The Swedish Childhood Cancer Fund invests in multimodal AI to defeat childhood cancer

In Sweden, 85 percent of children diagnosed with cancer survive. Sweden is at the forefront of cancer care internationally, but still 15 percent of children die. Among the survivors, many are forced to live with severe complications caused by the disease or its treatment. Although the survival curve climbed steeply for decades, progress has leveled off since the early 2000s. New ideas are needed to fulfill the Swedish Childhood Cancer Fund’s vision in which all children diagnosed with cancer should survive and live long, healthy lives.

"We must find new ways to treat children with cancer, and for that, we need much more data we can collectively learn from", says David Gisselsson Nord, Professor at Lund University and Senior Advisor at the Swedish Childhood Cancer Fund.

New answers through advanced AI

BrainChild is one of the Swedish Childhood Cancer Fund’s most significant undertakings, aiming to help more children survive. The project generates AI and machine learning opportunities to collect, structure, and analyze health data to improve diagnoses and treatment. Researchers, the Swedish Childhood Cancer Fund, Microsoft, and the Swedish government are among the collaborators. Microsoft contributes secure IT solutions and tools that help researchers quickly detect patterns linking genes, treatments, and outcomes.

"It’s one of the most ambitious projects we’ve ever undertaken, and it’s possible because so many are pulling in the same direction", says Ola Mattsson. Anna Eklöf, BrainChild Project Manager, adds, "We believe many of the answers already exist. The challenge is to collect and structure the data, so connections become visible".

Ethics and responsibility

Handling data naturally raises concerns about personal privacy. Ola Mattsson stresses that while there is a justified focus on integrity, "We also believe it’s unethical not to use data that could save lives. Patient safety remains priority throughout the project”.

David Gisselsson Nord has a key role in BrainChild. He is the clinical lead for one of the project’s first pilot studies, combining clinical information with large-scale data from whole-genome sequencing and digital imaging. The material comes from over 500 Swedish children who were diagnosed with solid tumors and treated between 2021 and 2024.

Although Sweden’s survival rate is high, it drops to around 50 percent for children who relapse. "There is a pressing need to replace intensive chemotherapy treatments with more targeted drugs that typically have fewer side effects and long-term complications", explains David Gisselsson Nord. "To offer new, gentler treatments, we must understand why some children respond to them and others don't. We can only achieve that by learning from real patient data".

AI as a partner

Researchers are working closely with Microsoft to structure and analyze the collected data. AI models are trained to examine images and data to detect patterns that even the most skilled pathologists might miss. The field is evolving quickly, with new multimodal models for healthcare emerging. These models allow researchers and clinicians to combine insights from medical records, genomics, and imaging, while advanced language models provide more natural support.

"AI can find patterns we cannot see with the naked eye. It’s not about replacing doctors but about giving us better tools to make the right decisions", says David Gisselsson Nord.

BrainChild is transforming our understanding of childhood cancer

David Gisselsson Nord believes the technology could fundamentally change how we understand various childhood cancer diagnoses.

"We’ve visualized genetic information and survival data in the same tool, and it completely changes how one sees the disease. Suddenly, some children with the same genetic change survive, and others do not", he says. "It raises new questions and prompts researchers to ask what other factors might influence outcomes".

The pilot study demonstrates the project's potential and shows that it works. Today, the project is scaling up, involving more data sources, broader usage, and additional diagnostic groups.

"We have a clear plan for the next few years. We’re building further technically, but we’re also working on advocacy, regulatory frameworks, and partnerships. Everything needs to happen in parallel", says Anna Eklöf.

From Sweden to the rest of the world

The project also holds global potential. David Gisselsson Nord sees great value for lower-resource countries that have a shortage of pathologists and low childhood cancer survival rates.

"AI could become a way to bring advanced diagnostics to places lacking experts today. It’s about democratizing and revolutionizing childhood cancer care globally", he says.

The technology and model developed through BrainChild can be scaled and adapted. While the work starts with childhood cancer, David Gisselsson Nord believes the approach could eventually be applied to other types of cancer and even to entirely different diseases.