Keen Eye
Propelling Pharma on the Computational Pathology Revolution

In the wake of the COVID-19 pandemic, pathology—like all other branches of healthcare—was disrupted almost overnight. The social distancing mandates forced the healthcare industry to limit on-site work to just one or two pathologists at a time or sometimes even ship slides to pathologists’ homes for testing. This hastened shift to remote work led the labs to increasingly consider digital solutions that foster remote interactions and data sharing between pathologists. Thus, pathology, a field of study that had remained largely unchanged for almost a century and a half, had to accelerate its transition to the digital realm, marking a paradigm shift in how it is carried out!

With digital pathology, slides are moving from the narrow field of a microscope to broader, diverse, and more precise digital whole-slide imaging (WSI)—opening a completely new window of opportunities. It is paving the way to novel approaches like computational pathology, which can revolutionise not only pathology as a practice but also healthcare as a whole by transforming healthcare, biopharma services, and drug makers.

Today, increasingly, pathologists are witnessing the upside of AI-driven computational pathology to automate, augment, and even provide predictive insights during pathological processes. However, the road to making that possible is like walking a tightrope, thanks to the constantly evolving regulatory framework. That is where Keen Eye is setting a benchmark in the domain by simplifying the adoption of computational pathology and guiding pathologists and drug makers in driving better drug development.

To that end, Keen Eye puts forth a unique model to implement computational pathology; it combines proprietary deep learning with cloud-based technologies while adhering to regulatory compliance. “When it comes to applying such a mature AI technology, there is a wide gap between creating great AI algorithms and being able to deliver a truly scalable AI-based product in a regulatory environment, as in clinical trials,” notes Sylvain Berlemont, PhD, founder and CEO of Keen Eye. “Our core mission is to bridge that gap; our one-of-its-kind platform delivers the best quality AI insights to every stakeholder of a study (sponsors, CROs, and pathologists alike) while complying with the regulatory standards of the market,” Berlemont adds.

Keen Eye’s web-based computational pathology platform, in that regard, has already cemented its cornerstone—playing a phenomenal role by offering drug makers the much-needed agility and efficacy in operations while staying compliant with the regulations. Notably, Keen Eye’s platform has been enabling pharmaceutical companies to streamline the design, development, and validation of AI algorithms serving their biomarker assays, and at the same time, provide access to remote pathologists and facilitate data sharing from various sources. Pharmaceutical companies, as a result, can not only deploy the AI algorithms into their clinical studies (on their premises or with their preferred CROs) but also speed up the delivery of quantitative, standardised, and regulatory-compliant data to support clinical trials.

The 360-degree Approach to Computational Pathology

Be it optimising translational research, enhancing CDx/ IVD assay development, or accelerating clinical trials, Keen Eye’s platform is built from the ground up, considering every touchpoint of the research process.

Keen Eye’s deep learning-powered platform has been playing a phenomenal role in digital pathology by offering pharmaceutical companies the much-needed agility and efficacy in operations for image analysis while staying compliant with the regulations

In translational research, researchers are seeking new ways today to characterise their samples with robust and quantitative insights. The platform supports multiple imaging modality in that regard. The team at Keen Eye has developed a series of deep learning algorithms to quantify data from very complex images and provide the necessary readouts that researchers are looking for. Besides, the researchers can benefit from the platform’s powerful annotation system that can handle millions of annotations. With it, researchers can perform automatic analysis on whole tissue sections with extremely accurate quantifications. Keen Eye’s product also offers a wide range of analyses for multiple fields of application (oncology, neurology, inflammatory diseases, etc.). Furthermore, each of these analyses performed on the platform can be shared securely between different partners of the study, such as pathologists, data scientists, and clinical managers.

At the same time, as the platform completely supports Good Clinical Practices (GCP) and Good Laboratory Practices (GLP) standards, histopathology analysis can be performed directly in it with a validated application. Interestingly, the image analysis step is swift, reproducible, and standardised in the platform, thanks to its robustness of deep learning algorithms. The platform also has a highly controlled and customised workflow for each study purpose, which guides the users step-by-step. In addition, data integrity is ensured by a complete audit trail that offers traceability to each action performed on the platform. At last, when a study is finished, the final validation of data is performed with electronic signatures.

Taking Clinical Research to the Next Level

How Keen Eye’s deep learning-powered platform is enhancing the efficacy of clinical research can be best reflected through a recent client success story. One of the clients of Keen Eye has been undertaking a project to explore the tumour microenvironment in tissues with a final objective to establish a clear correlation between specific proteins expression and localization and a clinical outcome. To achieve this goal, understanding the spatial organisation of multiple biomarkers was required through immunofluorescence-based multiplex staining of the different samples. But the client faced multiple challenges on that front. First, they wanted to set up an agnostic analysis tool that was able to sustainably manage heterogeneity between different scanning systems and was powerful enough to extract critical information from a haystack of biomarkers. Then, it also had to be accurate, scalable, and accessible to different people at different places.

Keen Eye was able to create and validate a strong AI model that met every requirement of the project with unparalleled accuracy for every biomarker’s detection. “Now the next step of the project can be triggered to scale up the number of samples,” comments Berlemont.

This is but one of the numerous successful stories of Keen Eye. Today, tissues and biomarkers that have to be analysed are ever-increasing, especially with healthcare rapidly inching towards personalised medicine. Keen Eye, with its extensive deep learning competencies, is well poised to play a pivotal role in that endeavour. Concurrently, Keen Eye is vying to integrate new modalities, including spatial genomics and transcriptomics, to fuel this transition towards personalised medicine. “We are convinced that the future of clinical research will be built upon deep learning solutions, transforming the way image analysis is performed,” remarks Berlemont. “Our ambition is to cater to those maximum number of projects, without forgetting the quality of each one,” he concludes.