Master’s Thesis Defense at the College of Information Technology on Predicting Phenotypic Representations of Cancer Cells Using Siamese Networks

Duhaa Fadill Abbas
The Department of Software Engineering at the College of Information Technology – University of Babylon witnessed the defense of a Master’s thesis submitted by the student Tayba Hussein Shaman Hussein, entitled:“Image-Based Prediction of Phenotypic Representations of Chemotherapy-Treated Cancer Cells Using Siamese Networks”,under the supervision of Assistant Professor Dr. Sura Zaki Naji. The defense was conducted in a rigorous academic and scientific atmosphere, reflecting the advanced level of the research and its significance at the intersection of artificial intelligence, medical image analysis, and modern biological research.

The thesis highlighted the transformative role of high-content cellular microscopy in biological research and pharmaceutical drug discovery, particularly in enabling large-scale and precise identification of cellular phenotypes under various chemotherapeutic treatments. It also discussed the limitations of traditional image analysis techniques based on hand-crafted features, which often suffer from sensitivity to experimental noise and a limited ability to capture the complex variability inherent in cellular phenotypic patterns.

The researcher proposed an advanced framework that integrates Masked Siamese Networks (MSN) with weak supervision to extract meaningful and biologically relevant phenotypic representations from high-content fluorescence microscopy images. The framework is built upon a vision transformer backbone, leveraging siamese networks to align and compare representations derived from different masked views of the same cell image. This approach encourages the model to focus on essential structural information without relying on extensive and strongly annotated labels, which are typically scarce and costly to obtain.

Overall, the study confirms that combining self-supervised representation learning with weak supervision provides a robust and biologically meaningful framework for high-content cellular microscopy analysis. This work represents a valuable contribution to the fields of medical image analysis and computational biology, and underscores the growing potential of artificial intelligence techniques in supporting biomedical research and the development of advanced therapeutic strategies.