By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.
February 15, 2024

Optimizing Data Annotation: Understanding the Process Behind the iToBoS & isahit Project

February 15, 2024

The Crucial Role of Data Annotation in the iToBoS Project

Artificial intelligence (AI) is at its peak right now. And we know its applications in healthcare are numerous. But when it comes to health data labeling, precision is necessary. This is in fact the case in projects like iToBoS, where AI technology, with a mix between technology and Human-in-the-loop technics, is used to accurately detect melanoma. This article digs into the the isahit annotation process and workflow that we applied to iToBoS, enlightening our pivotal role as a data labeling services company, in producing quality annotations and driving successful outcomes.

Introduction to the isahit Annotation Process applied to itobos

Data annotation is the key in AI model training, particularly in complex tasks such as identifying melanoma from skin photography and images. Within the iToBoS project on which we work on at isahit, the annotation workflow is tailored to enhance the efficiency and accuracy of melanoma detection through several key strategies:

  • Assigning Specific Color to Lesions: Annotators assign a distinct color to lesions, facilitating easy identification and categorization during the training process of the AI.
  • Annotation of Lesions Based on Size: Only lesions surpassing a predetermined size threshold are annotated. This targeted approach ensures that annotators focus uniquely on significant lesions, reducing unnecessary annotation workload while maintaining accuracy.
  • Degree of Sun Damage Assignment: Additionally, sun damage levels are manually assigned through image annotation. This granular labeling at the individual tile level ensures comprehensive data enrichment, enabling the AI model to discern nuanced features crucial for accurate diagnosis.

These image annotation technics and steps in the workflow not only streamline the annotation process but also contribute to the precise training of AI models for effective melanoma detection.

Annotator Responsibilities and Workflow

Central to the success of the annotation workflow within the iToBoS project is the collaborative communication between annotators, reviewers, and the oversight provided by the Customer Success Management (CSM) team located in Paris. Using a multi-teams approach ensures accurate and consistent annotations, which are crucial for effectively training AI models in melanoma detection.

  • Annotators are “recruited” based on their skills acquired through the digital academy provided by isahit, in addition of some instruction provided on the V7 tool and a specific training to the iToBoS project. This ensures their proficiency in accurately annotating melanoma-related features in images.
  • Reviewers, experienced in image annotation tasks, work alongside annotators to increase the quality of their annotations. Their meticulous and precise feedback ensure high standards throughout the annotation process.
  • The CSM team (a team of engineers located in Paris), led by a project manager, oversees the entire workflow, ensuring alignment with project objectives. Additionally, a person responsible for managing the labelers, a former annotator, provides valuable insights and support, enhancing the efficiency of the process.

Quality Control and Feedback Mechanisms

Maintaining annotation accuracy requires regular check-ins and feedback loops. An open chat allows annotators, reviewers, and project managers to exchange insights and address challenges. Weekly meetings with clients and medical professionals ensure the annotated data meets specific requirements. Clinicians from the consortium review the data post-annotation for accuracy and relevance.

This collaborative approach ensures high-quality annotated data for effective AI model training related to Melanoma Detection.


In conclusion, the meticulous image annotation process structured on the iToBoS project, facilitated by isahit's expertise and collaborative workflow provided by V7 labeling tool, underscores the critical role of data annotation in training AI models for melanoma detection. By leveraging specialized training, rigorous quality control measures, and proactive feedbacks, isahit and iToBoS are advancing the frontier of medical diagnosis and treatment.

You might also like
this new related posts

Want to scale up your data labeling projects
and do it ethically? 

We have a wide range of solutions and tools that will help you train your algorithms. Click below to learn more!