Abnormal Bone Image Detection Method Based On Convolutional Neural Network

In recent years,deep learning technology has been gradually introduced into the field of medical image processing,and the automation of medical facilities is one of the important directions of scientific research.The interpretation of X-ray images requires high medical expertise.In the face of the increasing number of patients with skeletal diseases,medical personnel need to face a large number of X-ray skeletal images every day.Reading and diagnosis,using traditional methods for investigation,can easily cause visual and nerve fatigue,resulting in pathological diagnosis errors.The method of manually reviewing the film can achieve the standard of clinical accuracy,but it will aggravate the pressure of medical resources and is inefficient.At present,there are few studies on the method of automatic detection of bone abnormalities.Constructing a precise auxiliary diagnosis system can greatly improve the diagnosis efficiency.In view of the above problems,this paper proposes an X-ray bone image anomaly detection method based on deep learning.The main research contents are as follows :Firstly,in order to solve the problems of blurred bone edge lines and unclear boundaries with other tissues in bone images due to imaging reasons,this paper proposes an image enhancement algorithm combining wavelet transform and histogram equalization.Firstly,wavelet transform is used to strengthen the details of the image to remove the noise in the image,and then histogram equalization is used to enhance the global contrast of the image,so that the bone structure in the image can be more clearly displayed,which is convenient for the subsequent network to extract the features of the image.By comparing the images before and after processing,it can be seen that the method proposed in this paper better highlights the details of the image and the bone structure is clearer.At the same time,the experimental results show that the information entropy and contrast of the image are improved.Based on the Densenet network,this paper proposes an improved automatic detection method for abnormal bone images.This method is improved on the basis of Densenet network,and the network structure is adjusted.The main problem is to improve the training accuracy of neural network and obtain higher anomaly detection accuracy.CBAM attention mechanism is added to the original Densenlayer,which integrates spatial attention and channel attention,suppresses unnecessary features,and strengthens the feature extraction ability of the network.The activation functions in the backbone network are replaced by the Mish activation function,and the Transition block is optimized.The average and maximum pooling strategies are combined to increase the anti-interference ability of the network.The improved network is verified by using the large public radioskeletal image dataset MURA.The results show that compared with the traditional network,the improved Densenet has improved the accuracy of skeletal anomaly detection in all parts,and its average accuracy is 2.6 % ～ 5.9 % higher than the benchmark method.The area under the receiver operating characteristic curve is increased by 4.2 % ～ 9 %,and the balance accuracy is increased by 2.3 % ～ 6.8 %.Figure [41] Table [12] Reference [81]...