BPU 12 Congress

Evaluation of parameters of different architectures for segmentation of cell images

10 Jul 2025, 14:30

30m

Cntral Library 2nd Floor Main Hall ("Politehnica" University)

Poster S06 – Interdisciplinary Physics, Mathematical and Computational Methods Poster Session 3

Speaker

Griselda Alushllari (Epoka University)

Description

Griselda Alushllari 1,2, Arban Uka 1, Margarita Ifti 2
1 Department of Computer Engineering, EPOKA University, Tirana, Albania
2 Department of Physics, Tirana University, Tirana, Albania

Abstract

Unstained brightfield images pose a significant challenge for image analysis. U-Net architectures are a family of fully convolutional neural network which are used for image segmentation. These architectures are widely used for segmenting biomedical images due to their ability to detect fine spatial details in microscopy data. Understanding the impact different biomaterials have on cell images requires reliable segmentation methods. In this study, different architectures are used to segment brightfield microscopy images to determine cell confluence and assess the impact of various biomaterial concentrations on cell health. Results were compared for different loss functions, batch sizes to identify the optimal parameters for cell segmentation. Manual labelling of the images from an original dataset was conducted under the supervision of the medical practitioners. Among all the trained models, the ones achieving highest accuracy results were chosen for further analysis and cell confluence. Hybrid function with different weights of a combination of binary cross entropy and dice loss was found the best one.

References

1. Caicedo, J. C., Goodman, A., Karhohs, K. W., Cimini, B. A., Ackerman, J., Haghighi, M., ... & Carpenter, A. E. (2019). Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl. Nature methods, 16(12), 1247-1253.
2. Uka, A., Ndreu Halili, A., Polisi, X., Topal, A. O., Imeraj, G., & Vrana, N. E. (2021). Basis of image analysis for evaluating cell biomaterial interaction using brightfield microscopy. Cells Tissues Organs, 210(2), 77-104.
3. Uka, A., Polisi, X., Barthes, J., Halili, A. N., Skuka, F., & Vrana, N. E. (2020, August). Effect of preprocessing on performance of neural networks for microscopy image classification. In 2020 International Conference on Computing, Electronics & Communications Engineering (iCCECE) (pp. 162-165). IEEE.
4. Duro, X. P., Uka, A., Alushllari, G., Halili, A. N., Karras, D. A., & Vrana, N. E. (2025). Optimizing combination of feature extraction and classifiers in supervised classification of cells images. Edelweiss Applied Science and Technology, 9(2), 682-707.