فهرست مطالب سیده شهربانو فلاحیه حمیدپور

Breast cancer is one of the most common causes of death among women around the world, but early and accurate diagnosis of this type of cancer can dramatically improve treatment. Thermal imaging is one of the primary methods of diagnosing breast cancer. The computer diagnosis system can also be used to help physicians to increase the accuracy of interpretation of results. This paper presents an intelligent computer diagnostic system for the detection of breast cancer using thermal imaging. The proposed intelligent computer diagnosis system includes SFTA method for feature extraction and SVM, kNN and D-Tree algorithms for classification of results. The performance of the proposed intelligent computer diagnosis system is evaluated using the DMR-IR and Fluminense Federal University databases and MATLAB2018, when using the cuckoo feature selection algorithm and without using the feature selection algorithm. The results show that the average accuracy, sensitivity and specificity are 99%, 99.5% and 98.03%, respectively, using the cuckoo feature selection algorithm and SVM classification algorithm. Also, the presented computer diagnostic system has advantages compared to other computer diagnosis systems. These results indicate that the use of SFTA feature extraction method, cuckoo feature selection algorithm, SVM classification algorithm and DMR-IR database in the proposed computer diagnosis system can improve the evaluation results.

Nowadays virtual colonoscopy has become a reliable and efficient method of detecting primary stages of colon cancer such as polyp detection. One of the most important and crucial stages of virtual colonoscopy is colon segmentation because an incorrect segmentation may lead to a misdiagnosis. In this work, a hybrid method based on Geometric Deformable Models (GDM) in combination with an advanced region growing and thresholding methods is proposed. GDM are found to be an attractive tool for structural based image segmentation particularly for extracting the objects with complicated topology. There are two main parameters influencing the overall performance of GDM algorithm; the distance between the initial contour and the actual object’s contours and secondly the stopping term which controls the deformation. To overcome these limitations, a two stage hybrid based segmentation method is suggested to extract the rough but precise initial contours at the first stage of the segmentation. The extracted boundaries are smoothed and improved using a modified GDM algorithm by improving the stopping terms of the algorithm based on the gradient value of image voxels. The proposed algorithm was implemented on forty data sets each containing 400-480 slices. The results show an improvement in the accuracy and smoothness of the extracted boundaries. The improvement obtained for the accuracy of segmentation is about 6% in comparison to the one achieved by the methods based on thresholding and region growing only.Discussion and The extracted contours using modified GDM are smoother and finer. The improvement achieved in this work on the performance of stopping function of GDM model together with applying two stage segmentation of boundaries have resulted in a great improvement on the computational efficiency of GDM algorithm while making smoother and finer colon borders.