Health

Paraná Hospitals Use Google Cloud AI to Cut Cancer Evidence Searches From a Week to an Hour

· 6 min read · Health

Paraná Hospitals Use Google Cloud AI to Cut Cancer Evidence Searches From a Week to an Hour

Paraná has become the first Brazilian state to use a Google Cloud artificial intelligence tool that helps oncologists search the medical literature, cutting the time needed to gather scientific evidence for personalised treatment decisions from a week to an hour, UOL reported on November 12. Called Project Capricórnio, the system runs at the Hospital do Câncer in Londrina and the Hospital São Vicente in Guarapuava and is starting to reach other states. It was originally developed in the Netherlands at the request of doctors at the Princess Máxima Center for paediatric oncology in Utrecht, who needed to keep up with a growing volume of studies on childhood cancer.

The tool works on PubMed, the public biomedical database maintained by the US National Library of Medicine, which holds 35 million articles and receives between 1.5 million and 1.7 million new studies a year. Priscila Cruzatti, Google Cloud’s innovation and digital health specialist for Brazil, acknowledged that doctors could access the data without Google, but said keyword-based Boolean searches are slow and laborious. Bruno Henrique Bressanini de Almeida, medical director of the Hospital do Câncer de Londrina, said doctors usually squeeze such research into gaps between outpatient appointments.

What it means in practice

Paraná Hospitals Use Google Cloud AI to Cut Cancer Evidence Searches From a Week to an Hour — contextual photo

The doctor describes the case in Portuguese, as to a colleague, including markers, condition and treatments already tried. Gemini then extracts the actionable events that could change the clinical outcome, and the system returns between 8 and 50 of the most relevant articles in a table explaining why each matters, including study type, population and journal. The final decision stays with humans, and cases still go to multidisciplinary tumour boards. In one case UOL described, a patient whose liver nodules returned was treated with hepatic ablation, which destroys the tumour with heat or cold, instead of conventional surgery.

The project is already being piloted at the Hospital Estadual da Criança e do Adolescente in Goiás, run in partnership with the Hospital de Câncer de Barretos, and is being tested in the private sector. It joins other clinical AI efforts in Brazil such as Finep’s R$280 million loan for Einstein’s AI clinical intelligence platform and the CUTIS-AI skin disease tool tested with the Health Ministry in SUS. What to watch is how quickly it spreads to more public hospitals and whether studies measure its effect on patient outcomes and costs.

← Back to AI4Brazil