Annotating organs without any prior medical skills and/or background

Client Profile

The company provides streamlined medical imaging data management processes for clinical trials.

Challenge

The client required precise segmentation of different human organs on medical imagery for clinical research and trials. They created a small batch with the help of a few researchers, doctors and radiologists but it was not financially viable for the client to ramp up a team with such skills for a critical volume mass required for the research.

Solution

We started a trial with 2 annotators having no medical skills and/or background to assess how long it took to train them in annotation based on human anatomy; to understand the medical imagery and differentiate between different organs. After 1 week of trial, we were confident that we could deliver the tasks with our regular annotators with some training in human anatomy.

After a month, the client chose to scale up the team from 2 to 9 team members (7 annotators and 2 QA). The team managed to minimise the AHT of most organs compared to client’s set benchmark. Also, our QA executives achieved more than 99% accuracy in all organ segmentations till now.

The success in the project was the product of following:

• Client’s involvement early in the project and specially when training and selecting annotators and also setting up the common benchmark for each type of tasks.
• Proper communication process with escalation and screen sharing sessions with the client for new instructions or edge-case scenarios.
• Report edge case scenarios to the client and come up with solutions from experience and discussions.
• Created a compiled version of the short instruction manuals and a FAQ document for recording edge-case scenarios for future reference.
• Detail trackers and dashboards to monitor the progress and where the team needed any assistance.
• Dedicated QCs and QA-process to ensure the utmost quality of work.

Benefits for the Client

The client has a substantial amount of quality datasets for their required research even with a small research budget.