| As the incidence rate of tumor increased year by year,the number of primary brain tumors and brain metastases is also growing.Compared to tumors from other parts of body,the effect of brain tumors on the quality of life and survival time is more serious.Now in the area of diagnosis and treatment of tumor,the top problem is to find the tumors early,which will make the best period of treatment gone.There for,early detection of brain tumors can be the key to prolong the survival time and improve the quality of life.From now on,the best way to show brain tumors is Magnetic Resonance Imaging(MRI).With the medical expenses continuous reducing and the national economic conditions continues improving,the huge amounts of MRI brain images become a burden for clinical doctors.Radiologists read large amounts of images by eyes for long time which cause misjudgment or false negatives by visual fatigue or experience of other human factors.However,with the continuous developing of computer image process and analysis technology,computer-aided detection(CAD)technology become an effective way to solve this problem.CAD technology has been applied in many fields,such as satellite image recognition,plate recognition,street signs,autopilot,machine vision,image retrieval and so on.Moreover,with 3D image processing technology matures,the requirement of efficient,accuracy,and false positive of CAD for detection algorithm increased.The research of CAD based on 3D images becomes hotspot.In medicine,lesion detection is especially difficulty in the study.According to different method of detection,template matching method and classifier can be concluded.After decades of efforts,domestic and foreign researchers have developed many detection methods for specific problems,but obviously,these methods still have a long way to go from mature in medical field.Some research achievements in the field of brain tumors CAD system from recent years are summarized in this paper,including the parenchymal extraction method,region of interested extraction methods,detection method based on template matching and classifier method.3D adaptive template matching method was further studied in this paper.A complete feasible optimal solution for the parenchymal extraction with tumors was also mentioned.This research focus on the 3D adaptive template matching method for brain tumor detection.Follows are the main contributions.1.In the view of the brain extraction algorithm with tumors,the improved BET algorithm was mentioned,and the extraction effect of original algorithm was improved.An effective method which can be used to reduce the amount of calculation of brain extraction while not affect the extraction effect was also put forward.More importantly,3D information of the image was made full use.2.The extraction algorithm of region of interested adopted can more efficiently extract brain with region of interested as much as possible.Based on which,the template building algorithm and detection algorithm was mentioned.The sensitivity of the region of interested algorithm for tumor is 99.38%.3.The 3D adaptive templates ware built in this paper,and faster matching algorithm was adopted to avoid the disadvantages of search volume data from usual matching algorithm.The ROC curve was used to analyze the performance of detection algorithm in this paper.Experimental results show that the sensitivity rate of the method in this paper was 88.7097%,while the false positive was 16.03%.Compared with the template matching method reported in recent years,the detection performance of the method has obvious improvement.The algorithm is faster than which reported in recent years by nearly 60% to 83%. |
PDF Full Text Request