Quick Annotator: an open-source digital pathology based rapid image annotation tool.

Area of Science:

• Digital pathology
• Computational biology
• Biomarker discovery

Background:

• Accurate segmentation of histologic structures in whole slide images (WSIs) is crucial for image-based biomarker discovery.
• Manual annotation of these structures is time-consuming and often impractical for large datasets.
• Existing methods lack efficiency, hindering large-scale analysis.

Purpose of the Study:

• To introduce Quick Annotator (QA), an open-source tool designed to dramatically improve the efficiency of histologic structure annotation in WSIs.
• To demonstrate QA's effectiveness in accelerating the annotation process while maintaining high accuracy.
• To provide a scalable solution for generating annotated WSIs for downstream biomarker studies.

Main Methods:

• QA utilizes a deep learning (DL) model that is concurrently optimized with user annotations in an iterative feedback loop.
• Users interact with an intuitive web interface, accepting correct segmentations or correcting errors to refine the DL model's performance.
• The tool was evaluated across three use cases involving the annotation of cell nuclei, tubules, and epithelial regions in WSIs.

Main Results:

• QA achieved significant efficiency gains, with annotations per second increasing by 102× for nuclei, 9× for tubules, and 39× for epithelial regions.
• The tool maintained high accuracy, with f-scores consistently above 0.95 across all use cases.
• Demonstrated successful annotation of hundreds of thousands of structures across multiple WSIs.

Conclusions:

• Quick Annotator (QA) offers a substantial improvement in annotation speed for digital pathology WSIs.
• The tool's iterative DL approach enables efficient and accurate segmentation of histologic structures.
• QA is a valuable asset for researchers aiming to accelerate biomarker discovery through large-scale WSI analysis.