The present study aimed to investigate the design of a computer-assisted pathology system for diagnosis and clustering of
cancerous lesions in magnetic resonance imaging of breast, using computer code in MATLAB software. In the analysis of
breast segmentation by Atlas method, mass tumors 4 and non-mass tumors 5 are identified and segmented. Characteristics of
the morphology, kinetics and matrix of the gray level co- occurrence of the tumors are extracted. In this study, a new feature
called “dual-tree complex wavelet transform (DTCWT” was extracted and five characteristics associated with this type of
property were extracted. After extracting these properties, the feature vectors were assigned to the clustering with different
kernels and the combined clustering, which combine the linear discriminate analysis method and the nearest neighbor, and
clustering of the tumors was performed into two benign and malignant categories. Using the new feature introduced in this
study and applying it to the SVM cluster, AZ values for mass tumors, non-mass tumors and their combination were 0.71, 0.77
and 0.70, respectively, and by applying it to the combined cluster s LDA and NN-k were 0.70, 0.44 and 0.69, respectively.
Also, in the Atlas-based segmentation, the FCM cluster was used for them first time. The use of this cluster caused that
there is no empty cluster and the accuracy of the results would increase. In the feature extraction section, the feature of
dual-tree complex wavelet transform (CWT-DT) was applied for the first time in magnetic resonance images of the breast
and on mass and non-mass tumors and a combination of them was applied. Detection and extraction of non-mass tumors is
the main challenge of this study, and applying the proposed feature group of non-mass tumors created an acceptable result,
and the value of AZ increased compared to previous studies.
Detector of 3-D salient points based on the dual-tree complex wavelet transform for the positioning of hippocampi meshes in magnetic resonance images
