Quantitative Statistical Study On Microvascular Network Of Liver Tumor By Micro CT

Objective: This study was based on the analysis of X-ray phase contrast microscopic images of mouse liver tissue and postoperative liver tumor tissue to analyze three-dimensional microvascular network images and tumor typing studies.The aim is to perform non-destructive three-dimensional microscopic imaging of liver tumors using new imaging techniques Improve the accuracy of liver tumor micro-pathological structure analysis and auxiliary diagnosis.Methods: Image acquisition of liver tissue based on X-ray phase contrast imaging,AVIZO software is used to perform qualitative and quantitative analysis on the microvascular network of ROI images.After preprocessing the ROI images by MATLAB software,gray histograms,gray co-occurrence matrices,improved gray co-occurrence matrices,Gray gradient co-occurrence matrix,Tamura texture feature,Hu invariant moment,discrete wavelet transform,and gray difference statistics are 69-dimensional comprehensive features.Features are optimized using AUC area and PCA two-dimensional reduction methods.Logistic regression,NB,SVM,RF,Bagging,and BP neural network classification algorithms,using 10-fold cross-validation to classify diffuse,ulcerous,and microcapsule-type liver tumor X-ray phase contrast images,and evaluate the classification effect of each algorithm through parameter evaluation.Results: Animal liver tissue dehydration studies have shown that after 10% neutral formalin fixation and ethanol dehydration concentration gradients are slowly transitioned from 40% to 100%,the microvascular edges of tissue tomographic images are sharp,the three-dimensional vascular network is regular,and both sides of the blood vessel edges are gray.The degree of symmetry is high,the grayscale changes are obvious,the blood vessel diameter deformation is small,and the CNR region of the image is high.The results of the liver tumor vascular network show that the pathological sections and X-ray contrast images correspond to each other in spatial structure.Comparing the microvascular network of the two-dimensional and three-dimensional images,it is found that tumors with hematogenous development cause the vascular network to gradually cross warp,the spatial structure is disordered,and the intrahepatic A large amount of collagen deposition and regeneration nodules were formed,and the vascular tree structure disappeared.With the development of the tumor,the blood vessel diameter,microvessel density,and fractal dimensions showed an upward trend.There were statistical differences between the groups(P <0.05).The liver tumor classification results show that the comprehensive features and AUC screening features are suitable for the classification of diffuse,ulcerous,and microencapsulated liver tumors.Among them,the diffuse classification has the best effect,and the classification accuracy has reached 90% and above.The BP neural network and The SVM classification algorithm is more suitable for the classification of three types of liver tumors,and the classification accuracy rate is more than 98%.Conclusion: The liver dehydration experiment in mice provides a reference for the treatment of soft tissue X-ray phase contrast imaging samples.The qualitative and quantitative analysis of liver tumor microvascular network provides a new basis for the study of liver tumor microstructure and early diagnosis.The liver tumor typing results can be used to assist doctors Patient diagnosis and evaluation of drug efficacy.