Project description:
- The Lam lab and associated collaborators, funded by the NIH, have identified a need for easy-to-use software-based tools to analyze the information-rich imaging data obtained with ever-improving novel in vitro microscopy-based assays, including microfluidic assays.
- Computational methods capable of processing/interpreting large amounts of imaging data efficiently represent a solution. However, application requires a level of computational expertise impractical for most researchers.
- We present a user-adaptable toolkit packaged into the open-source, standalone Interactive Cellular assay Labeled Observation and Tracking Software (iCLOTS).
- iCLOTS has been designed to bring advanced image processing and data science approaches to all researchers and clinicians, regardless of computational experience.
Application information:
-
iCLOTS is comprised of four categories of image processing applications, a machine learning application, and a suite of video editing tools to help users format data properly.
-
Image processing capabilities are divided into four main categories: (1) adhesion, (2) single cell tracking, (3) multi-scale microfluidic accumulation, and (4) velocity time course/profile analysis.
-
iCLOTS-generated Excel files (or any files following formatting guides) may be used in iCLOTS' machine learning clustering algorithms. iCLOTS implements K-means algorithms to detect and mathematically characterize natural groupings and patterns within clinical data sets, such as blood cell subpopulations or healthy-clinical sample dichotomies.
-
iCLOTS does require specific data formats, e.g. image, image series, or video, depending on application. Users may want to shorten, rotate, crop, or normalize images for most ideal use. A series of video processing tools assists users in preparing their data for successful analysis.
Downloading and using iCLOTS: iCLOTS is open source, e.g. completely free to use and modify. Software for Mac and PC operating systems, detailed documentation, and methods to contact the development team are available at https://www.iCLOTS.org/. Methods-only scripts for interested users are available at https://www.github.com/LamLabEmory and software source code is available at https://www.github.com/iCLOTS.
All applications have been tested using three key analyses to ensure result veracity:
- Sensitivity analysis: reasonable changes in parameter (numerical factors that describe how an algorithm should be applied) values should not produce significantly different results.
- Repeatability analysis: multiple analyses of the same data set and/or similar datasets should produce the same/similar results.
- Comparison to manual analysis: while manual analysis is susceptible to human error or even bias, all iCLOTS results have been compared to hand analysis performed by a trained hematologist to ensure results are within a reasonable range of values.
Results of these analyses will be published in our accepted Nature Communications manuscript.
iCLOTS v0.1.1 identifies cells based on either particle-finding (for brightfield microscopy) or threshold-based (for fluorescence microscopy) algorithms. Should users need more detailed image segmentation methods, users can use software Ilastik (https://www.ilastik.org) to identify individual cell shapes, then upload binary outputs as the inputs to iCLOTS.
Manuscript "iCLOTS: open-source, artificial intelligence-enabled software for analyses of blood cells in microfluidic and microscopy-based assays" by Fay et al. has been accepted for publication in scientific journal Nature Communications. The authors will edit this wiki to add the link to the manuscript upon publication.
Publication in process.