top of page

Retina AI, learnings in fAIr LAC Jalisco for the use of AI in the detection of diabetic retinopathy

Constanza Gómez Mont, Founder and Principal of C Minds; and José Roberto Mejía, Coordinator of fAIr LAC Jalisco for C Minds


  • C Minds accompanied the implementation of Retina IA, a pilot use case for the early identification of diabetic retinopathy through AI systems in the public health sector in the State of Jalisco.

  • This initiative strategically contributed to Jalisco consolidating itself as one of the leading entities in the piloting of AI pilots for public innovation.

  • 1,053 patients were treated in three health centers in the Guadalajara Metropolitan Area, resulting in reducing diagnostic time from 6 months to 8 weeks and generating valuable learnings on the use of AI to provide health care in Mexico.



****

Diabetic retinopathy is an ocular condition derived from diabetes mellitus, generally detected in very advanced stages, causing irreversible damage such as blindness. This chronic and progressive condition has a prevalence of 32% percent (of patients with diabetes) in Mexico. Moreover, the number of people in Mexico with this disease could reach 17.2 million inhabitants in 2030.


Given this context, in our co-initiative fAIr LAC Jalisco, we developed an applied research project, to explore the use of Artificial Intelligence (AI) systems to influence this public health problem.


In this sense, since 2020, we have participated in the identification, design, and implementation of the use case Retina IA which aimed to streamline the processes for early detection of this disease through the development of neural networks and computer vision.


From 6 months to 8 weeks

The conventional diagnostic process tends to happen in people with Diabetes who attend periodic checkups. In general, the process of taking the image, analysis, communication of results, and medical referral can take up to 6 to 8 months in the public health system of Jalisco. Using the AI ​​tool to support the diagnosis, as well as adjustments to the screening process, it was possible to shorten the time to an average of 6 to 8 weeks, with a minimum of 4 weeks. This difference in time is crucial when talking about progressive diseases. In turn, the benefits of telemedicine and the existing health system were used, modifying the flow of information and patient care to generate a technically adequate and systematically validated solution.


Seeking to capitalize on the opportunities to take advantage of systems that use AI for the public sector, in Retina IA C Minds emphasized its responsible adoption. This included a development and implementation based on ethical principles of AI and Human Rights, understanding the possible benefits of technological use for this context, but also taking into account ethical dimensions such as data privacy, the margin of reliability of the model, and timely follow-up of patients, among other factors. For this, a work structure was designed where there was a continuous collaboration with the public and academic sectors, having the development ofa solution to the problem having the patients at the center of the approaches, thus avoiding technocentric responses.


The pilot lasted two years, from its design to the end of its implementation, where 1,053 patients were treated in three health centers. We found a health ecosystem eager for alternatives to improve patient services, and also learned greatly from the team developing a tool for doctors to help solve efficiency challenges. Additionally, processes were validated to integrate AI technologies within preventive medicine in the current health system, which opens the doors to similar pilots focused on other problems.


Furthermore, this pilot was based on a philosophy of multi-sector collaboration. In this sense, theforces of the partners of fAIr LAC (including C Minds, Government of Jalisco through the General Coordination of Government Innovation, IDB, and Tecnologico de Monterrey) joined with health practitioners. In this sense, the success of the pilot also had the collaboration of agencies such as the Jalisco Health Secretariat, Jalisco Health Services, OPD Salud, the Specific Action Program for Cardiometabolic Diseases, the Civil Hospital, the Siloé Institute, the College of Ophthalmologists of Jalisco, the Government Innovation Coordination and the managers and medical staff of the health centers of La Aurora and Esperanza, No.4 Yugoslavia and Paraísos del Colli to test the benefits of AI systems for this purpose.


The multi-stakeholder and multidisciplinary characteristics of the use case resulted in a benchmark in the region in the use of responsible AI for social good.


Image of team members from one of the participating health centers, retrieved from: http://retinaiajalisco.com

Local and regional impact

  • Generation of inputs for the creation and strengthening of AI tools and resources for social good for Latin America and the Caribbean. One of the interesting factors of Retina AI is that a comprehensive methodology was followed. It included the stages of risk analysis, prospecting, conceptualization and design, data collection and management, model development, iterations, and the final stage of use and monitoring. The learnings from each step helped to strengthen different AI tools and guides for social good of the regional initiative led by IDB, fAIr LAC. Several of the resources are already available in the library of the initiative, as well as those presented in fAIr LAC in a box.


  • Generation of public policy recommendations. Among the strengths of C Minds, is our ability to translate learning from specific use cases into public policy recommendations, especially on issues of innovation, digital transformation, and technological ethics. For this reason, we closely monitored the implementation of the pilot to develop a report of public policy recommendations, fueled by the experiences, inputs, achievements, and learning of the implementing team.

The public policy recommendations that we develop focus on three axes:

  • Articulation of the ecosystem.

