Cancer Cell Consensus Map

Computer Vision

Pharma

1 min read

We’ve utilized computer vision and unsupervised learning algorithms to enhance the analysis of microscopic images, assisting doctors and healthcare providers in the diagnosis and treatment of various diseases. Our clustering algorithm creates a consensus map for hand-labeled cancer cell images, improving diagnostic accuracy and generating valuable training data for machine learning models.

‍

Book a call with us.

Choose a date and let’s talk business.

Book a call

The problem overview

Our training dataset comprised heterogeneous microscopic images, where cancer cells were hand-labeled by doctors. The primary challenges included managing the complexity of modeling these images, handling large data sizes, and ensuring scalability of the solution.

Our solution

We developed an analytical algorithm for cell labeling consensus using an unsupervised machine learning approach, specifically HDBSCAN, to automatically identify optimal clusters. Multiple clustering techniques, including distance-based, hierarchical, and density-based methods, were evaluated for their effectiveness, scalability, and computational efficiency. DBSCAN was ultimately selected for its ability to accurately target cancer cells while effectively removing outliers. The resulting labels were augmented with domain knowledge and hierarchical annotation rules to create a robust consensus map.

The results

Exceptional results were achieved through innovative solutions, turning the challenge at hand into measurable success.

Greater accuracy

Our algorithm enabled medical professionals to achieve greater accuracy and consensus on cell labeling through a shared map, potentially improving the targeting of cancer cells.

Enhanced Applicability

The use of unsupervised machine learning methodologies allowed for the development of an algorithm that does not rely on annotated/ground truth data for every new data problem, enhancing its applicability in the field of cell labeling.

Valuable Training Data

The consensus map can now serve as a labeled dataset for other machine learning algorithms, facilitating further advancements in medical diagnostics and research.

Liked it?

Download this success story

Get in touch and apply it in your company

Ready to transform your business? Contact us today to learn how we can apply these solutions to your company’s challenges.

Boost my company

Partners

Awards

Choose a journey 

Start here Build your AI ecosystem Scale your AI

Resources

Blog Pressroom Events Success Stories

Join us Contact Us

Cookies Policy