Research On Application System Of Liver Tumor Auxiliary Diagnosis Based On Super-resolution Reconstruction

With people’s attention to health,regular physical examination has become very common.Liver diseases have occurred frequently in recent years,and the mortality rate of liver cancer is also gradually increasing.During the physical examination,the internal organs such as liver are usually examined by computed tomography(CT).Considering the radiation,low-dose CT equipment is usually used.As a result,the image resolution is not high and the details are not obvious,which puts great pressure on doctors to read the image information.In addition,the current diagnostic technology relies too much on doctors’ experience and is subjective.In the face of complex pathological features,doctors are prone to make wrong diagnosis.Aiming at the problem of low resolution of physical examination images,this paper proposes a new super-resolution reconstruction algorithm C-VDSR to improve the resolution of images.By replacing the L2 loss function adopted by VDSR(Very Deep Super Resolution)algorithm with Charbonnier penalty function,high-frequency features can be better captured.The experimental results show that the network model proposed in this paper has better performance than VDSR.Taking the 2-scale reconstruction effect as an example,the PSNR and SSIM indexes of C-VDSR in this paper reach 32.20dB and 0.9182 respectively,and the reconstruction time can be reduced to a certain extent.For the reconstructed high-resolution CT images of the physical examination liver,the deep learning algorithm U-Net network was used to perform automatic liver segmentation.After obtaining the liver regions,the algorithm was further used to realize tumor segmentation.The results showed that the performance of the segmentation network was improved after image reconstruction.For example,the Dice score of the liver segmentation network reached 87.4%.According to the above research results,combined with the system requirements,this paper designed a liver tumor auxiliary diagnosis system composed of image enhancement,liver segmentation and tumor segmentation.And this project is realized,in addition to the three functions before the addition of user management,CT image management and other functions needed in practical applications,and the function and performance of the system was tested.As the third party service system of OpenEHR,this system can realize such functions as super-resolution reconstruction of CT images in patients’ electronic medical records and auxiliary diagnosis of liver CT images.