AI in healthcare is accelerating medical breakthroughs
Artificial intelligence (AI) is rapidly transforming the way we view medicine and biomedical research. Whereas researchers once spent months analysing data, algorithms can now detect patterns that are barely visible to the human eye. Yet this revolution also brings new challenges for researchers as well as doctors. Kris Laukens (Faculty of Science) and Tim Van De Looverbosch (Faculty of Pharmaceutical, Biomedical and Veterinary Sciences) share their enthusiasm and concerns about AI with Stroom.
In brief
- AI helps to analyse vast amounts of information and identify new connections.
- Experiments are becoming faster and cheaper.
- Results still require validation and checks for bias.
- Proper training and a critical eye remain essential.
- AI makes personalised medicine possible.
AI connects bioinformatics and image analysis
At first glance, image analysis and bioinformatics may seem like very different fields. Yet they share the same goal: to better understand biological processes. AI brings these fields together. Both image analysis and bioinformatics start from a common challenge: making sense of complex biological systems.
‘Ultimately, it all revolves around biological data,’ says bioinformatician and date mining expert Kris Laukens. ‘AI helps us analyse vast amounts of information and uncover connections we couldn’t see before.’ Cell biology researcher and expert in AI-driven analysis Tim Van De Looverbosch adds: ‘Our worlds are converging. We can now combine image data with sequencing data, which gives a far more complete picture of what happens inside cells.’
What are sequencing data?
These are the data generated when reading genetic material, such as DNA or RNA. They form the foundation for understanding biological processes and they are crucial for personalized medicine.
AI makes it possible to handle a data tsunami in ways that were previously unthinkable and helps us analyse vast amounts of information and uncover connections we couldn’t see before.
Deep learning: a breakthrough in biomedical research
AI isn’t just the latest hype, both researchers emphasise. ‘I started working with techniques that can be seen as precursors to AI around 2005,’ says Laukens. ‘At that time, it became clear that the volume of biological data was growing exponentially. Traditional computational methods could no longer keep up.’
For Van De Looverbosch, the turning point came around 2017: ‘That’s when deep learning became a real game changer. Suddenly, you could feed raw data into neural networks [computer models inspired by the human brain] and the system would learn for itself which patterns were relevant. That was revolutionary.’ A well-known example is AlphaFold, an AI model from DeepMind capable of predicting the 3D structure of proteins. ‘What had been considered unsolvable for decades suddenly became feasible thanks to AI,’ says Laukens. ‘That sent shockwaves throughout the life sciences.’
AI as a smart assistant for researchers
AI makes biomedical research more efficient on multiple levels:
- Faster and more accurate image analysis: for example through models such as CellPose, trained on enormous datasets from scientists worldwide
- More efficient data collection: AI automatically detects rare events in cells and only stores those images in high resolution, saving researchers both time and storage space.
- Lower costs: during DNA sequencing, AI can decide whether it is worthwhile to continue analyzing a sequence or to stop early.
Even so, the road from algorithm to clinic remains long. ‘There are often years between an AI discovery and its application in patients,’ says Laukens. ‘Still, the role of AI in healthcare is growing rapidly. AI can help doctors recognise warning signs more quickly and make recommendations. The technology won’t replace physicians, but it can support them in making better decisions.”
Van De Looverbosch also emphasizes that AI makes research more accessible for smaller labs. “In the past, certain experiments were expensive, but today open AI tools allow researchers to perform analyses that were once only feasible at top institutions.”
New skills needed in the AI era
The impact on day-to-day work of a researcher is enormous. AI supports researchers in analyzing complex datasets, generating hypotheses, optimizing experiments and accelerating analyses.
‘AI has become a valuable sparring partner, but it still requires an active, critical approach. By asking precise questions from your own expertise, you sometimes open up unexpected directions.’ ‘says Van De Looverbosch. Laukens observes a similar shift on a larger scale: ‘At our spin-off ImmuneWatch, we use AI to analyse whether candidate vaccines trigger a T-cell response [a specific reaction of the immune system to intruders such as viruses or cancer cells]. Before AI it required weeks of laboratory research and was therefore time-consuming and expensive.’
The rise of AI is also reshaping the researcher’s profile and this shift needs to be embedded in education. According to Laukens, the essential skills researchers now need include:
- grasping how models work,
- interdisciplinary communication and collaboration,
- critically evaluating AI results,
- algorithmic thinking and programming.
Laukens stresses the importance of interdisciplinary communication: ‘A computer scientist must understand what a biomedical researcher means, and vice versa. That requires a mindset of collaboration and lifelong learning. What’s relevant today may be obsolete in ten years.’
The risks of AI: bias and validation
Yet the pace at which AI is advancing also carries risks. ‘It’s still crucial to validate results,’ stresses Van De Looverbosch. ‘We still compare our AI outputs with manual annotations to measure how well the model truly performs.’
Laukens adds: ‘Bias is a major concern. If a model is trained mainly on Western populations or certain data types, it risks drawing false conclusions. At the Antwerp Center on Responsible AI (ACRAI), we not only test model accuracy but also assess whether models are free of bias, do not reinforce stereotypes and certainly do not discriminate.’
AI is sometimes a better sparring partner than a colleague: the model always has time, no agenda, and sometimes thinks in patterns you’d never come up with yourself.
Collaboration as the key to reliable medical AI
The key to reliable AI lies in collaboration. The BIOMINA network brings together experts in computer science and the life sciences This cross-pollination is essential,” says Laukens. “Our annual symposium AntwerpHealth.AI, for example, brings together computer scientists, physicians, and biologists to share insights. That creates real value for everyone involved.”
Van De Looverbosch agrees: ‘Sometimes, someone from a completely different discipline suddenly offers an unexpected insight during a lunch meeting that advances your research. In addition, BIOMINA organizes the inter-university workshop From Pictures to Numbers, where researchers learn how to use AI for image analysis.’ analysis.’
What will the biomedical world look like in ten years’ time?
Laukens is optimistic: ‘I foresee personalised therapies and immunotherapies that are far more tailored to individual patients. AI will support doctors by translating complex data into clear insights, giving them more time to talk to their patients.’
Van De Looverbosch looks to the fundamentals of biology itself: ‘We will understand cells better than ever. By combining imaging and sequencing, we’ll gain insights into how abnormalities in DNA or RNA translate into cellular problems. That opens the door to more targeted treatments.’
AI promises to revolutionise healthcare, but only if used responsibly. ‘The potential is enormous, but we must remain critical,’ concludes Laukens. ‘AI isn’t a cure-all, it’s a tool. It demands collaboration, transparency and constant reflection. Our human curiosity and sense of responsibility remain essential. AI can help us understand more — not stop us from asking questions.’
Should universities fully embrace AI?
The evolution of AI is moving incredibly fast—too fast, according to philosopher Anthony Longo. He would prefer to slow things down. David Martens, founder of ACRAI, on the other hand, believes that universities should fully embrace AI.