Designing a system for the diagnosis and classification of uterine cancer using a combination of graph cutting and color separation
,1,* Azam letafatfarashbandi
,2 Mohammad ghalandari
,3 Jasem jamali
1. Azad University of Kazeroun Branch
2. Azad University of Kazeroun Branch
3. Azad University of Kazeroun Branch
4. Azad University of Kazeroun Branch
Although auto-detection systems have improved, the histological examination of cervical biopsy by pathologists to detect cervical cancer maintains a clinical standard. however, due to the large volume of histologic data and the individuality of the histological examination of tissue samples and the lack of experience in evaluating samples, the results are significantly influenced by these factors. get it histologic images of the cervix, including the background, scaly tissue and structural tissue (stroma). an elemental pathologic component is a scaly tissue that contains important diagnostic information. scalable cladding tissue contains nuclei that display diagnostic information. the classification of cervical cancer is done using local and global methods for the segmentation and analysis of cervical tissue images. the global method defines the target area (roi), which is the same scalable texture. this method involves the fragmentation of the scaling tissue from the entire cervical histology image. the next process is a local method that is applied to a scaled tissue. the local method involves the segmentation and analysis of the cores and the contents of the scaffold cover. as a result, the nuclei are detected and the disease is graded. the overall purpose of this system is to propose a partitioning algorithm that identifies and separates the target area from the entire cervical histology picture to provide a computer-assisted diagnostic system for categorizing cervical cancer. the process in the proposed system includes a global approach, a local approach, and grading and classifying the disease. in the first step, the cervical histology image is processed using a global method to partition the cervical fissure. secondly, the cores are parsed and analyzed. finally, the image is graded and categorized. this process uses a combination of graph cutting and color separation.
This paper presents a computer aided diagnostic support system that helps pathologists in inspecting cervical biopsies. this paper investigates various components of an efficient computer aided diagnostic support system: image acquiring, preprocessing, segmentation, feature extraction, classification, disease grading and identification. the main goal of the proposed system is abnormality identification and determination of cancer grading in a systematic and repetition able situation. using combination of graph cut and color segmentation in computer aided diagnostic support system for cervical cancer classification is a novel research that uses sliding block algorithm for analyzing of nuclei. block moves in horizontal and vertical direction in order to cover squamous epithelium, and analyses the existence of nuclei wit good details. as a result, this method provides better performance as compared to k-means clustering and gabor wavelet in terms of specificity and false positive.
It can be seen that the method of cutting the graph, based on the quantitative measurements described in the next paragraph, makes a fine particle cladding texture. in order to measure the performance of the segmentation of the graphic cut, visual evaluation and quantitative measurements are used to evaluate the accuracy of the process. in figures, it can be seen that in comparison with hand-parting, the graph cutting method can partition the texture of the flute. in order to evaluate the partitioning process, hand segmentation has been used as a reference criterion. in the tables shows the results of segmentation for 475 cervical histology images in the form of a chart, which indicates the amount of error occurred in the segmentation of the graph cutting method.
In this paper, an automatic diagnostic system was designed and introduced by introducing the method of segmentation of the graphic section and the method of color classification and combining these two methods together. the function of this system was investigated on various images of cervical histology. it was found that this method is useful in the diagnosis and evaluation of cervical cancer and can help pathologists.
Cervical cancer classification, computer-assisted diagnostic support, k-means clustering and graph a