DIGITAL HEALTH/DIAGNOSTICS Stratipath Breast is an AI-based image analysis and decision support platform that aims to reduce the incidence of breast cancer relapse. Stratipath received early-stage support from Karolinska Innovations and the product is now going from being a pilot project to being rolled out in hospitals nationwide.

When Johan Hartman and Mattias Rantalainen realised the potential of AI-based image analysis, Stratipath was born. They recently launched Stratipath Breast, which is set to lead the way for precision medicine and diagnostics in cancer treatment.
“We’d been working together on various projects, including bioinformatics and gene expression, since 2014. But when we saw the improvements in image analysis, we both knew it was an area for the future,” says Hartman, professor in tumour pathology at Karolinska Institutet and co-founder of Stratipath.

Identifying those at higher risk

Stratipath Breast is an AI-based tool that should make it easier to find breast cancer patients who have a high risk of relapse. Today, breast cancer cases are divided into three categories: high, low, and intermediate risk. But because it is so difficult to determine the risk of relapse, almost half of patients are classified as intermediate risk.

In practice, this means that patients who should receive treatment miss out, while other patients are given unnecessary treatment, often with severe side effects.
“We see several problems in current cancer treatment that we can address with the products we’re developing. Sometimes, diagnoses are completely wrong and vary depending on where in the country you live,” says Hartman.

Stratipath’s solution has been made possible thanks to the digitalisation of the country’s clinical pathology labs and the rapid increase in computing power in recent years. Digitalisation provides easier access to high-resolution imagery and the improvement in computing power has paved the way for the development of new AI models.
Today, tissue samples taken in the course of cancer treatment are examined digitally by pathologists. But making the right diagnosis requires extensive knowledge and experience. In addition, most countries face a shortage of pathologists.

Hartman and Rantalainen quickly saw how well suited AI was to image-based analysis.
“Our vision is that our image-based analyses provide attending doctors and oncologists with the help they need to diagnose each individual patient. At the same time, we want to be very clear that our tools should complement the role of pathologists.”

Stratipath’s AI model has been trained on existing data from previous patients collected from several hospitals in the country. This information contains scanned images and patient outcomes. Using this data, the AI model has learned to recognise which patients have a high risk of relapse and can flag these up for the pathologist.

“By connecting thousands of images with patient outcomes, our model is able to draw reliable conclusions about each patient. We cannot say exactly which patterns the computer identifies, but we know that results are replicated time after time.”

Hartman and Rantalainen contacted Karolinska Innovations early in the development of Stratipath Breast, which in turn put them in touch with a business coach. That business coach, Fredrik Wetterhall, is now also CEO of the company.
“As a team, we complement each other perfectly. I’m a clinical expert, Mattias has deep technical knowledge and Fredrik has the knowledge to build a company.”

The move to production proved unexpectedly tough

Stratipath took in its first venture capital in 2021 to accelerate development of Stratipath Breast and the company itself. Today, they have almost 20 employees and the product is CE-marked according to the In-Vitro Diagnostic Medical Devices Directive, the so-called CE-IVD. However, the step from research to developing a product that could be approved was considerably greater than the pair had anticipated.

“Compared to creating a medical device, research and publishing are pretty straightforward. Work to obtain the required approvals has been extremely extensive,” says Hartman.

Now the goal is to give all the country’s breast cancer patients access to the decision making support platform.
“Our first product is aimed at breast cancer, but we’re working further to develop systems for more of the major cancers.”

An alternative to AI-based image analysis is to analyse which genes are active and inactive in cancer tumours, so-called gene expression analysis. However, this is a costly method where results can take up to two weeks to come through. Stratipath’s analysis currently takes just under one hour and the answer is delivered directly to the attending doctor.

“Our solution is much more cost-effective than the methods based on gene expression analysis. In addition, our product can be integrated with existing technologies in the lab.”

Hartman is convinced that the use of AI analysis will become an integral part of cancer healthcare going forward.
“This will completely change the way we look at cancer diagnosis as early as the coming five to seven years. The technology represents a major step towards more equal cancer care with precision diagnostics for patients.”