Image: Ewa Krawczyk / National Cancer Institute

Published on 06/10/2024

By Roseli Andrion  |  Agência FAPESP – Artificial intelligence (AI) has been part of everyday life for some time, and a growing number of business organizations are using it to develop solutions of all kinds, The startup Huna, for example, uses AI to detect chronic diseases such as breast cancer by analyzing ordinary blood work.

In 2018 its researchers helped set up an AI laboratory at the Federal University of Minas Gerais (UFMG), the first Brazilian university to have one. “We realized the data in blood work was underutilized,” said Daniella Castro, co-founder of Huna and the researcher responsible for the project, which is supported by FAPESP’s Innovative Research in Small Business Program (PIPE).

Biological processes in the human organism are not linear, especially in the case of complex diseases, Castro explained, so certain subtle details of a blood test may go unnoticed, not least because markers interact. “AI can be used to identify patterns invisible to the human eye. For example, cholesterol interacts with glucose, leukocytes and many other components of blood,” she said.

This characteristic of blood tests – the fact that they contain hidden information – lies behind the healthtech’s name. In Maori, the native language of New Zealand, kura huna means hidden knowledge and wisdom protected as treasure.

“That’s exactly what we aim to produce by analyzing blood work in search of unseen patterns with the help of AI,” said Marco Kohara, co-founder of the startup.

Accordingly, the scientists have used data from blood tests to point to the possible development of Alzheimer’s disease four years in advance of the appearance of any symptom and to diagnose COVID-19 during the pandemic before vaccines became available, for example.

“When a friend was diagnosed with metastatic breast cancer, I thought of using the method to screen for this disease, which is more likely to be cured when it’s diagnosed at an early stage,” Castro said, noting that 80% of Brazilian municipalities do not have mammography machines. “Screening mammography is essential for early diagnosis but isn’t available to some 20% of Brazilian women. The possibility of detecting the disease at an initial stage is a major benefit of our method,” Kohara said.

The process consists basically of identifying patterns in blood work that are characteristic of specific diseases. “We built the world’s largest database of hemograms labeled for breast cancer – at least the largest scientifically published one. On that foundation, we created a risk stratifier for use in screening programs,” Castro said.

Ancillary tool

The solution is designed to identify patients with up to three times more risk than is typical for the disease. “So if only 20% of the population have access to screening, we set out to find a way of making it more effective. A complete blood count test is inexpensive, and the goal is to prioritize women at greater risk of breast cancer on the basis of the patterns identified,” Castro said.

The method is not intended to replace diagnostic tests already in use, such as mammograms and breast ultrasound. “Our goal is to prioritize high-risk women and help reduce waiting times for mammography,” she said.

In future, the method can be adapted for the detection of other diseases. “We want to create models for other types of cancer, especially those with high incidence rates and that cost the health system a great deal,” Kohara said.

Given the routine nature of the test and its low cost, the startup wants its use to be as widespread as possible. “Hopefully, it will be available from the SUS [Sistema Único de Saúde, Brazil’s public healthcare network]. Besides being cheap, hemograms are prescribed for many reasons, including detection of anemia and infections, for example. So it has multiple uses and can be a value-added medical tool,” Castro said. 

It should also be a tool for policymakers and administrators, he continued, explaining that they can use it to decide which groups at risk of disease should be prioritized. “Blood counts don’t have significant value at the level of the individual, but they’re essential to define ways of prioritizing high-risk patients. Physicians will be able to use it to decide whether to prioritize particular patients, refer them sooner to specialists, or prescribe supplementary tests and exams,” Kohara said.

The startup wants to export the technology in future. “Cancer is a world problem. Theoretically, our market is global,” he said. “In addition, the population is aging and broader use of a technology like this will be particularly helpful to people without sufficient access to healthcare.”

Early diagnosis

Cancer was chosen because of the possibility of optimizing early diagnosis with technology. “AI is useful precisely to predict the occurrence of alterations, which is very important in the case of cancer and can represent better outcomes for patients by permitting less aggressive, complex and expensive treatments in most cases,” Kohara said.

Besides early detection of disease, the scientists are studying the creation of models to predict risk for specific patients and prognosis in the context of a specific treatment. “We can drill down into the patient’s experience of the disease. For cancer patients, for example, it would be possible to determine who is likely to have a relapse and optimize their treatment accordingly,” he said.

The technology is currently being tested in partnership with the Barretos cancer hospital (Hospital de Amor), which is part of the SUS, the Fleury group (a private laboratory chain), and health maintenance organizations. The tests are geared to proof of concept. The startup receives private investment as well as support from FAPESP.

 

Source: https://agencia.fapesp.br/51907