Research On Lung Nodule Detection Based On CT Image Fusion Model

In recent years,with the popularization of public health knowledge and the improvement of the overall medical level,the diagnosis and treatment of lung cancer have been improved significantly.However the mortality rate of patients with advanced lung cancer is still at a high level,so early diagnosis has become the key to lung cancer prevention and treatment.The current early diagnosis method for lung cancer is mainly low-dose computed tomography technology.However,due to the large number of CT images scanned by this technology and the diverse morphology of lung nodules,it is very possible for diagnosticians to make misjudgments under high-intensity work.Therefore,it is imperative to adopt auxiliary methods to help doctors detect lung nodules.With the development of hardware and software,deep learning has been widely used in natural language processing,image processing and other fields.In medical field,the combination of medical imaging and deep learning has also become an important research direction in the industry.Therefore,we will investigate lung nodule detection based on convolutional neural network in lung CT images.In this paper,we adopt two effective image target detection algorithms-Yolo-v3 target detection algorithm and Faster R-CNN target detection algorithm,for the LUNA16 data.Through the comparison of evaluation criteria,the analysis shows that the Yolo-v3 model performs better.In view of the under-fitting phenomenon in the Yolo-v3 model detection results,combining the Faster R-CNN model,we not only improve the detection precism,but also reduce the detection error.