Our Team blog

Advancing Medical Image Analysis Through Data Science, Artificial Intelligence, and Deep Learning

By Eunice Stennett, David Guet, Hippolyte Heuberger - 02 Mar 2021

Data science is the art of extracting and interpreting relevant information from large amounts of complex data. It has evolved rapidly over the years and is now recognized as a valuable medical image analysis tool. Until recently, the field of computational pathology relied on machine learning (ML) methods. These are highly-supervised algorithms designed to learn human-crafted annotations, and extract and decipher relevant features from their analysis. Keen Eye image-based biomarker analysis, on the other hand, relies on deep-learning (DL) methods. Those semi-supervised or unsupervised algorithms improve feature extraction and interpretability from translational to clinical research.

Data science for computational pathology


More preclinical and clinical R&D organizations in pharma, biotech, and CRO are expanding their in-house data science team. They seek to improve their operational efficiency from biomarker discovery and drug development to patient care at the bedside. This could include applying data science to whole slide image analysis, which requires a high level of expertise in image processing and analysis.

While third-party software can analyze images, most rely on pre-trained ML algorithms for given biomarkers. That means they lack adaptability and require specific training. Keen Eye’s AI platform is designed to allow high flexibility for data scientists to develop and apply their knowledge and expertise in deep learning, without the ties of a closed system and pre-trained algorithms. This enables Keen Eye data scientists to craft custom AI models to decipher complex data and improve knowledge. Clients come to Keen Eye precisely for this expertise and scalability. They all face a common challenge: extracting information from microscopy images to understand a biological process.




This process could be treatment response, signaling pathways, or cell co-expression, for example. Keen Eye agnostic approach to designing DL image-based assay lets clients research numerous therapeutic areas, including oncology, colon inflammation, neurology, ophthalmology, or otology.  Keen Eye data scientists leverage their advanced computing skills to write algorithms that extract data from pixels and provide data modeling. Hippolyte Heuberger, leader of the data science team, explains that “image analysis today is largely a matter of writing code and enabling deep learning through algorithms.” Keen Eye data scientists also explore other fields, from medical research to innovative computational diagnostic tools in a challenging scientific environment. This diverse knowledge enables them to create custom AI model algorithms for each client. Being passionate doesn’t hurt either, as data scientist Louis Jeay adds that “seeing an algorithm start learning is an exhilarating, almost magical moment.”


Developing expertise through diversity


Applying deep learning in computational pathology is a relatively new discipline,  and experience across the industry remains thin. Data scientists at Keen Eye perfect their skills through numerous client projects and research initiatives with medical centers. The wide range of image types, scientific objectives, and medical applications handled provide excellent development opportunities. It also opens the door to transferring models and algorithms created across different projects. For example, Keen Eye successfully adapted models used to transfer a multiplex biomarker to another biomarker.

The collaborative culture at Keen Eye provides data scientists more opportunities to continuously adapt to new challenges and seek creative solutions. The entire team is debriefed on new projects, and members can take the ones they want based on various motivations.



For example, Louis Jeay recently volunteered to lead a multiplexing project because of its applications to fight cancer. The technical originality of the challenge of dealing with multiple biomarker channels attracted him. He believes that this diversity of projects helps team members broaden their horizons. The diverse skill sets among Keen Eye data scientists also enable them to make discoveries and develop reliable models that advance translational and clinical research up to patient care. They share findings and analysis workflows during weekly meetings, discuss the latest published models in AI and medical research, and investigate new techniques or methods that can be applied to their projects. Regular consultation with expert pathologists in different fields such as breast cancer, lung cancer, or other therapeutic areas also increases their understanding of the medical problems they try to solve.

Advancing medical image analysis through AI and DL


The potential of artificial intelligence and deep learning in medical image analysis is promising. As data scientists and medical researchers increase their collaboration, more translational and clinical projects will succeed. This active listening between two fields that seem so far apart helps data scientists strengthen their expertise and knowledge in digital image analysis based assay. While this technology is still in its early days, its benefits for society are clear. Through more research and projects, data scientists will continue to develop their expertise and create new models. Keen Eye understands it and is proud to participate in initiatives that advance medical image analysis.


Eunice Stennett - Former Chief Marketing Officer 

David Guet - Product Application Specialist 

Hippolyte Heuberger - Head of Data Science