Brazilian startups use artificial intelligence to diagnose Covid-19

Many Brazilian startups are betting on Artificial Intelligence to cope with covid-19.
In July, an edict launched by the government of the state of São Paulo, through the IdeiaGov program, opened a public call to select technological solutions that used artificial intelligence algorithms, capable of helping physicians in the diagnosis of coronavirus from x-ray scans and computed tomography.
The challenge, which included 21 enrolled, selected three startups, which will have their proposals validated at the Hospital das Clínicas of the School of Medicine of the University of São Paulo (USP). After this period, which should take up to eight weeks, the solutions can be adopted by the state.
According to the government of the state of São Paulo, today, the only solution of this kind in the world commercially available is in China.
In practice, artificial intelligence allows machines to learn to reason in a similar way to humans. It was from this possibility that the startups involved in IdeiaGov worked.
They developed convolutional neural networks and trained them with images of lungs with Covid-19 and without the disease, so that the elaborate system could learn to diagnose the patient in less than a minute. Proponents could present solutions for x-ray reading and tomography or only one of the types of examination.
NeuralMind, which participated in the selection through a consortium with the Laboratory of Medical Imaging Computing of the Faculty of Electrical and Computer Engineering (MICLab FEEC), of the State University of Campinas (Unicamp), ranked first in the evaluation for chest x-ray application and tomography. In both cases, convolutional neural networks, Deep Learning algorithms that capture input images and assign them seized weights and vieses are used. The researchers were also able to convert tomographic images into x-ray-like images in order to facilitate the use of the same neural network.
The solution can be used in the diagnosis and also in the prognosis of patients already diagnosed with the disease.
NeuralMind operates within the Scientific and Technological Park of Unicamp, involving eight people. The team was already studying the applicability of artificial intelligence in x-ray diagnostics and tomography since the emergence of coronavirus.
The startup Otawa Health has been developing technology involving AI in the oncology area since 2016 and began this year to research ways to apply the tool in the diagnosis of Covid-19. Five professionals are involved in the product. The solution proposed by Otawa Health provides for the classification of x-ray images to be processed in three levels of probability of Covid-19 – low, medium and high.
An application was developed in Portuguese, Spanish and English to help professionals. Doctors can send the x-ray image of their patients to the server either by cell phone, by photo, or by connection to hospital equipment.
The startup Visibilia, from São Carlos, was approved with a solution focused on computed tomography, FADCIL. The system also automatically identifies cases of coronavirus on medical images, through Deep Learning and according to the anatomical variability of patients.
Composed of six associated researchers, the company has been in the market since 2017 with solutions applied to the business world. Fadcil's project is the leader of the researcher Doctor Gabriel Humpire, a professional with experience in analysis and processing of medical images, in particular, computed tomography for automatic detection of diseases such as cancer.
The Visibilia product integrates with the PACS tool, already used in clinics and hospitals for archiving images, and which can be connected directly to ct scans and x-ray machines. There is also the possibility of developing a web interface that allows the doctor to directly upload the images to the FADCIL server in the cloud. The FADCIL server would generate this diagnosis and deliver it to the doctor.