Classification of CTC on Fluorescence Image Based on Improved AlexNet
Kohei Kisanuki, Guangxu Li, Tohru Kamiya · 2022 22nd International Conference on Control, Automation and Systems (ICCAS) · 2022
These days, cancer has been the most primary cause of death in Japan. Cancer often progresses by repeating metastasis, so early detection and early treatment are important. Analysis of Circulating Tumor Cells (CTCs) has come to gather attention as a new biomarker that CTCs can detect primary cancer in human body. However, the number of CTCs in a billion blood cells is only a few, and detecting CTCs is very hard. Accordingly, we propose an automatic detection method of CTCs from fluorescence microscopy images to enable quantitative analysis by computer. This method consists of two parts. The first part, we use some series of filtering to the images and, new dividing method some overlapping nucleus then, from the images cut out the region of interest (ROI). The second part is distinguishing images by using CNN. We applied the proposed method to 5040 images of 6 samples. As a result, we obtained TPR:94.59%, FPR:6.544% by using AlexNet based model.