Interview with Lorenzo Venieri, Data Scientist from Datrix’s R&D team.
The Digital Staining project based on Generative Adversarial Networks that Datrix is working on concerns a technology developed as part of the European research project OrganVision, funded by the European Union’s Horizon 2020 research and innovation program.
The project involves collaboration between Datrix and a consortium of prestigious universities and research centers, including: Arctic University of Norway (UiT), University of Southampton, JenLab GmbH, University Medical Center Hamburg-Eppendorf (UKE), Universitat de Barcelona, and University Hospital of North Norway.
This technological innovation opens a new era for medical research and aims to revolutionize the analysis of tissues and cells and ideally replace fluorescence microscopy with a faster and cheaper alternative based on artificial intelligence.
Through digital staining models, which predict fluorescence using deep learning algorithms, it is indeed possible to obtain the same result as traditional fluorescence microscopy starting from transmitted light images, but this solution offers a much more accessible alternative to expensive traditional techniques, expanding its application even to non-specialized laboratories.
It is Datrix’s Research & Development team that is following the project on the front line. The team, led by Matteo Bregonzio, Chief Technology Officer who coordinates the development activities, is composed of Data Scientists Lorenzo Venieri, responsible for the development of the AI model from data pre-processing to training and testing, and Andrea Masella, supervisor of deep learning model development.
Lorenzo, a young researcher expert in Artificial Intelligence with a solid academic background in mathematics, tells us more about the project.
What does the project consist of and what is Datrix’s role?
“The project focuses on the application of artificial intelligence (AI) to replace fluorescence microscopy with digital staining techniques. Starting from transmitted light images, the AI generates outputs equivalent to those obtained with fluorescence methods, greatly reducing costs and times,” says Lorenzo, adding: “Datrix plays a crucial role in the development of these AI models, thanks to its solid R&D experience gained in other Horizon projects such as CS-AWARE, CRIMSON, and BETTER, and international collaborations with academic and industrial institutions.”
Going into more detail, what is the difference between a standard “fluorescence staining” process and a “deep learning based” digital staining? How does Artificial Intelligence play a relevant role?
“The standard “fluorescence staining” process requires complex preparations and specific, expensive instrumentation, with a high dependency on specialized laboratories,” Lorenzo explains. “Classical staining has limitations due to the markers used to physically label the samples. These fluorophores can alter the normal behavior of the structures being observed. In addition, the spectral overlap between available markers makes it possible to observe only a very limited number of structures on the same sample, whereas this does not happen with digital staining.” Venieri specifies the advantages of adopting this technology, highlighting how “digital staining can reduce the need for physical preparations, using AI to virtually replicate the same results from transmitted light images, increasing scalability and accessibility.”
What are the benefits of adopting this technology and what are the practical use cases?
“The biggest advantage is that digital staining can drastically reduce laboratory costs, eliminating the need for expensive reagents and machinery and simplifying processes. Regarding practical use cases, it can speed up pharmaceutical research, facilitate diagnostics in hospital laboratories, and improve efficiency in biotechnology.”
Given the uniqueness of this field, how did you come to work on the project? What do you take away as a researcher from this work?
“I started collaborating with Datrix on this project for my master’s thesis and I am proud to work in an environment of such high technological competence,” Lorenzo says with pride. “This work represents an exciting challenge, as it combines my passion for artificial intelligence with the ambition to have a concrete positive impact on biomedical research. As a researcher, what I take home is the satisfaction of seeing how AI can transform traditional approaches, improving accessibility and efficiency in critical fields like biomedicine for the benefit of society.”
Facing AI challenges is Datrix’s mission: what are the future challenges/opportunities in adopting digital staining technology?
“The main challenges include the integration of AI technology into existing clinical and research processes, ensuring the robustness of models on a large scale and under varying conditions,” Lorenzo replies. “We are working on precisely this point, testing and training our model on data from different tissues and cell cultures acquired with different instrumentation and methodologies, to ensure that the model maintains high levels of accuracy and reliability regardless of variables.”
He concludes by explaining: “We collaborate with international strategic partners to obtain continuous feedback and adapt our solution to the real needs of laboratories. Our goal is to make digital staining a standardized and accessible practice, thus contributing to greater efficiency and innovation in the biomedical sector. The adoption of this technology represents an opportunity to democratize access to advanced analysis techniques, breaking down barriers of cost and complexity, especially in the pharmaceutical and biomedical sectors, and therefore we hope that this technology is implemented as much as possible in the healthcare world.”
Lorenzo Venieri studied mathematics and specialized with a master’s degree in Artificial Intelligence from the University of Bologna.
In the digital staining project, he worked on the development of the AI model, from data preprocessing to model training and testing.
He is part of the R&D team at Datrix, where he focuses on artificial intelligence and data analysis.
