Blood vessel segmentation method and device, electronic device and storage medium

Abstract

The embodiment of the invention discloses a blood vessel segmentation method and device, an electronic device and a storage medium. The method comprises the following steps of respectively inputting an image to be segmented into at least two pre-trained set neural network models, and generating an initial segmentation result corresponding to each set neural network model; carrying out weighting processing on each initial segmentation result by utilizing a target model weight value corresponding to each set neural network model to generate a weighted segmentation result, the target model weightvalue being determined when the set neural network model is trained; and carrying out image post-processing on the weighted segmentation result to generate a segmented blood vessel image. According to the technical scheme, the Boosting modeling idea is applied to the field of blood vessel segmentation for the first time, the problem that a blood vessel segmentation result obtained through a single neural network model is low in precision is solved, and the effect of improving the blood vessel segmentation precision is achieved.

Description

technical field [0001] Embodiments of the present invention relate to medical image processing technologies, and in particular to a blood vessel segmentation method, device, electronic equipment, and storage medium. Background technique [0002] Vessel segmentation in medical images is a fundamental problem. For example, the segmentation of blood vessels in the liver is widely used in the diagnosis and treatment of liver lesions and the planning of liver surgery. [0003] At present, a common method for blood vessel segmentation is traditional image processing methods, such as threshold segmentation method or region growing method, etc., but the accuracy of blood vessel segmentation of these methods is limited, and it is difficult to meet clinical needs. Another commonly used method is the image processing method based on deep learning, such as fully convolutional neural network model (Fully convolutional networks, FCN), the fully convolutional neural network model U-net bas...

AI Technical Summary

Problems solved by technology

[0003] At present, a common method of blood vessel segmentation is traditional image processing methods, such as threshold segmentation method or region growing method, etc., but the accuracy of blood vessel segmentation of these methods is limited, and it is difficult to meet clinical needs.

Another commonly used method is the image processing method based on deep learning, such as fully convolutional neural network model (Fully convolutional networks, FCN), the fully convolutional neural network model U-net based on two-dimensional medical images, and the fully convolutional neural network model based on three-dimensional medical images. Convolutional neural network model V-net, etc. These methods are far superior to traditional image processing methods in terms of accuracy and robustness, but due to the limitations of imaging equipment and the complexity of the morphological structure of blood vessels, using any of the above convolution The blood vessel segmentation results obtained by the neural network model are prone to problems such as fewer blood vessel branches, wrong blood vessel borders, and segmented blocks.

Images