Authors
Maciej Szymkowski, Łukasiewicz Research Network – Poznań Institute of Technology, maciej.szymkowski@pit.lukasiewicz.gov.pl
Bartosz Jura, Łukasiewicz Research Network – Poznań Institute of Technology, bartosz.jura@pit.lukasiewicz.gov.pl
Kacper Perz, Łukasiewicz Research Network – Poznań Institute of Technology, kacper.perz@pit.lukasiewicz.gov.pl
Jakub Gołąb, Łukasiewicz Research Network – Poznań Institute of Technology, jakub.golab@pit.lukasiewicz.gov.pl
Krzysztof Trusiak, Łukasiewicz Research Network – Poznań Institute of Technology, krzysztof.trusiak@pit.lukasiewicz.gov.pl
Aim of the project
The project’s general aim was related to the analysis of artificially grown cardiomyocytes, especially in terms of contractions and their directions. To observe these two factors, the authors used image analysis, processing methods, and Artificial Intelligence-based techniques (e.g., neural networks). In the project, the authors prepared a procedure for analyzing previously recorded videos and real-time observations performed with a high-quality microscope or camera. It needs to be pointed out that the novelty of the proposed approach lies in the usage of 3D videos of single cardiomyocytes. Until now, in recent research papers, their authors mostly focused on 2D microscopic videos (to analyze cardiomyocyte behavior); in this work, the authors used 3D videos to observe frequency, time of single contraction, and other measurable cardiomyocyte parameters. The researchers had not previously explored this path.
Short description of the problem addressed by this project
The problem the authors would like to solve is related to real-time observation of a single cardiomyocyte. It needs to be pointed out that new medicaments for cardiac diseases cannot be tested on humans or animals. It is too dangerous – especially when a new treatment is proposed (we are not yet aware of the way in which it will affect the heart). The question is – how can we solve this problem? The novel approach utilizes artificially grown cardiomyocytes, so that we can apply specific medicament to them and observe their behavior. However, the following problem is that sometimes the issue (e.g., halting of the cardiomyocyte or incorrect contraction, when e.g., only a specific part of the cardiomyocyte is actively taking part in the process instead of the whole tissue) is observable after a couple of hours from the administration of a drug dose. This is why the tissue needs to be observed automatically, and its parameters must be calculated in real-time (to observe the moment of problem occurrence easily).
Main results and achievements
The developed algorithms state the results of the work. During the experimental phase, we created 5 different approaches for contraction observation and measurement of the liveliness parameters (as BPMs, TTP90, or TTB90). The proposed solutions use both image/signal processing and analysis and Artificial Intelligence-based approaches (e.g., for image segmentation). It must be pointed out that the best methodologies guarantee high precision in measured parameters. It was observed that the maximum difference reached does not exceed 5%.
Moreover, the time needed to analyze the recorded video is near real-time. It means that the difference between the duration of the video and the time needed for its processing was 4 seconds at maximum. Finally, the authors also developed the software that the medical doctors or lab technicians can easily use. Both functionalities, i.e., observation in real-time and analysis of the videos are working as expected and were included in the program’s latest version. The results were also discussed with experienced researchers from the Polish Academy of Science – Institute of Human Genetics.
Conclusion
The project’s main goal was to observe whether it is possible to detect contractions and analyze their correctness within the videos of artificially grown human cardiomyocytes. That aim was reached – the authors not only proposed five different algorithms to fulfill that goal but also implemented them in the software that lab technicians or medical doctors can quickly and efficiently use. The results are also satisfactory – differences between the fundamental values of the parameters (as BPMs, TTP90, or TTB90) and those measured with our software do not exceed 5%. The authors would like to continue the project and propose new multimodal algorithms to observe many more parameters and implement efficient methodologies for detecting pathological changes. The proposed procedures can find practical use in new drug development processes.
Acknowledgments
The project was funded by Łukasiewicz Research Network – Poznań Institute of Technology, under the grants S-6117-0-2023 and S-6251-0-2024 and supported with resources for research by the Ministry of Science and Higher Education in Poland.