For this new edition of the Scientifica Journal, our guest is Noemi Sgambelluri, a Physics graduate from the Alma Mater Studiorum – University of Bologna.
Noemi is also one of the winners of Thesis, the program promoted by Scientifica Venture Capital to support STEM culture.
Through her experience, Noemi will walk us through how Artificial Intelligence is transforming the field of brain imaging.

Hi Noemi, welcome and congratulations again on your Thesis win! Can you tell us how your research journey began and what led you to focus on Artificial Intelligence applied to MRI?

Of course. After graduating in Physics and completing my Master’s in Applied Physics, I decided to pursue a career exploring the applications of Artificial Intelligence in Magnetic Resonance Imaging.
I was driven by the desire to have a tangible impact on people’s lives by improving healthcare systems.
During my research period at the University of Oxford, I had the opportunity to witness firsthand the potential of AI in this field, especially in neuroimaging. Working on these topics showed me just how much AI is revolutionizing the way diagnoses are made—boosting precision and reducing image acquisition times.

You’ve focused on a specific area: diffusion-weighted brain MRI. Could you explain what that is and how Artificial Intelligence plays a role?

Diffusion-weighted MRI, or dMRI, is a technique introduced in 1985 that studies the movement of water molecules along different directions within the body, particularly useful for the brain.
We can imagine the brain as a cluster of cities (neurons) connected by “highways” through which water molecules move. dMRI traces this motion, almost like a GPS, and is highly sensitive to lesions and conditions such as Alzheimer’s or multiple sclerosis. However, its complexity often leads to long scan times, limiting its clinical use.
AI can truly make a difference here. Models like Generative Adversarial Networks (GANs) can quickly learn the features of diffusion signals and predict missing data, reducing the number of measurements needed. This not only speeds up acquisition times but also enhances image quality, thanks to deep learning algorithms that reconstruct images from incomplete datasets and remove noise and artifacts.

Very interesting. So, Artificial Intelligence not only accelerates the process but also improves image quality?

Exactly. In addition to shortening acquisition times, AI enables high-quality imaging even in resource-limited settings, making the technology more accessible.
One interesting example is the DeepDTI study, which uses a convolutional neural network to reconstruct diffusion images from undersampled data, cutting acquisition time by up to 50% without compromising image quality.
In healthcare, AI holds tremendous potential to improve clinical diagnostics while also paving the way for broader access to cutting-edge tools that, until recently, were available to only a few.

Can you explain this idea of democratizing technology a bit more?

In a world where state-of-the-art MRI scanners can cost millions of euros, AI presents a way to improve performance even in low-resource settings.
In smaller hospitals with only one scanner and limited operating hours, AI can speed up scans, enhance image quality, and reduce overall costs, ultimately improving the experience for both doctors and patients.

The implementation of low-cost MRI solutions is still a subject of debate. What are some of the approaches currently being explored?

Over the past 30 years, scientists have identified low-field MRI – which uses simpler magnets – as a potential solution. These scanners lower installation and maintenance costs and require less infrastructure.
Recently, a Science article by Professor Ed Wu from Hong Kong University showed that full-body scanning can be achieved using a 0.05T system, a major step toward commercializing more affordable and potentially portable MRI systems.
The main challenge remains image sensitivity, but new deep learning algorithms are already proving effective on these systems.
Improving image quality, reducing costs, and increasing practical benefits would accelerate the adoption of this technology, making MRI accessible to the remaining 50% of the global population.

What advice would you give to someone starting a STEM academic journey?

My advice is to look beyond your comfort zone. Great innovations often happen at the intersection of different disciplines.
Working in multidisciplinary teams can truly make a difference. If I could go back in time, I’d approach university life more proactively, seizing every opportunity and viewing failure not as an obstacle, but as a moment for growth.
After all, science is made up of trials and the ability to learn from mistakes is crucial for innovation.