RS-SVGG: An Intelligent Circulating Tumor Cell Detection Method Based on VGG
Haosen Huang, Zhenning Wu · 2022
The image-based C irculating T umor C ell detection is a powerful tool for both the diagnosis and the recovery process of cancer. However, healthy cells are dominant in the captured image in most scenarios. That makes the task of recognition CTCs from healthy cells can barely be manually achieved. This paper presents RS-SVGG, a machine learning based automatic CTC detection method based on VGG network. RS-SVGG adopts Multi-scale Region Selective Search algorithm to split the original large size image into small fixed-sized suspected target CTC images. Then final CTCs are detected from these target CTC images by SVGG, which is an optimized VGG network. The proposed method was evaluated using real clinical images. The experimental results show that RS-SVGG outperforms traditional image processing methods and can achieve up to the accuracy of 97.56%, which also outperforms the detection accuracy of human.