HistoJS: Web-Based Analytical Tool for Advancing Multiplexed Images

Mohamed Masoud, David A. Gutman, Sergey M. PlisJ. Open Source Softw 2024 · DOI · DBLP

Introduction

Recent advances in high-plex imaging technologies, such as MxIF, CODEX, and MIBI, have revolutionized spatial biology by allowing scientists to capture single-cell data in unprecedented detail. These methods can detect over 50 different antigens within a single tissue section, providing deep insights into disease mechanisms and tumor microenvironments. However, the resulting massive datasets—often saved as large-scale OME-TIFF mosaics—present a significant analytical bottleneck. Existing commercial tools are often prohibitively expensive and difficult to customize, while open-source alternatives like QuPath are primarily desktop-bound, limiting accessibility and seamless collaboration.

HistoJS addresses these challenges by providing an open-source, web-based platform designed to visualize and analyze massive, highly-multiplexed immunofluorescence (mIF) images. By combining a robust data-hosting backend with browser-based machine learning, HistoJS enables researchers to perform real-time cell segmentation, phenotyping, and spatial analysis without the need for heavy local installations or expensive software suites.

System Architecture and Methods

HistoJS utilizes a modern, web-centric client-server architecture to handle the scale and complexity of multiplexed data:

  • Backend Data Management: The platform uses the Digital Slide Archive (DSA), a containerized system that employs a MongoDB database and the Girder data management system. This setup handles image data, metadata, and secure user access through RESTful APIs.
  • Frontend Visualization: Integration with OpenSeadragon allows for high-performance viewing of massive images. Custom JavaScript layers enable dynamic panning, zooming, and the real-time compositing of multiple protein channels.
  • In-Browser Analytics: A key innovation of HistoJS is its use of TensorFlow.js with a WebGL-2 backbone. This allows machine learning tasks to be performed directly on the user's local GPU, ensuring a responsive interface while keeping server requirements lightweight.
  • Spatial and Morphological Analysis: The tool incorporates fast techniques for extracting cell boundaries using star-convex polygons. It also computes morphological features (such as eccentricity and solidity) and spatial relationships using Delaunay neighborhood graphs.

Integrated Analytical Interface

The HistoJS interface is designed to bridge the gap between raw data storage and high-level biological insights. Users can transition seamlessly from viewing raw tissue slides to generating biomarker histograms, classifying cell types, and mapping spatial neighborhoods within a single environment. The platform provides immediate, interactive visual feedback, allowing for iterative analysis and discovery.

The HistoJS user interface showcasing multi-channel visualization, cell segmentation, and integrated data plotting.

Performance and Results

The primary achievement of HistoJS is its functional capability on standard consumer hardware. The authors demonstrated that the platform runs efficiently on systems equipped with an Intel Core i7 CPU, 16 GB of RAM, and a mid-range GPU (e.g., GeForce GTX 1050 Ti with a 4GB buffer).

Utilizing the publicly available Lung dataset-1 (Rashid et al., 2019), the tool successfully executed real-time segmentation and phenotyping. These results prove that client-side computation is a viable strategy for big data in pathology, delivering fast response times for rendering and processing complex immunofluorescence images without specialized server-side infrastructure.

Conclusion and Future Work

HistoJS represents a significant step toward democratizing spatial biology. By moving complex histological analysis to the browser, it removes the friction associated with software installation and hardware limitations. This hybrid approach—leveraging client-side compute for big data—makes cutting-edge spatial analysis accessible to a broader range of biomedical researchers and facilitates seamless collaboration across the community. Future work may involve further optimization of the WebGL-2 backbone to scale to even larger and more complex datasets.

Summary updated

BibTeX
@article{masoud2024histojs,
  author = {Mohamed Masoud and David A. Gutman and Sergey M. Plis and Mohamed Masoud and David A. Gutman and Sergey Plis},
  title = {{HistoJS: Web-Based Analytical Tool for Advancing Multiplexed Images}},
  journal = {J. Open Source Softw.},
  year = {2024},
  doi = {10.21105/joss.06317},
}
Figure