Gland cell segmentation method based on multilevel feature fusion network

Abstract

The invention discloses a gland cell segmentation method based on a multilevel feature fusion network, and belongs to the technical field of medical image processing. The gland cell segmentation network model provided by the invention comprises an encoder and a decoder; in the encoder stage, feature maps of an input end are subjected to down-sampling to generate feature maps of different scales, and then the feature maps are spliced with feature maps of corresponding proportions generated by maximum pooling in the encoder to realize multi-level feature input, and the propagation of image features is enhanced; in a decoder stage, according to feature map information of different levels reserved by down-sampling, the feature maps with corresponding sizes are spliced when up-sampling operation in the decoder stage is carried out, shallow image features are combined again, the loss of pixel positions is compensated, errors generated when feature image pixel positions are predicted and positioned in the decoding process are reduced, and an efficient gland cell image segmentation task is achieved.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a segmentation method for glandular cell images. Background technique [0002] A normal gland is composed of a tubular structure with a luminal area and an epithelial nucleus surrounding the cytoplasm. Malignancies arising from glandular epithelial cells are called adenocarcinomas. Conventional treatment plans often depend on the grade and stage of the adenocarcinoma. Annotating and segmenting gland morphology in histopathological images is an important step and means for medical experts to judge the grading of cancers such as colon, breast and prostate. This work is very important for the treatment of patients' conditions, because accurate gland segmentation helps targeted and personalized treatment design, thereby improving the cure rate of patients. However, manual annotation and segmentation of glandular cells by medical experts requires a high ...

AI Technical Summary

Problems solved by technology

[0004] (1) The shape of benign and malignant glands is very different. The benign glands generally have a circular structure, while the malignant glands have an irregular structure. Larger, which increases the difficulty of accurately delineating individual glandular cells from tissue;

[0005] (2) The gap between adjacent gland cells is narrow, and a higher resolution is required for the division of the boundaries of each gland cell, which increases the difficulty of accurately dividing each gland cell;

Regardless of the aspect, it is a great challenge to achieve the desired effect.

Images