Circulating tumor identification using neural networks for monitoring cancer progression
Obonyo, Stephen Oketch
Cancer is the third most killer disease in Kenya after infectious and cardiovascular diseases. It contributes to a significant portion of annual national deaths, led by breast and prostate cancer. Existing cancer treatment methods vary from patient to another based on the type and stage of tumor development. The treatment modalities such as surgery, chemotherapy and radiation have been successful when the disease is detected early and constantly monitored. Ineffective treatment method and development of complications such as cancer relapse must be monitored as they are likely to cause more deaths. Detection of circulating tumor cells (CTC’s) is a pivotal monitoring method which involves identification of cancer related substances known as tumor markers. These are often excreted by primary tumors into patient’s blood. The presence, absence or number of CTC’s can be used to evaluate patient’s disease progression and determine the effectiveness of current treatment option. This research work proposed an adaptive learning-based, computational model to help in cancer monitoring. It identifies and enumerates CTC’s based on the auto-learned features from stained CTC images using deep learning methodology. The 3.0% error rate model, without human intervention, automatically learned the best set of representative features from labelled samples. The representations were used in enumerating and identifying CTC’s given a new test example.
Thesis submitted in partial fulfillment of the requirements for the Degree of Master of Science in Information Technology (MSIT) at Strathmore University
CTC Detection, CTC Enumeration, Neural Networks, Deep Learning