We blend human insight with safe and fast data access to train high performance models.

Collaborate

We combine cutting-edge technologies, like federated learning, with critical clinical expertise and a network of academic researchers. This allows research to be accelerated in a safe and privacy-preserving way.

Challenges

What if we could share global data insights seamlessly?

Strict regulation to protect patient data

Regulatory requirements such as GDPR and HIPAA and a fear of compromising competitive assets or patient privacy makes data sharing complex.

Disjointed data

Complementary datasets cannot be easily accessed and combined for model training.

Data harmonization

Differing standards of data structure, digitization, storage, privacy compliance, and geographical distribution stack up and compound each other.

Strict regulation to protect patient data

Regulatory requirements such as GDPR and HIPAA and a fear of compromising competitive assets or patient privacy makes data sharing complex.

Disjointed data

Complementary datasets cannot be easily accessed and combined for model training.

Data harmonization

Differing standards of data structure, digitization, storage, privacy compliance, and geographical distribution stack up and compound each other.

Owkin’s solution

We unite Clinical, IT and Data specialists and groundbreaking technologies within an integrated network to overcome these challenges for faster and more effective research.

Federated Learning

Connect disjointed data to build accurate models without sharing or pooling data with Federated Learning

Unlocking access

Federated learning powers precision medicine by unlocking access to the right quality and quantity of data to train high-performance models.

Data stays local

With Federated Learning the data stays local - only models and insights travel between the servers, protecting privacy and ensuring compliance.

More accurate models

This technology allows research communities to benefit from access to more datasets, capturing the full heterogenous picture, resulting in more robust and accurate models.

Testimonial

01 / 01

“In terms of security, Owkin provided the methodology, as well as the external audits which, for us, are a guarantee of security. Owkin’s philosophy of excellence and speed is closely aligned to that of cancer centers – there’s a great cultural fit. Today, if we wanted to apply AI to our data, I wouldn’t think twice – I’d call Owkin.”

Thierry Durand

Head of IT Systems at Centre Léon Bérard

Substra (formerly Owkin Connect)

Expand access to health data while protecting privacy with our open source Federated Learning software, Substra

Substra technology powers collaboration in healthcare by ensuring data confidentiality and compliance.

Substra for clinical research

The HealthChain project unites AI experts and medical researchers to foster medical innovation in anatomo-pathology, dermato-oncology and fertility while preserving patients' privacy. Requirements such as GDPR and HIPAA and a fear of compromising competitive assets or patient privacy makes data sharing complex.

BioPharma companies and healthcare institutions deploy Substra in various settings for clinical research and drug discovery.

Substra for drug discovery

The MELLODDY Consortium tackles the $1.3Bn challenge of bringing a new drug to market. This project enhances predictive machine learning models on the decentralized data of 10 pharmaceutical companies, without exposing proprietary information.

Substra documentation

Discover how to use and deploy Substra in a real-world setting.

09/10

Visualize predictions in the slide viewer, per individual patient, alongside associated clinical information and annotations.

10/10

Heatmaps are overlayed onto slides to assist the user in finding regions of high concordance with the predicted features. Whole slide images can easily be zoomed into with linear and diverging heat maps turned on/off at a click of a button.

01/10

Once the federated run is complete, the researchers can examine the overall performance of the federated model throughout the evaluation steps.

02/10

Registering a new dataset can be done easily via the graphical user interface by a medical researcher, or via the python library by IT or data science specialists.

03/10

Owkin Connect is powered by a Python library which makes it possible to operate the platform from your Python environment. It also comes with documentation.

04/10

Owkin Connect's home page will display the federated learning channels the user has access to. Each channel consists of one or more datasets which can be used for training a federated model to provide specific results.

05/10

Once a new dataset has been uploaded, Owkin Connect provides cohort metrics and an easy way to visualize all data within the cohort before launching any models.

06/10

Users can track the model training development (called the compute plan), with key metrics at each step of the process.

