Boundary Extraction of Stress Granules with Semantic Image Segmentation
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3170113Utgivelsesdato
2024-10-01Metadata
Vis full innførselSamlinger
- Master theses [220]
Sammendrag
This thesis examines the extraction of stress granule boundaries from imaging data using semantic segmentation, particularly focusing on overcoming the limitations of traditional methods when handling granules’ diverse shapes and sizes. Currently, this task is performed by the Granule Explorer program, which relies on an intensity-gradient based approach which fail to accurately capture complex geometries, limiting the subsequent analyses to a subset of all granules.
Given the importance of stress granules in cellular biology, particularly their malfunction, which has been linked to neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS), improving boundary extraction is critical for advancing our understanding of granule dynamics. In this work, an encoder-decoder-based machine learning approach is proposed to address these challenges. The result shows that the machine learning approach for boundary extraction is successful on a wider range of granules. However, whenthe extracted boundaries were used to estimate interfacial tension and bending rigidity through flicker spectroscopy, the results diverged significantly. This discrepancy suggests that further investigations into the stability of the underlying granule behavior model employed by Granule Explorer are necessary.