An improved yolov3 algorithm for pulmonary nodule detection

Lung cancer is one of the highest incidence rate and mortality rate in the world. Lung nodule is an early manifestation of lung cancer. Therefore, the early detection of pulmonary nodules is of great value in preventing lung cancer. At present, the target detection method of deep learning is widely used in medical image processing. In order to improve the efficiency and accuracy of pulmonary nodule detection, and reduce the missed diagnosis rate and misdiagnosis rate of pulmonary nodule, this paper proposes a pulmonary nodule detection method based on improved yolov3. This method first proposes to use the superimposed extended convolution, and then processes the 104 * 104 feature map output from the second residual block in darknrt-53, which is the backbone feature extraction network of yolv3, and the superimposed extended convolution. Finally, after twice down sampling and 52 * 52 detection feature map fusion, a new 52 * 52 detection feature fusion map is formed to improve the network feature extraction ability. A large number of experiments were carried out on luna16 open lung CT image data set. The experimental results show that the average accuracy of the improved model is improved from 70.5% to 73.9%, and the convergence effect of the improved model is better than that of the original model.