Dortmund, 13th May 2026
Dr Davide Panzeri is a biophysicist in the AMBIOM – Analysis of Microscopic BIOMedical Images research group, specialising in AI-based analysis. With his new project PRISM – Physics-aware Generative AI for Double-Blind Spectral Unmixing he aims to significantly improve fluorescence microscopy. His goal is to separate fluorescence signals more quickly and accurately, even when these strongly overlap. For this work, the 30-year-old has been awarded a two-year Marie-Skłodowska-Curie Fellowship (see info box) from the European Union, starting from June 1, 2026, and worth 218,000 euros.
Fluorescence microscopy is a powerful technique for visualising and analysing cellular structures within a sample. Fluorophores (fluorescent labels, such as dyes or fluorescent proteins) are used to tag target molecules (like proteins, DNA, enzymes, viruses). When excited by specific wavelengths of light, the labelled cell components emit light in other colours, such as red, green or blue. These colours reveal important information, for example about cellular organisation, interactions, and dynamics. However, despite its strengths, analysing fluorescence microscopy images is often time-consuming and challenging.
With PRISM, Panzeri wants to establish a double-blind approach. Unlike previous methods, it does not require prior knowledge of the spectral profiles or the number of fluorophores involved.
© ISAS
The problem: overlapping colour spectra & autofluorescence
One major challenge is spectral overlap. “The emission spectra of different fluorophores can overlap, making it difficult to clearly assign fluorescence signals to specific structures,” explains Panzeri. Signal overlaps complicate both the qualitative interpretation and quantitative analysis of the data. To address this, the ISAS biophysicist conducts research on AI-based methods for signal separation.
Currently, computer-assisted spectral unmixing uses predefined reference spectra to estimate each fluorophore’s contribution to the overall signal. However, this well-established method requires prior knowledge of both the number of fluorophores and their spectral profiles. “Spectral unmixing needs careful pre-calibration,” says Panzeri. “Autofluorescence remains another challenge: certain biological structures naturally emit light without any fluorescent dyes, further complicating the signal.”
Blind separation without knowledge of singular spectral profiles
Unlike previous methods, with PRISM, Panzeri wants to establish a double-blind approach that eliminates the need for prior knowledge of the number of fluorophores and their spectral profiles. A blind algorithm is intended to determine these spectral profiles from the measured microscope images without any prior knowledge of the underlying data. Panzeri will be developing an AI-powered method that bypasses time-consuming and costly pre-calibration, enabling biomedical researchers to measure fluorophore overlap in cells faster and with greater precision. Using real data and simulations, Panzeri will train a neural network to cluster cellular components with similar properties and to identify functional patterns in complex tissues. Autofluorescence also plays a significant role in this context.
Marie Skłodowska-Curie Actions
The Marie Skłodowska-Curie Actions Postdoctoral Fellowships support outstanding research projects while fostering collaboration across countries, sectors, and scientific disciplines. As one of the EU’s most prestigious postdoctoral fellowships, they provide funding for excellent researchers across a wide range of fields. Funding is provided for projects lasting twelve to 36 months. Researchers of all nationalities, affiliated with an EU institution, and with no more than eight years of postdoctoral experience are eligible to apply.
Panzeri plans to systematically incorporate natural light emission into the analysis. Unlike previous approaches, which primarily regard autofluorescence as a disruptive signal that needs to be removed, he aims to evaluate its specific contribution and include it directly in the analysis. This would enable biomedical researchers to observe complex biological processes that were previously obscured by spectral signal limitations.
One of about 17,000
For Panzeri, the Marie Curie Fellowship is a huge recognition and a strong endorsement of his research project: he is one of around 1,600 EU postdoctoral researchers who will receive a fellowship in 2026, chosen from over 17,000 applicants.
(Saskia Schlesinger)
