MarkIt: A Collaborative Artificial Intelligence Annotation Platform Leveraging Blockchain For Medical Imaging...

Area of Science:

• Medical Image Processing
• Artificial Intelligence in Healthcare
• Blockchain Technology Applications

Background:

• Supervised machine learning for medical imaging requires large, well-annotated datasets.
• Current annotation tools are insufficient for high-volume processing and scalable systems.
• Lack of efficient annotation methods hinders the development of advanced AI in medicine.

Purpose of the Study:

• To introduce MarkIt, a novel web-based platform for collaborative medical image annotation.
• To leverage artificial intelligence and blockchain technologies for efficient data annotation and reward distribution.
• To assess the efficiency and potential of MarkIt in accelerating the annotation process for medical imaging tasks.

Main Methods:

• Development of MarkIt, a collaborative, web-based annotation tool supporting DICOM and non-DICOM images.
• Integration of AI and blockchain for efficient annotation and user activity tracking for reward calculation.
• Proof-of-concept experiment involving three radiologists annotating 1,000 chest X-ray studies for multi-label classification.

Main Results:

• MarkIt demonstrated efficiency in annotating medical images for classification and object detection tasks.
• The platform facilitated the calculation of inter-rater agreement and estimated dataset value for cryptocurrency-based reward distribution.
• The annotation process was accelerated, and user activities were tracked for fair compensation.

Conclusions:

• MarkIt significantly enhances the efficiency of the medical image annotation process.
• The platform provides a scalable solution for evaluating data value and trading annotation results.
• MarkIt holds potential for future development of robust AI systems in medical imaging through improved data accessibility and annotation.