Research On Medical Image Classification Method Based On Convolutional Neural Network

With the development of information technology and medical imaging technology,the era of medical big data has come.The explosive growth of medical images requires more experienced doctors to complete the diagnosis of diseases,which will undoubtedly bring heavy pressure to doctors,and misdiagnosis and missed diagnosis occur from time to time.Computer aided diagnosis system can overcome the subjective factors such as subjective experience and fatigue degree,which can not only improve the diagnostic efficiency of radiologists,but also improve its accuracy.At present,convolutional neural network has been used more and more in image recognition.It has many advantages,such as no manual intervention,automatic extraction of image features and strong learning ability.In this paper,convolutional neural network is used to analyze and study two kinds of medical images,one is lung X-ray image for screening lung diseases,the other is skeletal X-ray image for identifying skeletal injury.Shallow convolution neural network has the characteristics of light structure,few parameters and strong applicability,but the effect of medical image classification is not ideal.In this paper,the shallow convolution neural network is improved,and a medical image classification method based on improved convolution neural network is proposed.The experimental comparison with classical convolution neural network proves the superiority of the improved method.It effectively solves the problem that the classification effect of shallow convolution neural network is not ideal.Aiming at the shortage of medical image data and the problem of over-fitting in training,the transfer learning is applied to the field of medical image classification,and a medical image classification algorithm based on transfer learning is proposed.The validity of transfer learning is verified by comparing with the traditional model of random initialization parameters.Transfer learning has been applied to the field of medical image classification successfully.On the basis of the medical image classification algorithm based on transfer learning,a medical image classification algorithm based on multi-scale transfer learning is proposed.The feasibility of this algorithm is verified by comparative experiments.This study has certain reference value in medical image feature analysis and classification.