Artificial intelligence (AI) has been one of the main innovations in healthcare in the last few years, augmenting the sector with processes automation, errors reduction, and quality improvement. AI also helps healthcare professionals with data analytics, lesions measurements, and reduction of repetitive tasks, providing them more time for patient care.

As in many other parts of the world, AI is being more and more in the daily routine of hospitals, clinics, and diagnostic centers in Brazil. The main use cases are in radiology, for detection, characterization, and management of conditions such as pulmonary embolism, lung nodules, and brain hemorrhage, speeding up magnetic resonance protocols and aiding physicians with the clinical evaluation of chronic diseases like hypertension or diabetes.

The COVID-19 pandemic prompted many of the healthcare centers and hospitals in Brazil to look for strategies to support the response to the viral outbreak. Besides AI algorithms for detection and quantification of the pulmonary ground-glass opacities caused by the coronavirus, many Brazilian startups were positioned to create AI solutions tailored for local needs, in Portuguese and much cheaper than those offered by big international companies. A consortium of different players from the academia, healthcare organizations, civil society, institutes, and foundations gathered and created a big data lake of chest radiography (CR) and computed tomography (CT) images of COVID-19 patients, powered by AI algorithms dedicated to the assessment, detection and quantification of the pulmonary findings related to the viral infection. The Institute of Radiology of the University of São Paulo spearheaded the initiative, named RadVid-19, which was sponsored by the Brazilian College of Radiology and had the participation of almost 50 hospitals, 40% of which from the public healthcare system, and diagnostic centers in Brazil, and gathered more than 25,000 CR and CT studies from different Brazilian regions.

Every case submitted RadVid-19 was automatically evaluated by AI algorithms, providing a short assessment of the likelihood of COVID-19 and its severity, based on the quantification of the chest CT findings (figure 1). The first AI algorithms used at RadVid-19 had been developed by international companies (Siemens and Huawei), which were very powerful but presented some drawbacks. Cost was one of the main concerns, and customization of those AI solutions to the local market was also an issue. For those reasons, there was a decision to make a partnership with startups to develop new algorithms, tailored to the local needs, engaging Brazilian innovators, programmers, physicians, and data scientists, helping to nurture a national AI ecosystem.

Figure 1. An example of AI detection and quantification of COVID-19 pulmonary findings on a chest CT.

RadVid-19 local team analyzed the main COVID-19 guidelines in Brazil and worldwide, interviewed physicians and other healthcare professionals, and explored the main bottlenecks for COVID-19 diagnosis in Brazil, a country where testing was not widely available at the beginning of the pandemic. The project team designed a workflow suitable for less experienced and expert radiologists, summing clinical data and PCR results to the radiologists’ own visual analysis. The system was rapidly incorporated by Brazilian radiologists and proved to be a reliable tool to be used as a first reader for less experienced physicians and as a second reader for experts in chest imaging.

In conclusion, RadVid-19 was one of the most impressive initiatives in Brazil during the pandemic, helping patients and healthcare professionals all over the country with AI evaluation and quantification of COVID-19 findings on chest X-ray and CT images. In a very short period, a big and strong collaboration among different players from diverse segments fostered the Brazilian startup ecosystem, developed new professionals and innovators, diagnosed more than 25,000 patients with COVID-19 and supported the national healthcare system to fight against the pandemic.

**** 

RadVid-19 was a collaborative effort among many different companies, academia, institutes, foundations, and people. A special thanks goes to Dr. Claudia Leite, Dr. Marcelo Félix, Marco Antônio Bego, Dr. Bruno Aragão, Dr. Shri Jayanthi, Dr. Márcio Sawamura, Dr. Cesar Nomura, and Dr. Giovanni Cerri.

Publicado na coluna de outubro de 2022 na The Yuan.