Lung tubercle automatic detection method based on 3D convolutional neural network

Abstract

The invention discloses a lung tubercle automatic detection method based on a 3D convolutional neural network(CNN). According to the invention, detection is creatively divided into two phases: (1) a candidate lung tubercle detection phase and (2) a false positive lung tubercle screening-out phase. In every phase, a unique 3D CNN is constructed and trained to be suitable for detection and screening-out of lung tubercles. The 3D CNN of the first phase is used to detect candidate lung tubercle positions of suspected lung tubercles preliminarily, and the 3D CNN of the second phase is used to screen out the false positive lung tubercles of the candidate tubercles, and finally, the positions of the tubercles in a whole lung CT image are found. The existence condition of the tubercles in the whole lung CT image is automatically detected, and by comparing with a conventional artificial tubercle detection way, the lung tubercle automatic detection method provided by the invention has advantages of high detection accuracy, strong robustness, high efficiency, short consumed time, and realizing convenient and effective detection of lung tubercles.

Description

technical field [0001] The invention belongs to the technical field of lung CT image detection and screening, and more specifically relates to a method for automatic detection of pulmonary nodules based on a 3D convolutional neural network (CNN for short). Background technique [0002] At present, due to long-term smoking, air pollution and other reasons, the number of lung cancer cases is increasing rapidly all over the world. Lung cancer is a type of cancer with high morbidity and mortality in the world. According to data, the average 5-year survival rate of lung cancer in the world is only 16%, while the 5-year survival rate of early stage (I stage) lung cancer can reach 65%, but unfortunately only 10% of patients can survive lung cancer. Disease is detected at an early stage and treated accordingly. Evidence shows that annual lung computed tomography (CT) screening of lung health among high-risk groups can reduce lung cancer mortality by 20%. [0003] Lung nodules are...

AI Technical Summary

Problems solved by technology

Such traditional methods often have problems such as heavy workload, time-consuming, error-prone and omissions, etc., and the screening results also mostly depend on the professional technical level of the individual medical staff

Images