CytoBrowser

CytoBrowser enables collaborative visualization, annotation, and structured export of large-scale microscopy images, including z-stacks (focus stacks) from cytology and histology whole-slide images, to generate ground-truth datasets for analysis and machine learning.

Key Features:

• Annotation types: Supports rectangle and polygon region annotations, point-based annotations, and free-form text fields for labeled image markup.
• Annotation classes: Allows each annotation to be associated with a selectable class to produce structured labels for downstream analysis.
• Customizable annotation schemes: Provides interfaces compatible with the Bethesda cancer grading system and permits integration of other annotation schemes.
• Support for z-stacks and whole-slide images: Handles large-scale whole-slide images and z-stacks (focus stacks) common in cytology and histology.
• Structured storage and interoperability: Stores annotations server-side in a structured format and supports JSON-based import/export for interoperability with other tools.
• Collaborative annotation capability: Enables multi-user collaborative sessions to produce consensus annotations and aggregate expert inputs for dataset curation.
• Ground-truth dataset generation: Facilitates creation of high-quality annotated datasets suitable for method evaluation and training machine learning models.

Scientific Applications:

• Cytology and histology annotation: Annotation of whole-slide images and z-stacks for cytological and histological studies.
• Cancer grading and pathology studies: Support for workflows such as Bethesda-based cancer grading and other pathology annotation projects requiring expert consensus.
• Machine learning dataset curation: Generation of structured ground-truth datasets for training and evaluating computer vision and machine learning algorithms in biomedical imaging.
• Consensus-driven studies: Aggregation of multiple expert annotations to produce consensus labels for research and validation studies.

Methodology:

Implemented using JavaScript and Node.js; annotations are stored server-side in a structured format and exported/imported as JSON.