Lung brush film cell identification and segmentation method based on deep learning

Abstract

The invention discloses a lung brush film cell recognition and segmentation method based on deep learning, and the method comprises the steps: obtaining a lung adenocarcinoma hyperspectral original image, and carrying out the preprocessing of the lung adenocarcinoma hyperspectral original image; labeling the preprocessed original hyperspectral image of the lung adenocarcinoma, and extracting small hyperspectral images; inputting the small blocks of hyperspectral images into a two-dimensional classification network, inputting the central point spectral curve of the small blocks of hyperspectral images into a one-dimensional classification network, and calculating supervision loss values and consistency loss values of the two-dimensional classification network and the one-dimensional classification network to obtain a final two-dimensional classification network; inputting the original hyperspectral image of the lung adenocarcinoma into the small sample segmentation network to obtain a test image; and classifying the test image by adopting the final two-dimensional segmentation network, and mapping a classification result back to the original hyperspectral image of the lung adenocarcinoma to obtain a segmentation result of the lung brush film cells. Accurate cell edges can be obtained, extracellular matrixes are prevented from affecting cell segmentation, and better performance is achieved.

Description

technical field [0001] The invention relates to the field of pathological diagnosis of lung brush slices, in particular to a method for identifying and segmenting lung brush slice cells based on deep learning. Background technique [0002] Lung cancer has been the leading cause of tumor-related death due to its aggressiveness and lack of effective treatment strategies, of which lung adenocarcinoma (PUAD) is the most common subtype of lung cancer. In addition, tumor cell invasion and metastasis are important criteria for pathologists to diagnose lung adenocarcinoma. Therefore, cell detection and segmentation of lung adenocarcinoma plays a pivotal role in lung cancer diagnosis. At present, with the rapid development of deep convolutional neural networks, the application of deep learning in medical image segmentation has risen significantly. The current pathological cell segmentation algorithms are mainly divided into semantic segmentation and instance segmentation. Semantic...

AI Technical Summary

Problems solved by technology

In addition to the problem of noise labeling, the labeling of microscopic pathology images is also too time-consuming

Images