Lung lobe segmentation method based on 3D full convolutional neural network and multi-task learning

Abstract

The invention discloses a lung lobe segmentation method based on a 3D full convolutional neural network and multi-task learning, and belongs to the field of automatic lung lobe segmentation. The method comprises the following steps: preparing lung lobe related data, and calibrating the lung lobe related data; preprocessing the original lung lobe CT image data and the lung lobe related data with the accurate label, and removing redundant background information; constructing a 3D full convolutional neural network based on multi-task learning; training the constructed 3D full convolutional neural network by using the pre-processed calibrated lung lobe related data and the synthesized learning error; and performing lung lobe segmentation on the input three-dimensional lung lobe CT image by using the trained 3D full convolutional neural network, and outputting a predicted lung lobe label. The lung lobe CT image data of the original size can be received, and the accurate lung lobe segmentation result can be automatically and rapidly generated.

Description

technical field [0001] The invention relates to the field of automatic lung lobe segmentation, in particular to a lung lobe segmentation method based on 3D fully convolutional neural network and multi-task learning. Background technique [0002] With the popularization and use of computed tomography (CT) technology in hospitals, CT has become one of the main technologies for the diagnosis and treatment of lung diseases. Lung lobe segmentation is crucial in the qualitative and quantitative analysis of lung diseases, such as the location of pulmonary nodules and the generation of diagnosis and treatment reports. The lungs are anatomically divided into five functionally independent parts: left upper lobe, left lower lobe, right upper lobe, right middle lobe, and right lower lobe. The boundaries of the lung lobes are called fissures. Since the fissures are usually incomplete and unclear, it is very cumbersome to segment the lung lobes manually. It usually takes 2-4 hours to mar...

AI Technical Summary

Problems solved by technology

However, the neural network based on the U-Net and V-Net structure has a large number of parameters and a large memory requirement, so the input is generally resampled and a lot of original information is lost. At the same time, the receptive field obtained by the network is also limited. These shortcomings make the network difficult. Produce accurate lung lobe segmentation results

The PLS-Net based on 3D depth separable convolution, dilated residual dense convolution block and input enhancement does not have to be resampled because of the small number of parameters, and can quickly extract more widely because the dilated convolution increases the receptive field. However, the results are still not accurate enough in the segmentation of the lung lobe boundary area

Images