top of page

Join the DigiLut Challenge: Advancing Lung Transplant Rejection Detection

🚀 Join the DigiLut Challenge: Elevate Lung Transplant Research!

Exciting News for The Data Scientists Community! is thrilled to announce the upcoming launch of a cutting-edge data challenge, 🌟"DigiLut: Detection of Graft Rejection Following Lung Transplantation." This challenge is led by Pr. Antoine Roux from Foch Hospital, who is a pneumologist and has a PhD in immunology. Data scientists across the globe are invited to participate in this pivotal research aimed at improving outcomes for lung transplant recipients.

This challenge is organized by the Foch Hospital, in partnership with the Health Data Hub and financed by the Banque Publique d’Investissement (Bpifrance) . The challenge is sponsored by the “Grand Défi: Improvement of medical diagnoses through Artificial Intelligence” program led by the French Secretariat General for Investment.

Imagine unraveling the mysteries of lung graft rejection (picture above) using cutting-edge machine learning. Your mission? Develop mind-bending algorithms that decode digitized lung biopsy slides, revealing the hidden secrets of graft rejection.

Example of lung biopsy slide that will be used in the DigiLut challenge

Sign up for the DigiLut Challenge today , we will notify you as soon as the competition launches!


Introduction of the DigiLut Data Challenge by Dr. Antoine Roux (video from Health Data Hub)

The Challenge Background

Lung transplantation is the only curative treatment for end-stage respiratory diseases, with around 4,000 transplants performed globally each year. Despite its life-saving potential, transplantation success critically hinges on one major factor: graft rejection. Rejection is a complex and heterogeneous process, currently identified through histological examination, which suffers from poor reproducibility.

Foch Hospital is France's leading lung transplant center, and one of the top in Europe: over 1,000 transplants have been performed to date, including more than 70 per year over the past 3 years, with an active file of almost 600 patients. Hôpital Foch is also a leader in lung transplant research, with a particular focus on ex-vivo reconditioning and rejection.

Digital pathology (a sub-field of pathology that focuses on data management based on information generated from digitized specimen slides) is revolutionizing conventional anatomo-pathology, both in day-to-day practice (training interns, sharing expertise with colleagues) and with the prospect of diagnostic aid tools based on Artificial Intelligence

(AI). In the future, we expect these AI tools to improve the quality and reliability of diagnoses.

The Objective of the challenge : IMPROVE Acute rejection diagnosis

  • Dive into a treasure trove of digitized biopsy data.

  • Unleash your AI prowess to pinpoint pathological zones.

  • Ascend the leaderboard by detecting graft rejection like never before! 🏆

To develop a medical decision-support tool that will consolidate the diagnosis of graft rejection episodes in lung transplantation. Competitors will develop algorithms capable of identifying pathological zones in digitized transbronchial biopsy slides, determining the presence and severity of graft rejection, based on an unique image bank of digitized graft biopsies from lung transplant patients followed at Foch.

An international panel of expert pathologists (USA, Canada, France, Germany, Italy, UK, Netherlands) generated the annotations of the zones of interest and their annotation in accordance with the LASHA Grid “Lung allograft standardized histological analysis”. 🌎🔍

The aim is to apply state-of-the-art machine learning approaches to the analysis of digitized biopsy slides and pathology characterization, to develop an algorithm capable of generating region-of-interest boxes for type A lesions on new, non-annotated slides, using both the annotated and non-annotated data provided.

The use of non-annotated data should be optional, to leave teams free to use the approaches they wish. However, using more data can be beneficial for the performance of different models.

For more technical information about the background and goal of the challenge check out the following presentation from Dr. Roux :

🌟 Why Participate? Unlock the Power of Data for Lung Transplant Patients!

Calling all data scientists, statisticians, and AI specialists around the world to take a part of this exciting challenge, and shape Shape the future of lung transplant patients’ care!

  • Impact: Your skills can help improve survival rates for lung transplant recipients. 🏥💓

  • Challenge: Tackle a massive dataset of high-dimensional medical images—approximately 8 TB in size. 📊🔬

  • Resources: Take advantage of free CPU and RAM resources in our hosted Jupyter notebook environment. Need more power? Paid GPU resources are available for complex computations! 💻⚡

  • Collaboration: Collaborate with leading data scientists and industry experts, including Pr. Roux, who is specialized in the clinical and immunological aspects of lung transplantation. 🌍🤝

Example of a lung biopsy

🔍Key Details

  • 📅 Start Date: May 31st, 8 AM CEST

  • Duration: 2 Months

  • 🏆 Prize : The top-three performing solutions developers will share a €25,000 prize money.

    • 1st Place : 12 000 €

    • 2nd Place : 8 000 €

    • 3rd Place : 5 000 € 

For this challenge , provides a Jupyter notebook equipped with free CPU and Memory resources. You can manage and analyze the vast dataset directly inside the Jupyter environment, create a compressed version of your dataset and download it for local model training.

📊A baseline model will be provided by to all participants, illustrating a basic approach to dataset compression and model training.

Sign up today, the challenge is open to ALL!

Sign up for the DigiLut Challenge today ( )

Mark your calendars for May 31st—that's when the dataset will be accessible! 🗓️🔭

Don't miss out—let's innovate together for a brighter, healthier future!

In the meantime, if you have any question feel free to reach out to us at

17 views0 comments


bottom of page