Cell cluster detection of thyroid FNAB-WSI via deformable convolution with frequency channel attention

Meng Sun, Jing Zhang, Shimei Zhao, Xiaoguang Li, Li Zhuo

Abstract: Highlights•The sliding window and ADA are used to solve the issue that the light-colored cell cluster objects are easy to lose.•We combine DCN and FCA to overcome the impurity interference and the irregularity of cell clusters.•We introduce a Sparse R-CNN model with sparse learnable proposals to learn the location information of sparse cell clusters.•We collect a real-world dataset thyroid FNAB-WSI for cell cluster detection.