytometry (and Fluorescence Activated Cell Sorting, Image Flow Cytometry) can detect, sort and collect cells with desired properties, these techniques do not permit dynamic monitoring of cell responses as the data is collected at a single time point [3]. Considering these limitations, microscale technologies such as droplet microfluidic devices and microfluidic cell trap arrays allow for facile collection and segregation of single cells to enable real-time investigation of cellular processes [4, 5]. Droplet microfluidic devices in particular, have an advantage of working with picoliter to nanoliter volumes of solution that increases sensitivity, specificity, and precise quantification of real-time intra and extracellular processes [3]. The development of a wide variety of sophisticated cellular fluorescent probes in recent times has enabled easy tracking and detection of cellular activities by incorporating static microdroplet trapping arrays with fluorescence microscopy platforms to eliminate the need for high-speed cameras and expensive fiber optics used in large-scale cytometric tools [6, 7]. This technology has found a diverse set of applications in disease detection and diagnostics ranging from single cell analyses to droplet-based quantitative PCR and electrokinetic assays [8–11]. One such example in cellomics is the use of fluorescent stains and organic dyes in droplet microfluidic devices to sort cells based on their dynamic fluorescent responses to external stimuli [12, 13]. Similarly, fluorescent proteins, quantum dots, and luminescent nanoparticles have been used to track protein-protein interactions, intracellular enzyme activities, and identify biomolecules or biomarkers within single cells encapsulated in droplets [14–17]. In addition to cellomics, massively parallelized high-throughput droplet generators are used in combination with fluorescent barcodes to perform single cell DNA- and RNA- sequencing [18, 19]. Digital droplet microfluidics are also extensively used in the quantitative immunoassays and development of biosensors [20]. Beyond disease detection and diagnostics, fluorescence-based droplet microfluidics also finds applications in renewable energy, pharmaceutical industry and managing environmental issues [21–24].

The FluoroCellTrack algorithm consists of several steps as depicted in Fig 1. This schematic outlines the pipeline that can quantify a wide range of data generated using the microfluidic droplet trapping array. The algorithm begins with a basic feature to read fluorescent microscopic images that were collected and stored in folders. Approximately 30–40 unprocessed, overlaid brightfield and fluorescent images were batch fed into the algorithm. The major steps of the algorithm included: (1) pre-processing to remove noise from the images; (2) feature detection including droplet and contour detection; (3) post-processing to extract essential information from the images; and (4) exporting the results for additional analyses. All the processing steps involved 16-bit processing techniques.

The multistep Python algorithm, FluoroCellTrack was developed to process and analyze folders of microscopy images obtained from global droplet microfluidic experimentation. The algorithm has several distinct features including the automatic detection of droplet subpopulations (e.g., empty droplets, droplets with single cells, droplets with multiple cells), the quantification of single cell responses to drugs, and the quantification of intracellular fluorescence in intact cells. FluoroCellTrack was successfully implemented in three different systems: (i) live/dead subpopulation studies to understand cellular responses to different doses of drugs, (ii) quantification of cell and nanoparticle co-encapsulation for droplet tracking information, and (iii) quantification of CPP uptake in single intact cells based on fluorescent intensity. This algorithm was found to be superior to the commonly used feature detection techniques like Edge Detector thresholding, template matching and had well-defined and precise steps to eliminate false positives such as debris (e.g., cell or peptide) across the trapping array. Manual control analyses conformed with the Python algorithm with an average similarity of ~92–99% from a mean population of 320 cells. Moreover, automated image analysis took about <1 min to count all the cells trapped in the device and <30 min to quantify the fluorescence intensity across the entire population of cells, proving it to be a powerful tool for microscopy data analysis. This was far superior to the ~60 min required for manual cell counting and ~20 h needed for manual analysis of intracellular fluorescence. While the work here demonstrated the utility of FluoroCellTrack with a low-throughput droplet data (~787 droplets), the algorithm has a potential to quantify high-throughput droplet data (~10,000 droplets) in a couple of hours in comparison to days of manual analysis. Finally, FluoroCellTrack was found to overcome the limitations of existing non-droplet microfluidic algorithms and has the potential to be integrated with several different types of microfluidic devices, trapping arrays and non-microfluidic platforms with easy user-defined modifications. Beyond tracking and quantifying intracellular fluorescence, this algorithm has multiple applications in tracking cellular movements, through time-lapse imaging and position detection and can also be potentially extended to understand subcellular molecular processes by analyzing intracellular localization of biochemical stains through intracellular pixel quantification.   Source: http://doi.org/10.1371/journal.pone.0215337

 

Leave a Reply

Your email address will not be published.