Research On Medical Image Diagnosis Algorithm Based On Integrated Deep Learning

In the 21st century,cancer is still a serious problem that plagues people's lives.According to a survey by the American Cancer Organization,there are nearly 1.8 million new cases of lung cancer worldwide and 1.6 million deaths from lung cancer.CT images are one of the most commonly used and robust imaging techniques for tumor detection,diagnosis,and subsequent treatment.By visualizing human tissue through X-ray based absorption,radiologists can assess the degree of malignancy based on the density and morphology of the tumor,but the reliability of the prediction is highly dependent on the physician's experience and different radiologists may make different diagnosis.Due to the complex relationship between tumors,its appearance does not necessarily mean the occurrence of cancer.In some complicated situations,it is difficult for even experienced radiologists to reach a consensus.Therefore,there is a great need and interest in developing an automated diagnostic system based on CT images to assist radiologists in the diagnosis of lung cancer.Computer vision-based models can quickly examine lung CT images at the same level,and they are unaffected by physical and mental states.Computer aided diagnosis(CAD)refers to the establishment of a stable and reliable machine learning model by means of imaging technology,image processing methods and other available physiological and biochemical methods,and then assists doctors in judging lesions through computer analysis and calculation.In order to achieve the purpose of improving the accuracy of the diagnosis results.Artificial neural network is one of the most commonly used machine learning models in medical image classification.It simulates the working mechanism of human brain neurons to complete highly nonlinear processing of signals.And it has many excellent properties such as self-improvement,memory,and prediction,so it can achieve the effect of assisting doctors in diagnosis.In medical image classification and disease diagnosis,artificial neural network-based machine models show better and stable performance than traditional methods(such as probability statistics).The field of image recognition based on deep learning has made remarkable progress,mainly due to the availability of large-scale annotation datasets(such as ImageNet)and the development of deep convolutional neural networks(deep CNN).For data-driven learning based algorithms,large-scale annotated data sets with representative data distribution characteristics are critical for learning more accurate or generalized models.However,due to the difficulty of data collection and the high quality labeling,there is currently no large-scale annotated medical image dataset like ImageNet.This makes the machine learning model for medical image classification easier to fall into the over-fitting problem,i.e.,the network can fit very well on the training samples,but it performs very poorly for the diagnosis of new case samples.In this paper,we propose a corresponding solution around the above problems,and through a large number of experiments to complete the verification of the effects of key algorithms in the CAD system.We have completed the diagnosis of CT images by large-scale integration of appropriate deep convolutional neural networks.It has been proved that the integrated learning network can show better diagnostic capabilities.Finally,we verified the performance of the system on two public medical image databases and analyzed the experimental results.