Deblurring of Microscopic 3D Spheroid Images Using GANs
Spheroids are 3D aggregates of cells that have become increasingly important in the study of cancer and drug discovery due to their ability to mimic in real tumour microenvironments. However, spheroid imaging presents several challenges due to its complex structure, irregular shape, and optical properties. In this thesis, we experiment with deep learning approaches to address these challenges and improve the quality of spheroid images. Specifically, we use a modified U-Net architecture and generative adversarial networks (GANs) to generate high-resolution spheroid images. We evaluate our approach on a dataset of spheroids and compare the performance of unsupervised and supervised neural network architectures for the deblurring of spheroid images. Our work provides useful information for further research in spheroid image analysis and has potential applications in cancer diagnosis and drug discovery.
Graduation Thesis language
Graduation Thesis type
Master - Computer Science
Dmytro Fishman, Mikhail Papkov