    • Create a directory of key actors.

    • Develop a shared vision.

    • Promote co-creation and appropriation of the initiative.

    • Integrate a holistic collaboration model for innovation in the state.

  • Communication strategy.

    • Strengthen the internal communication of the Retina-IA collaborating team.

    • Generate communication with key actors related to the case.

    • Maintain communication with the benefited (or potentially benefited) public.

  • Implementation and scalability.

    • Create working groups with the participation of the target population.

    • Appoint counselors for each use case.

    • Develop a roadmap for the care of identified cases

    • Exchange of national, regional, and international best practices.

    • Streamline administrative processes.

    • Monitor the escalation process to maintain good AI ethics and data governance practices.

Image of the visualization of the fundus photography, retrieved from the website: http://retinaiajalisco.com/

Learnings
“The value of carrying out a pilot with the capabilities of an initiative such as fAIr LAC allows us to learn from successful and unsuccessful cases, and validate technical, systemic, and social assumptions about the application of technology. From organizations like C Minds, we have the responsibility of socializing the achievements, and learning, so that the ecosystem has a new regional point of reference”

José Roberto Mejía,

Coordinator of fAIr LAC Jalisco from C Minds.


“The implementation of the pilot provides valuable lessons and recommendations for any use case involving the public sector, both for the AI ​​ecosystem in Jalisco and for the regional ecosystem”

Constanza Gómez Mont,

Founder and Principal of C Minds.


Some of the insights we generated during this case study include:

  • The development and implementation of cases of public innovation that take advantage of AI systems must have from their conception principles of AI ethics and respect for Human Rights at the center of their operation. Beyond technical responses to public health challenges, there has to be a humanistic approach to ensure a real positive impact on communities and mitigate social risks.

  • Preventive medicine represents ample opportunities for improving the health of populations, however, it presents systemic challenges for its effective implementation and strengthening of primary care systems.

  • Despite the advantages of using AI systems to make resources more efficient, there are various social obstacles for people to access detection and treatment, and follow-up services.

  • Even a pilot who validates the cost-benefit with a positive balance will find administrative and budgetary difficulties in his establishment. The adoption in volume of service will require a higher budget and the improvement of multidisciplinary and therefore multi-institutional work capacities, necessary to systematize and scale the project.

  • AI represents an opportunity to face the limitations of economic and human resources in health institutions in the medium/long term, from a perspective of technological innovation. But the necessary budget must be faced to take the first steps.



next steps

The use case, which had a total duration of two years, ended its implementation in 2022. The next step to be evaluated by the Government is its scaling at the state level and its permanence in the three health centers where it is already part of the diagnostic processes for patients with Diabetes Mellitus. There is a great opportunity to capitalize on the lessons learned from furrowing the difficulties of generating solutions centered on ethical principles, Human Rights, and social good, within established bureaucratic systems. The learnings obtained from the pilot are fundamental so that people dedicated to the health service and the public function have comprehensive knowledge of the technology and its potential applications, in addition to adding inputs to regional knowledge networks.


From C Minds we will continue looking for opportunities to work with our partners in the development of similar use cases that result in positive impacts for local communities as well as valuable reflections for the AI ​​ecosystem for social good at a regional level.



***


The fAIr LAC Jalisco initiative is an example that regional collaboration is a driver of real impact, and that Latin America and the Caribbean is a region from which valuable experiences and knowledge are being generated.


We are thankful of the people who were crucial for carrying out the pilot and public policy recommendations including Constanza Gómez Mont, Jose Roberto Mejía, Lucia Trochez, Cristina Martinez Pinto, Luz Elena Gonzalez and Alejandra Perea from C Minds; Ricardo Swain, Enrique Cortes, Juan Roberto Hernandez, Tatiana Lefno, and Juan Alberto Amezquita, of the Tec de Monterrey; Gaspar Gonzalez Briseno, Coordinator of the diabetic retinopathy use case at fAIr LAC Jalisco; Alberto Ocampo, Head of the Department of Noncommunicable Diseases; Ricardo Garcia Gaeta, Director of Disease Control and Prevention; Edtna Jáuregui, Medical Liaison of the Heart Attack Code Program of the Cardiometabolic Program; Mayra Elizalde, Coordinator of the Cardiometabolic Program; Paul Díaz Preciado and Citlalli Becerra Fuentes, Physicians; Mario Arauz, Yunive Moreno, Adriana Diaz, Adriana Aceves, Ulises Moya, Erica Almaraz and Mayra Fernandez of the Government of Jalisco; and Cristina Pombo, Tatiana Vivriescas, Jose Alexander Soto, Luis Tejerina and Ricardo Perez-Cuevas, and Natalia Gonzalez of IDB.


We will publish soon the Learnings Report and Public Policy Recommendations for Use Cases of fAIr LAC Jalisco - Retina IA.


Comentários


bottom of page