Abstract

The morphological features of glands provide a reliable basis for pathologists to diagnose colon cancer correctly. Currently, most methods are limited in their ability to address blurred edges, adhesions, and morphological differences in glands when working in the spatial domain. In this study, a new perspective is proposed to segment glands by utilizing the interaction between Fourier and spatial domain features. Specifically, a Fourier transform is used to analyze the frequency characteristics of pathological images in which the phase component is related to image structural information and the amplitude component contains image intensity information. A Fourier-based colon cancer gland segmentation network (FFS-Net) is proposed, which gradually constructs feature representations in both the spatial and frequency domains through spatial frequency edge interactions. This network comprises spatial subnetworks, frequency subnetworks, and an edge phase fusion module (EPFM). To alleviate the interference of complex background information in feature discrimination, the spatial subnetwork collects high-frequency contour information regarding the glands. Then, the EPFM completes the spatial-frequency feature interaction to add edge-related constraints to the extracted features, enhance edge representation, and obtain global features from the frequency domain, thereby, addressing significant differences in gland morphology. Finally, learnable Gabor filters are introduced to enhance edge texture detail information in the frequency subnetwork, which further improves the segmentation of glands exhibiting blurred edges. The experimental results demonstrate that FFS-Net’s performance on two representative datasets was competitive and highly interpretable.

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.