CellProfiler Analyst Web (CPAW) - Exploration, analysis, and classification of biological images on the web

Keywords

Machine Learning Techniques, Life Sciences, Health, Medicine, Biology, Bioinformatics, Genomics, Machine Learning, Statistics, Modelling, and Simulation Applications, Data Analysis, Reasoning, Problem Solving, and Decision Making, Application Motivated Visualization, Task Abstractions & Application Domains, Human-Subjects Qualitative Studies, Human-Subjects Quantitative Studies, Image and Video Data

Abstract

CellProfiler Analyst has enabled the scientific research community to explore image-based data and classify complex biological phenotypes through an interactive user interface since its release in 2008. This paper describes CellProfiler Analyst Web (CPAW), a newly redesigned and web-based version of the software, allowing for greater accessibility, quicker setup, and facilitates a simple overall workflow for users. Installation and managing new versions has been challenging and time-consuming, historically, for CellProfiler Analyst. CPAW is an alternative that ensures installation and future updates are not a hassle to the user. CPAW ports the core iteration loop of CPA to a pure server-less browser environment using modern web-development technologies, allowing computationally heavy activities, like machine learning, to occur without freezing the UI. With a setup as simple as navigating to a website, CPAW presents a clean user interface to the user to refine their classifier and explore phenotypic data with ease. We evaluated both the old and the new version of the software in an extensive domain expert study. We found that users could complete the essential classification tasks in CPAW and CPA 3.0 with the same efficiency. Additionally, users completed the tasks 20 percent faster using CPAW compared to CPA 3.0. The code of CellProfiler Analyst Web is open-source and available at \url{https://mpsych.github.io/CellProfilerAnalystWeb/