English
Español
TARTAGLIA project (funded by the European Union – NextGenerationEU and the Recovery, Transformation, and Resilience Plan), which aims to create a federated network with artificial intelligence to accelerate clinical and health research in Spain, also seeks to design new methods for the early detection of diseases. Framed within the R&D Missions in Artificial Intelligence program of the Digital Spain 2025 agenda and the National Artificial Intelligence Strategy, it consists of 5 potential use cases, being the detection of diabetic retinopathy in early stages where Pixelabs is actively working together with the clinical and technical teams of Fundación Rioja Salud, Instituto de Investigación Sanitaria la Fe and the Galician Health Service (in which the Galician Knowledge Agency and the Ophthalmology Service of the Ferrol Health Area work and play an active part). In this particular case, it highlights the current need to find better and innovative methods based on artificial intelligence for the early detection of pathologies, as well as in the improvement of medical diagnostic support.
What is diabetic retinopathy?
The clinical teams working on the project give us the keys to understanding what diabetic retinopathy is, what causes it, the current diagnostic challenges and why research such as that being carried out in this area is so important.
As they explain, diabetic retinopathy (DR) is one of the complications of Diabetes Mellitus (DM) and the most important cause of blindness and visual impairment in people of working age in industrialized countries. Moreover, the incidence of DM continues to grow (1), and it is predicted that this disease will have an even higher incidence in the future (2). According to the WHO, by 2030 there will be 500 million people with DM, 25% of whom have some type of retinopathy and 2-10% have diabetic macular edema (3-6).
DR is defined as the microvascular abnormalities present in of people with DM, which can be determined by direct examination of the retina or through photography or retinography (7). The international classification divides DR into mild, moderate, severe and proliferative. In mild DR only microaneurysms, which are small saccular dilatations of the retinal capillaries, are observed. In moderate DR, hemorrhages, exudates or venous arrosariations (retinal venules of irregular caliber with successive areas of dilatation and stenosis) are also present. In severe DR, hemorrhages and venous arrosariation are more severe and intraretinal microvascular abnormalities may also be found. In proliferative forms, neovessels are observed, very fragile vessels that can rupture and produce large bleedings inside the eye (8). In all these cases, early diagnosis and intervention have proven to be able to improve the prognosis of patients (9); however, in order to implement them as standard practice, resources such as telemedicine are needed, which has proven to be a widely used alternative for DR screening (10,11).
Teleophthalmology, using photography (retinography) for screening diabetic patients, has proven to be not only effective but also cost-effective (12) and accepted by patients (13, 14). The use of this technology has contributed to displacing DR to second place as a cause of blindness in places where it has been possible to establish a systematic screening program(15, 16).
Challenges and automatic screening using AI
In our project presentation article we outlined the role that Pixelabs is playing, working to design and train an algorithm in a federated fashion with distributed data: where the model will travel to data centers to collect all the information needed to train the algorithm that will aid research into a consistent screening system. In our work as coordinators of this package, and in the first approach in this research, the algorithm will only screen automatically between those images that show signs of the disease and those that do not, and then add all the necessary features to complete and/or add new processes to the diagnosis.
This whole process is not without its challenges. IIS La Fe itself, where the algorithm will work with real-life data from patients, stresses that one of the biggest challenges is “to achieve a good database of good quality images, adequate labeling of these images that will enable the development, through artificial intelligence, of an effective system for classifying DR that will serve as a support for daily clinical practice”. And from SERGAS, they see it as indispensable, “to be able to establish criteria for ‘referable’ diabetic retinopathy or, on the contrary, that does not require referral to specialists for intervention, and depending on its severity, the desirable attendance times, so that it can function as an alert system and even ideally, be able to manage the relevant appointments.” And of course as a transversal and indispensable element “to be able to create an accessible tool, with CE marking and that meets all the legal and ethical requirements for clinical practice”, one of the biggest challenges of the project.
It is very important to point out that in addition to the development of the algorithm, the technical teams are carrying out other crucial tasks for achieving the project’s objective, in which other companies and entities of the consortium are participating, such as: implementation of the global infrastructure to provide the computing, coordination and application nodes, essential for carrying out research work with anonymized data (GMV); study and definition of the variables to be taken into account and harmonization of the data, a basic step for taking into account all the data needed for training and standardizing them (Veratech). It should be noted that in the development of the screening algorithm, different tests are being carried out with the images of the initial datasets provided by the different data providers (IIS La Fe, SERGAS and FRS) in order to analyze the possible lines of research (computer vision / predictive learning application) that could be applied; and to study how to humanize the results obtained with the model already developed. At this point, an interface has already been defined for clinical specialists that will serve as a validation tool for the algorithm (OPPINO).with all the objectives and challenges set, it is clear that TARTAGLIA is one of the most relevant projects being carried out today in the national territory to promote AI research in health systems and that it aspires to be a benchmark within the health sector.from our website, we will update all relevant information to publicize the progress of this.
