English
Español
Pixelabs is part of the public-private consortium formed by 16 entities that seek to accelerate and improve clinical and health research in Spain.
This consortium is working on the Tartaglia project framed within the Artificial Intelligence R&D Missions program of the Spain Digital 2025 agenda and the National Artificial Intelligence Strategy, Tartaglia is funded by the European Union – NextGenerationEU and the Recovery, Transformation, and Resilience Plan.
Pixelabs is leading the work package on research into predictive models for automatic screening of diabetic retinopathy using convolutional neural networks. Being able to publicize this type of clinical research is a way to contribute to the advancement of society and for us it is very important that part of our technological efforts as a company have a positive impact on institutions, society and citizens.
Eneka Carnicer, coordinator of the work package we are working on, tells us her vision: “We are excited to be able to participate in a project of this scale, with which we can contribute to the constant improvement of the medical world. From Pixelabs, and together with all our fellow travelers in the consortium and especially with the clinical part with which we are working more closely on a daily basis (Fundación Rioja Salud, Instituto de investigación Sanitaria la Fé and the Servicio Gallego de salud) we are contributing all our potential, both in management and development, to obtain positive results that we can put into practice as soon as possible”.
The project consists of different potential use cases. Each of them is led by different members of the consortium (with a total of 5 use cases), such as research on the early detection of Alzheimer’s disease; early detection and diagnosis of pathologies related to heart failure; early detection of clinically significant prostate cancer through AI techniques; simulating the evolution of chronic and complex diseases or the detection of diabetic retinopathy in early stages. And it is in this last use case where Pixelabs is actively involved.
Ecosystem, data and objectives
In the healthcare sector, the implementation of technology must pass strict evaluations and comply with the standards and certifications required in this environment. In addition, it is always subject to a final validation by a healthcare professional. In this way, Artificial Intelligence is presented as an ally and new tool to help improve repetitive tasks as well as to accelerate processes that until today are manual.
In order to train models with real data, the project’s premise is to use the federated data model (known as federated learning). This will allow the algorithm to travel to the data centers without having to leave each of the centers in order to preserve maximum privacy and security.
Figure representing a Federated Data Network
Tartaglia’s goal is not only to promote clinical and health research, but also to achieve new methods for the early detection of diseases. It also demonstrates that it is possible to adopt this type of technology, which is not yet mature in a sector with complex regulations and high standards. Other objectives of the project are to accelerate the development of prevention technologies, support for medical diagnostics and professionals. The fields of application and benefits of a project of this magnitude are many, from the quality of research and its improvement to the launching of final solutions for rapid implementation. The aim is also to improve the reliability of the algorithms, always with technical validation and certified as medical products.
Convolutional networks for early detection of diabetic retinopathy
An algorithm will be designed and trained in a federated manner with distributed data. This will comply with security and privacy regulations; the model will travel to the data centers, perform training and return to the trained central node.
This research will facilitate the implementation of a consistent screening system, which under the supervision of a professional in the area, will help to decongest the processes that until now have been carried out manually. Within the study, different clinical markers will be used to determine whether retinopathy exists and to what degree. Anonymized data from images of the patient’s fundus and structured data extracted from different analyses are available. We will seek to train models that use both data sources to provide the best possible answer. Similarly, the images will be analyzed with computer vision techniques to extract decisive variables such as the presence of hemorrhages, neovessels or microaneurysms and will be provided to the models. This image analysis seeks to replicate the procedure used by clinicians when categorizing a patient’s image. Therefore, in this task we will work hand in hand with the clinicians to understand their analysis work and translate it to the algorithms. The models will be trained in a federated manner seeking to maximize sensitivity and specificity.
Tartaglia is a project financed by the European Union – NextGenerationEU. However, the views and opinions expressed are solely those of the author(s) and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.
TARTAGLIA (con Nº de referencia TSI-100205-2021-11) está enmarcado dentro del programa Misiones de I+D en Inteligencia Artificial de la agenda España Digital 2025 y de la Estrategia Nacional de Inteligencia Artificial, y está financiado por la Unión Europea a través de los fondos NextGenerationEU. Las acciones realizadas se reportan al Ministerio para la Transformación Digital y de la Función Pública (Nº expediente MIA.2021.M02.0005), correspondiente a los fondos del Plan de Recuperación, Transformación y Resiliencia.
