Pulmonary nodule automatic detection method based on CT image

Abstract

The invention relates to a pulmonary nodule automatic detection method based on a CT image. The method comprises the following steps: 1, establishing a training and testing sample set by utilizing a public lung data set: reading an original file of a CT data set, synthesizing an image where a pulmonary nodule is located and adjacent images of a front layer and a rear layer of the pulmonary noduleinto a group of three-channel data, performing pseudo-colorization processing on the three-channel data, and expanding the sample set by utilizing a data enhancement method; 2, establishing a pulmonary nodule detection network: constructing a feature extraction backbone network, a multi-scale feature layer fusion network and a candidate box regression and prediction network; 3, training the pulmonary nodule detection network by using the training sample to obtain a trained detection model; and 4, verifying the detection model in the test data set, and detecting the pulmonary nodule position. The pulmonary nodule detection network is trained by using the training sample to obtain the trained detection model, false positive is reduced on the premise of ensuring the overall detection rate, and doctors are assisted to improve the diagnosis efficiency.

Description

technical field [0001] The invention relates to an automatic detection method for pulmonary nodules based on CT images, and relates to the field of computer-aided diagnosis of medical images. [0002] technical background [0003] Lung cancer is one of the malignant diseases with high mortality rate worldwide. In 2016, statistics from the International Agency for Research on Cancer (IARC) showed that the number of cancer patients is increasing every year, and lung cancer accounts for about 85% of various cancers, making it the largest disease. The clinical manifestations of lung cancer are not obvious in the early stage, and patients cannot pay attention to it. When lung cancer develops to an advanced stage, the spread of cancer cells makes it almost impossible to save the patient. Early lung cancer is mainly manifested as asymptomatic small pulmonary nodules. Therefore, regular lung examination can detect the occurrence of lung cancer early, and timely treatment can signif...

AI Technical Summary

Problems solved by technology

At the same time, the existing network loss function does not fully consider the prediction loss of difficult candidate boxes, which leads to the problem of low accuracy and large number of candidate boxes generated by the network model

Images