| Abstract [eng] |
Nowadays medicine, biology research has a lot of advanced technologies but one thing that is always slowing down the advancement of them is big counts of data proccessing, simmilarity findings, conclusion and relationship creation. To speed up the reasearch and find new relationships between the data – we can use artificial intelligence methods. The aim of this project – using artificial intelligence automate, speed up biological cells research, with the main goal of cell‘s gap junctions. To detect them, calculate them and evaluate the quantitive parameters. Artificial intelligence methods can detect alive cells but the rapid advancement in the field creates challenges when it is need to cross-compare models, to select most optimal one for the current task at hand. To choose the best model – cross-comparison is done between standard U-NET model vs „Detectron2 RCNN50“ vs YOLOv11. The research results show that „Detectron2 RCNN50“ achieves the highest performance, which are further improved by doing additional post-processing on results. The created cell segmentation and data extraction architecture marks the cells and calculates their counts in pictures. Using segmentation data it is capable of extracting cell quantitative parameters and as such using it is possible to calculate and plot cell‘s hemi-channels permeability scores after transfection, giving the ability to compare transfected cells data with non-transfected cells. |
