Quanti.us: a tool for rapid, flexible, crowd-based annotation of images

Area of Science:

• Computational biology
• Bioinformatics
• Image analysis

Background:

• Computational algorithms struggle with complex image analysis tasks.
• Expert annotation is time-consuming and not scalable.
• There is a need for efficient and accurate image analysis solutions.

Purpose of the Study:

• To introduce Quanti.us, a crowd-based image annotation platform.
• To evaluate the accuracy and efficiency of Quanti.us for image analysis.
• To assess the compatibility of Quanti.us annotations with machine learning.

Main Methods:

• Developed and implemented the Quanti.us crowd-based image annotation platform.
• Applied Quanti.us to various medium-throughput image analysis tasks.
• Compared analysis time and annotation accuracy against single expert annotators.
• Evaluated deep learning performance using Quanti.us-derived and expert-derived annotations.

Main Results:

• Quanti.us achieved 10-50x savings in analysis time compared to a single expert.
• Annotations generated by Quanti.us demonstrated accuracy equivalent to expert annotations.
• Deep learning models trained on Quanti.us data performed comparably to those trained on expert data.

Conclusions:

• Quanti.us provides an accurate and time-efficient alternative to computational algorithms for challenging image analysis.
• The platform enables scalable integration with machine learning workflows.
• Crowd-based annotation offers a viable solution for medium-throughput image analysis needs.