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.
References and bibliography:
1. IDF diabetes atlas – Home [Internet]. [cited 2019 Jul 23]. Available from: https://diabetesatlas.org/
2. Li JQ, Welchowski T, Schmid M, Letow J, Wolpers AC, Holz FG, Prevalence, incidence and healthcare needs. :30.
3. Pareja-Ríos A, Serrano-García MA, Marrero-Saavedra MD, Abraldes-López VM, Reyes-Rodríguez MA, Cabrera-López F,. Guías de práctica clínica de la SERV: Manejo de las complicaciones oculares de la diabetes. Retinopatía diabética y edema macular. Arch Soc Esp Oftalmol [Internet]. 2009 Sep [cited 2018 Jul 19];84(9). Available from: http://scielo.isciii.es/scielo.php?script=sci_arttext&pid=S0365-66912009000900003&lng=en&nrm=iso&tlng=en
4. Yau JWY, Rogers SL, Kawasaki R, Lamoureux EL, Kowalski JW, Bek T, Global prevalence and major risk factors of diabetic retinopathy. Diabetes Care. 2012 Mar;35(3):556–64.
5. Ding J, Wong TY. Current epidemiology of diabetic retinopathy and diabetic macular edema. Curr Diab Rep. 2012 Aug;12(4):346–54.
6. Klein R, Klein BE, Moss SE, Cruickshanks KJ. The Wisconsin Epidemiologic Study of Diabetic Retinopathy: XVII. The 14-year incidence and progression of diabetic retinopathy and associated risk factors in type 1 diabetes. Ophthalmology. 1998 Oct;105(10):1801–15.
7. Teo ZL, Tham YC, Yu M, Chee ML, Rim TH, Cheung N, Global Prevalence of Diabetic Retinopathy and Projection of Burden through 2045: Systematic Review and Meta-analysis. Ophthalmology.2021;128(11):1580-91.
8. Sociedad Española de Retina y Vítreo. Guía 3 de Práctica Clínica y Monografías. Manejo de las Complicaciones Oculares de la Diabetes: Retinopatía Diabética y Edema Macular. 2021.
9. Bäcklund LB, Algvere PV, Rosenqvist U. New blindness in diabetes reduced by more than one-third in Stockholm County. Diabet Med J Br Diabet Assoc. 1997 Sep;14(9):732–40.
10. Scanlon PH. Update on Screening for Sight-Threatening Diabetic Retinopathy. Ophthalmic Res. 2019 May 27;1–7.
11. Alonso Porcel C, Martínez Ibán M, Arboleya Álvarez L, Suárez Gil P, Sánchez Rodríguez LM. [Diabetic retinopathy screening programme in primary health care. Diagnostic concordance between family and eye care practitioners]. Semergen. 2016 Sep;42(6):357–62.
12. Rein DB, Wittenborn JS, Zhang X, Allaire BA, Song MS, Klein R, The cost-effectiveness of three screening alternatives for people with diabetes with no or early diabetic retinopathy. Health Serv Res. 2011 Oct;46(5):1534–61.
13. Kumari Rani P, Raman R, Manikandan M, Mahajan S, Paul PG, Sharma T. Patient satisfaction with tele-ophthalmology versus ophthalmologist-based screening in diabetic retinopathy. J Telemed Telecare. 2006;12(3):159–60.
14. Valpuesta Martin Y, Pacheco Callirgos GE, Maroto Martín TM, Piriz Veloso M, Hernández Santamaría S, López Gálvez MI. Satisfaction of patients and primary care professionals with a teleophthalmology-based screening programme for diabetic retinopathy in a rural area in Castilla y León, Spain. Rural Remote Health. 2020 Jan;20(1):5180.
15. Scanlon PH. The English National Screening Programme for diabetic retinopathy 2003–2016. Acta Diabetol. 2017;54(6):515–25.
16.Liew G, Michaelides M, Bunce C. A comparison of the causes of blindness certifications in England and Wales in working age adults (16-64 years), 1999-2000 with 2009-2010. BMJ Open. 2014 Feb 12;4(2):e004015.
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.
