Liver tumor recognition method based on self-supervised dense convolutional neural network

Abstract

The invention discloses a liver tumor recognition method based on a self-supervised dense convolutional neural network. The method comprises the following steps: obtaining a liver slice data set from a magnetic resonance image of a patient, carrying out the segmentation of the slice data set, training a constructed dense convolutional network, enabling the trained dense convolutional network to serve as a coding module, constructing a self-supervised learning network, and carrying out the recognition of a liver tumor through the self-supervised learning network. Through dense connection of the coding modules, a tumor region of a part of an image in a slice data set is manually marked, then the whole slice data set is segmented, segmented image blocks are adopted to train a self-supervised learning network, and the trained self-supervised learning network is adopted to automatically identify tumors in the image. The dense convolutional neural network based on self-supervision is used for liver tumor recognition, a puzzle task is set as a self-supervision upstream training task, useful representations are learned from a large number of images which are not subjected to medical labeling and are used for learning training of downstream target tasks, and therefore the purposes of automatically expanding training data samples and improving the recognition efficiency are achieved. The dependence on expert experience and historical data is reduced, and the recognition accuracy of the liver lesion region is improved.

Description

technical field [0001] The invention belongs to the technical field of medical image recognition, in particular to a liver tumor recognition method based on a self-supervised dense convolutional neural network. Background technique [0002] Liver cancer (hepatocellular carcinoma, HCC) is one of the malignant tumors with the highest mortality and morbidity in the world. Although surgical resection is the first choice for the treatment of liver cancer, the early symptoms of HCC are usually not obvious in clinical practice. In the middle and advanced stages, or due to conditions such as severe liver cirrhosis and abnormal liver function, less than 25% of clinically discovered HCC can be surgically resected. For patients who do not have the conditions for surgical resection, transarterial chemoembolization (Transarterial chemoembolization, TACE)) is the main choice. The theoretical basis of TACE for the treatment of HCC is that 70%-75% of the blood supply of normal liver tissue...

AI Technical Summary

Problems solved by technology

[0002] Liver cancer (hepatocellular carcinoma, HCC) is one of the malignant tumors with the highest mortality and morbidity in the world. Although surgical resection is the first choice for the treatment of liver cancer, the early symptoms of HCC are usually not obvious in clinical practice. In the middle and advanced stages, or due to conditions such as severe liver cirrhosis and abnormal liver function, less than 25% of the clinically discovered HCC can be surgically resected

U-net based on convolutional neural network technology is currently a well-known neural network framework in the medical field. Generate segmented images while ensuring the preservation of high-resolution information. Deep learning has the problem of insufficient training sample data in the field of medical images, because it is difficult to manually mark a large amount of data and require a relatively high accuracy rate.

Images