Classification Of MRI Brain Images And Liver Cirrhosis

The development of modern medical science has been continuously developing along with the development of medical diagnostic equipment. Devices for diagnosing make physicians from the visible surface damage, to the cell structure by using the microscopic, and till now we can contemporary imaging equipment for auxiliary organ structure, and constantly improve the auxiliary diagnosis and treatment of science.In 1611, Kepler designed the first stage of compound microscope,1874 Germany pathologist Virchow who was the first one using a microscope observation to observe human cells, created a precedent for a diagnosis, making the cognition of human medicine from the visible damage towards the cell pathological diagnosis.In addition, large imaging equipment has also experienced three milestones leap. German physicist roentgen discovered the X-ray in the laboratory (X-ray) in 1895, which made it possible for human beings to check the inside body in non-invasive examination method.1972, Hounsfield and Cormack announced the birth of the first brain-CT which was used for the brain damage(Tomography Computed, the electronic computer tomography), it allowed us to get a more clear bony structure through the method of tomography. In 2003, Mansfield and Lauterbur won the Nobel Prize for the invention of the Magnetic Resonance Imaging, and the new equipment let us to observe the three-dimensional structure of the lesion and its localization clearly.With more and more medical images, the diagnosis needs more sophisticated, the surgeons’burdens increasing, computer aided diagnosis (computer-aided diagnosis CAD) has become more and more important in modern medical. What is the CAD’s work is that converting the diagnostic process of the doctor’s experience into an image analysis algorithm and computer language, using automatic analysis and processing methods. In a word CAD is created to reduce the workload of doctors, to help the patients get the more details about their disease.At the present, image classification, which is used to classify the type of disease and the pathological stage of the disease has gradually become the focus of research in the field of image analysis. Because it is closely related to image classification and segmentation, segmentation is usually an important part of the classification, image classification can not only provide segmentation of intermediate results which is useful to help doctors extended image information. What’s more, the classifier can produce disease label to help doctors judge the disease type and severity.In this paper, we study on two tasks in classification 1)MRI classification of brain tumors,2) the quantitative staging of liver cirrhosis.Because of the danger in the special position and the harm of the operation, brain tumor is the highest incidence of systemic cancer tumors. At the same time, with increasing tumor volume brain edema and water increased, leading to increased intracranial pressure, serious caused the patients with sudden deathTherefore, the timely diagnosis of brain tumors has become particularly important. MRI has become the most widely used imaging diagnostic equipment for brain tumor because it is no radiation and has a super high soft tissue resolution. Using MRI as auxiliary equipment for brain tumor, can examine the early lesions, show a more clear tumor morphology, which is useful to judge the category of tumor and to pre symptomatic treatment and the therapeutic effect of tracking.Currently brain tumor incidence types, glioma, pituitary tumor and meningioma. Due to differences in the morphology and etiology, the treatment options also vary, so that how to classify the tumors accurately is very important to the doctor to help them formulate with treatment.At present, tumor classification is usually relied on the analysis of the doctor’s experience and judged based on the organizational structure of the tumor morphology, grayscale and the surrounding tissue information. In addition, the different modes of MRI images can reflect the different information of the tumor region, so when the doctors makes a diagnosis, they often combine with a variety of modes of the brain MRI image as a reference. For example, T1 contrast enhanced image can reflect the contrast enhancement region, while T2 image can reflect the water area. Therefore, the traditional classification methods depends heavily on the experience of doctors and intervention of doctors, it is difficult to meet with the growing number of. clinical needs.The main purpose of this paper is design a kind of automatic tumor classification algorithm, which is used to reduce the workload of physicians. The study showed that the same kind of brain tumors often have the same structural characteristics, while the different types of brain tumors have different surrounding tissues. Thus in this paper, we proposed a new classification method, which expand the region of brain tumor, take a certain range of neighborhood organizations of the tumor into consider.In general, the method we proposed in this paper, the classification of brain tumors for MRI include two main ideals. Firstly, because tumor surrounding tissues can offer important clues for tumor types, we take the augmented tumor region as the ROI instead of the original tumor region which was used in traditional methods. Second, the augmented tumor region is split into increasingly fine ring-form sub-regions, while the square patch was the most widely used features areas. In order to verify the effectiveness of the above ideas on the improvement of classification accuracy, we apply three common methods histogram, gray level co-occurrence matrix and Bag-of-Words(BOW) to show the different results.In this paper, we collected 3064 slices from 233 patients, containing 708 meningiomas,1426 gliomas 930 pituitary tumors. Compared with using tumor region as ROI, using augmented tumor region improves the accuracies to 82.31% from 71.39%,84.75% from 78.18%, and 88.19% from 83.54% for intensity histogram, GLCM, and BoW model, respectively. What’s more ring-form partition can further improve the accuracies up to 87.54%,89.72%, and 91.28%.Another work of this paper is to complete the quantitative classification of cirrhosis of the liver, in other words, the quantitative classification include two point 1) obtain the quantitative geometric parameters of cirrhosis,2) classify the stage of disease based on the severity of liver cirrhosis.Traditional pathological slices need to go through a series of processing steps, fixe the slice, dye the slice, scan slice and many other tedious steps in the whole process, which will lead to the systematic error in diagnosis. However, we still rely on to analyze of the doctor’s experience to judge the severity of the disease, which is based on the subjective judgment of the tissue structure of the pathological slices, inevitably leads to human error. What’s more they often usually use a single score index, which is difficult to realize the mutual authentication between different parameters. Quantitative pathological is mainly considering the structure characteristic of geometry numerical, this method is also dependent on the physician’s manual delineation and measurement, the complicated operation process always lead to miss diagnosis.In order to solve the above problems, in this paper we proposed the method named the classification and quantization of liver cirrhosis.1) We use the Second Harmonic Generati and the Two-Photon Excitation Fluorescence to get a stain-free, high resolution liver slice images, which can avoid the chemical contamination and dyeing errors in the traditional way of dyeing.2) Segment the image, the output of the segmentation image is used to assist the doctor’diagnosis, in which three image features of SHG/TPEF microscopy were defined as:vessel, collagen fiber and nodule.3) Then the geometric parameters of each pattern were measured automatically, which were used as reference for doctors, such as the area, diameter, thickness, percentage, and so on. Finally, quantitative parameters were statistically calculated for building up an evaluation index for sub-classification of liver cirrhosis.A serial procedures of multi-channel image processing, including self-adaptive threshold segmentation, distance-transform based feature extraction quantification, and morphological operations, were developed to discriminate different patterns of the features. Based on different recombination of these features, the principle cirrhotic patterns including nodule, septa, vessel and sinusoid were defined and segmented thereafter.In this paper, the method is validated on the animal model as well as the clinical samples. In the animal model,175 SD-rat were used to carry out the quantitative analysis 1 of liver cirrhosis. While in clinical samples, there were 105 samplesUsing the method which was proposed in this paper to quantify the characteristics of the liver cirrhosis, and establish the qCP-index index to classify the stage of the disease. The results were compared with and the pathological physician scoring which was scored by 3 pathological physician. The accuracy rate can reach 83.4%, which means the significance of the auxiliary clinical diagnosis.