07/10

Users can also visualize the algorithm code directly in the user interface for each federated model trained. This makes all computations reproducible.

08/10

Users can also return to the cohort to examine the results in more detail. For example, slide by slide to examine which tiles were the most predictive of certain features.

09/10

Visualize predictions in the slide viewer, per individual patient, alongside associated clinical information and annotations.

10/10

01/10

Once the federated run is complete, the researchers can examine the overall performance of the federated model throughout the evaluation steps.

02/10

Registering a new dataset can be done easily via the graphical user interface by a medical researcher, or via the python library by IT or data science specialists.

Substra documentation

Discover how to use and deploy Substra in a real-world setting.

01/10

Once the federated run is complete, the researchers can examine the overall performance of the federated model throughout the evaluation steps.

02/10

Registering a new dataset can be done easily via the graphical user interface by a medical researcher, or via the python library by IT or data science specialists.

03/10

Owkin Connect is powered by a Python library which makes it possible to operate the platform from your Python environment. It also comes with documentation.

04/10

05/10

Once a new dataset has been uploaded, Owkin Connect provides cohort metrics and an easy way to visualize all data within the cohort before launching any models.

06/10

Users can track the model training development (called the compute plan), with key metrics at each step of the process.

07/10

Users can also visualize the algorithm code directly in the user interface for each federated model trained. This makes all computations reproducible.

08/10

Users can also return to the cohort to examine the results in more detail. For example, slide by slide to examine which tiles were the most predictive of certain features.

09/10

Visualize predictions in the slide viewer, per individual patient, alongside associated clinical information and annotations.

10/10

Consortia

Accelerate your breakthrough research by building collaborative projects in a privacy-preserving way with Federated Learning.

Free up resources from data pooling and management work packages to focus on what really matters: your scientific question. Our decentralized-AI technology lets you set up secure and efficient consortia: data is never shared between partners; only the encrypted models travel between the data centers.

Data & KOL network

Solve the biggest questions in cancer by combining the very best biomedical expertise with artificial intelligence. Owkin has built a world-class research network of leading academic centers, KOLs and expertly-curated, high-quality patient data.

Owkin’s data network contains cohorts of rich longitudinal datasets that medical experts have curated for fit-for-purpose standard and AI analysis. They span many modalities; omics, imaging, histology, and clinical data, comprehensively capturing oncology’s heterogeneous nature.

Collaboration in healthcare

Owkin is proof that collaboration in healthcare could make precision medicine a reality.

Through data access and KOL insight, we’ve created breakthrough models

Professor Fabrice André, MD, PhD

Director of Research Gustave Roussy

"Owkin is leading the use of AI for prediction and drug discovery in cancer. It has developed methods to detect predictive signals that the human brain can’t. We developed a pathology tool to predict relapse in early stage breat cancer patients in one year, while we struggled for many years to do it with other teams."

Professor Francoise Galateau Salle, MD, PhD

Head of MESOPATH in the Department of Biopathology Cancer Centre Léon Bérard

“We had a wonderful experience with Owkin…they created an algorithm and the results exceeded our expectations. They were able to identify details from the histology slides that we knew about but had never previously recognized as significant prognostic indicators or biomarkers of treatment decisions.”

Professor William R. Jarnigan , MD

Chief of Hepatopancreatobiliary Service at Memorial Sloan Kettering Cancer Center

“I am excited to be part of this cutting-edge research collaboration that will have a positive impact on how clinicians evaluate and treat patients with intrahepatic cholangiocarcinoma, and potentially other cancers, as well.”

Professor Fabrice André, MD, PhD

Director of Research Gustave Roussy

Professor Francoise Galateau Salle, MD, PhD

Head of MESOPATH in the Department of Biopathology Cancer Centre Léon Bérard

Professor William R. Jarnigan , MD

Chief of Hepatopancreatobiliary Service at Memorial Sloan Kettering Cancer Center

Some of our partners