For the second year in a row, two Unicamp research groups were awarded by the Google Research Awards for Latin America. The research, developed in the field of Artificial Intelligence, at the Computing Institute (IC) and at the Faculty of Electrical and Computer Engineering (FEEC), seeks automatic screening models to detect Diabetic Retinopathy and Melanoma. "They are diagnostic imaging diseases, which is our expertise, and whose early identification is very important”, he explained Eduardo Valle responsible for the Melanoma project.
According to the professor, the objective is to develop instruments that help primary health professionals, general practitioners and nurses, to screen and refer patients, thus optimizing specialist time. Using images, researchers train the computer to identify each disease. “We show several cases and the program creates a model of what the disease is. From this model, it begins to make predictions and presents the probability of that image indicating the disease or not. With this information, the healthcare professional can start making decisions on how to refer that patient”, explained Valle.
The two projects follow similar paths, but are at different stages. According to Valle, the project “Data-Driven Automatic Triage of Diabetic Retinopathy”, coordinated by professor Anderson Rocha Instituto de Computação (IC), is at a more advanced stage of research and was the driver of the “Reliable automated identification of melanoma for the real world”. According to him, the success rate of the model that identifies airetinopathy is around 98%, while melanoma is 87,1%, in a computer laboratory. “In the area of retinopathy, we already have a research tradition. In melanoma, we are still an incipient group, but we had a great result at the beginning of the year, taking first place in an international competition for automatic melanoma detection”, he said. (Read matter)
The retinopathy project awaits an opinion from the Ethics Committee and is in partnership with the ophthalmology team at Hospital da Clínicas (HC) at Unicamp, to leave the research laboratory and be confronted with real clinical situations, in a simulation that will verify whether the System responses match those of human experts. The big challenge now, according to Valle, is to gain doctors' trust in using the technology. “This work is also part of our research. It's not just the development of artificial intelligence, but also being able to place the technology within the hospital. In addition to improving systems to build trust, we need to design convenient interfaces that speak the doctors’ language,” he pointed out.
The Unicamp students awarded by Google this year were Michel Fornaciali (doctorate, Feec) and Ramon Pires (doctorate, IC). Professors Jacques Wainer, Siome Goldenstein and Sandra Avila (IC/Unicamp), Flávia Vasques Bittencourt (FCM/UFMG), Herbert Jelinek (Charles Sturt/Australia), Lin Tzy Li (Samsung), and master's students also participated in the research. Afonso Menegola and Júlia Tavares (Feec/Unicamp). Check out the full list of winners at official blog.
Google Research Awards for Latin America
The Google program for Latin America offers one-year research scholarships for professors and master's and doctoral students in specific areas of Computer Science. In this year's edition, US$600,000 was distributed among the 27 selected projects. Each master's student can compete for two consecutive years and each doctoral student for three years.
Read also
Program makes computer 'understand' digital images in retinal examination
Unicamp team comes out on top in international automatic melanoma detection competition
Aggressiveness of good
Unicamp researchers are awarded Google grants
