Author: Hamdah Alanazi, Silvia Pella 👨🔬
Affiliation: FAU, Florida Atlantic University 🌍
Purpose: The appearance of breast cancer in the global list of most common cancers worldwide requires
research for ultimate treatment approaches including radiation therapy to reduce deaths from breast
cancer. The thesis investigates deep learning approaches for breast cancer segmentation within medical
imaging platforms to boost radiation therapy accuracy and performance. The approach will split breast
CT images by region to generate pixel-level classifications which will produce accurate class labels per
pixel. This classification method promotes peak accuracy in tumor area segmentation while supporting
a trained system to separate healthy tissue from organs at risk.
Methods: The research paper will develop a deep learning system and simulate its integration into existing
medical imaging platforms to automate clinical target volume (CTV) delimitations for breast cancer
treatment optimization. The breast CT image segmentation system development will use free open-
source programming in Python to create the deep learning framework.
Results: The precise targeting of radioactive isotopes towards tumors made possible by optimal segmentation
will improve both therapeutic targets and preserve vital organs from unwanted side effects. The
expected patient outcomes from the use of the proposed deep-learning model during radiation therapy
treatment of breast cancer will include superior tumor localization performance that delivers minimal
radioactive exposure to nearby healthy tissues and improved segmentation accuracy compared to
conventional methods, which enhance the detection of breast cancer tumors. The proposed system’s
automation of the tumor boundary detection process will improve the efficiency and reliability of
radiation therapy planning, hence higher quality of breast cancer treatment plans.
Conclusion: Through this thesis, the precision increment from the proposed segmentation method plays a central
role in clinical target volume detection and risk organ protection applications leading to better deep
learning-based oncological precision medicine.