Digital pathology is a major medical application using AI-powered imaging analysis. Hamburg-based Mindpeak is a global frontrunner, releasing products used in clinical practice.

Mindpeak convinces pharmaceutical companies worldwide

The year 2022 has been outstanding for Hamburg-based Mindpeak in many aspects: “We were able to release ten software products with a CE-IVD mark that are all based on our deep-learning platform,” says Co-founder and CEO Felix Faber. The second highlight of the year was that Mindpeak succeeded in entering the biopharma market, winning one of the largest oncology drug developers worldwide as a customer. “Both milestones result in a 1600% increase in revenue,” Faber says.

Further good news is that Mindpeak’s AI solutions are among the global frontrunners that are already in use in clinical routine diagnostics. One of the most promising fields of AI in clinical applications is digital pathology. AI-powered imaging analysis systems evaluate microscopic images of suspicious tissue biopsies, and trained software has learnt to distinguish tumor tissue from healthy tissue.

Mindpeak’s deep learning algorithms use multi-layer neural networks to analyze data, allowing pathologists to explore and extract information beyond human visual perception. “Our tools quantify cells or biomarkers and assist cancer experts in providing reliable and reproducible diagnoses. And our clients include pharma companies that want to discover and develop new biomarkers,” explains Faber, who is a computer scientist by training.

How does Mindpeak support cancer diagnostics?

Mindpeak, founded in 2018, applies a hybrid deep learning approach, where semi-supervised methods are combined with supervised methods. “A lot of work went into the first software and the platform behind it,” Faber says. Developing the platform itself took about 2.5 years. The first product recognizes and classifies breast cancer cells in tissue samples taken in a fraction of a second. The software received its CE-IVD mark in May 2021. “This made us the first company in Germany with such an approval in clinical routine diagnostics in pathology,” says Faber.

Working hard to develop a platform is now paying off for the product portfolio and the development pipeline. “We are now able to release algorithms much faster than before,” Faber says. He reckons that one of Mindpeak’s greatest assets is the algorithm analyzing the biomarker PD-L1, which is of major importance in the dynamic field of immuno-oncology. Faber says the IVDR and the limited number of Notified Bodies are major barriers now. “That is why we are likely to put only two products on the market in 2023,” he adds.

How does the new AI solution gain the trust of users?

The success of AI-based approaches relies on the quality and quantity of the data used to train the algorithm. Having access to tissue slides from partner laboratories is an important asset for Mindpeak in this respect. And there are also a limited number of customers which allow the start-up to train with their data. A network of over 30 pathologists annotate data based on rigid guidelines and create ground truth data to train the AI.

All Mindpeak products “explain” how they made a decision, allowing pathologists to understand and trust it. “We always show what the neural network detected and how it came up with the score.” Faber points out that explainability is absolutely necessary to get acceptance for AI-based solutions on the market.

A head start with zero-click solution

Another special feature is what the Mindpeak team calls a zero-click solution. The system pre-processes all the cells on the whole slide with two AIs and the pathologists are presented with all the results when they open the case. “It gives you a good overview and saves a couple of seconds, which doesn’t seem like much, but if you know how pathologists work, this is extremely important to the workflow.” For example, Mindpeak’s customer Unilabs, one of the largest European diagnostics providers, introduced its AI solution to the clinical routine to speed up breast cancer diagnosis and make it more accurate. It recently reported a 90% reduction in diagnosis time with the Mindpeak solution. “They say they are ten times faster and will now make it standard in Sweden.”

The start-up is funded by the Innovationsstarter Funds Hamburg (IFH), with support from

• APEX Ventures,
• Nina Capital,
• Motu Ventures,
• and prominent angel investors.

Mindpeak, with a team of 25 employees, has raised 9 million euros so far. “We want to grow our team in 2023,” says Faber. Since more than two thirds of revenue come from the business in the United States, Mindpeak is going to open a Sales Office there soon.

When it comes to recruiting, Felix Faber says it’s helpful that data scientists and machine learning experts show interest in the healthcare market. “With our purpose-driven approach, we are able to attract talent.”

Text: Philipp Graf

Featured image: © Mindpeak