Researchers have taken AI-based biological image analysis to a new level

Researchers at the Szeged Biological Research Centre and University of Helsinki have released a new version of their Advanced Cell Classifier (ACC) software and shown that it is a promising tool for exploring biological data in a continuous manner. The new method can help to understand the molecular mechanisms of cardiovascular diseases and even tumours.

Most image analysis software can only analyse biological processes in stages. To solve this problem, Péter Horváth and his research team together with FIMM researchers Lassi Paavolainen and Vilja Pietiäinen developed the so-called Regression Plane (RP) concept, which enables - by combining microscopy image analysis and artificial intelligence - high resolution continuous analysis of biological processes.

The team, with members from the ELKH Szeged Biological Research Centre in Hungary, the BIOMAG Group and Group Kallioniemi at FIMM and Faculty of Medicine, University of Helsinki, has just released a new version of the Advanced Cell Cassifier software and published it in Nature Communications.

In the publication, the researchers demonstrate the usefulness of the method by testing its performance in several complex biological processes. The results showed that the method can facilitate the exploration of mitosis (a crucial process for cancer development) in unprecedented details. The method was also capable of revealing sophisticated differences in the cell’s lipid storage system that plays a crucial role in the development of cardiovascular diseases.

The technology, in combination with a previously developed method by Péter Horváth's group, enables the isolation of a single target cell and assessing the molecular changes taking place within it. The software can also be used to study the process of cell differentiation. For this purpose, a fruit fly model was chosen because its cell differentiation process is very similar to that of humans. The research team examined a previously poorly characterized blood cell differentiation phenomenon. The results revealed that the two cell subtypes studied are actually two extreme stages of a cell size continuum instead of definitely distinguishable cell types.

Thanks to artificial intelligence, the new method is a fast, objective technology with a performance comparable to that of humans. Its sensitive classification system is made more efficient by using active learning algorithms, which allow the program to communicate with researchers. When it encounters a morphology that is unfamiliar to it, it asks the user for help in making a decision, just as a small child asks a question when unsure about a jigsaw puzzle.

The developers have made Regression Plane available to everyone. It has been developed with a strong focus on a user-friendly environment and ease of use, with tutorial videos to help users get the most out of it.

The project was initiated when Péter Horváth worked as a TEKES (currently Business Finland) funded FiDIPro Fellow at the University of Helsinki. FIMM research infrastructure services have also been utilized in the project: FIMM High Throughput Biomedicine Unit contibuted to the study by providing access to high throughput robotics and siRNA library and the FIMM High Content Imaging and Analysis unit for HC-imaging. 

Original publication: Szkalisity, A., Piccinini, F., Beleon, A. et al. Regression plane concept for analysing continuous cellular processes with machine learning. Nat Commun 12, 2532 (2021).