The programming of the project using MATLAB still needs to be improved as it produces the output that did not meet the author's expectation, especially in feature extraction. First of all thanks to Allah Almighty for giving me strength, wisdom and patience to be able to fulfill the entire project requirements. Last but not least, the author would like to express special thanks to the family for their understanding, invaluable support and constant love during the completion of the project.
As is known, breast cancer is one of the leading causes of female mortality in the world. Breast cancer is cancer that forms in tissues of the breast, usually the ducts (tubes that carry milk to the nipple) and lobes (glands that make milk). It occurs in both men and women, although breast cancer is rare in men. Primary prevention seems impossible because the causes of this disease are still unknown.
Mannograms do not prevent breast cancer, but they can save lives by finding breast cancer as early as possible. One of the important early breast symptoms on mammography is the appearance of a cluster of microcalcifications, which has a higher X-ray attenuation than normal breast tissue and appears as a cluster of small bright granular spots localized on mammography as Figure 1 below. Most breast specialists are encouraged by recent progress in CAD and look forward to more technical improvements and studies that help clarify its role in breast cancer detection.
PROBLEM STATEMENT
Masses can be caused by many things, including cysts (noncancerous, fluid-filled sacs) and noncancerous solid tumors (such as fibroadenomas), but they can be cancerous and should usually be biopsied if they are not cysts. Computers can help doctors identify abnormal areas on a mammogram by acting as a second set of "eyes." For standard mammograms, the film is fed into a machine that converts the image into a digital signal, which is then analyzed by the computer. The computer then displays the image on a video screen, with markers pointing to areas it "thinks" the radiologist should check particularly closely [2].
Early tests have found that CAD can help find some types of cancer that doctors might have otherwise missed. Some doctors feel that the device is not as effective as simply asking a second radiologist to review the fihn. Others are concerned that the device could lead to unnecessary biopsies by misidentifying benign abnormalities as suspicious for cancer.
Finally, dense tissue can easily be misinterpreted as calcification, causing high false positive (FP) rate.
OBJECTIVES
SCOPE OF STUDY
MICROCALCIFICATION DETECTION TECHNIQUES
IMAGE PROCESSING
Digital Image Formation
Image acquisition is performed to generate digital image from recorded data which includes optical system and sensor. To convert analog signal to digital form, we need to sample the function in both coordinates and amplitude. Each digital imaging subsystem introduces deformation or degradation to the digital image, such as geometric distortion, noise, and nonlinear transformation.
To identify and quantify structures in the image that may be indicative of the objects in the scene being viewed. To transform the image into an alternative representation where some operations can be performed more efficiently.
Digital Image Enhancement
Image Analysis
The objective of segmentation is to divide an image into meaningful regions with respect to objects within the scene. Gray level thresholding is the simplest segmentation process and is computationally inexpensive and fast. Pixel-based or local method - Detect and enhance edge or edge elements within an image.
Region-based or global approach - Create a region directly by grouping pixels with common characteristics into uniform areas. Edge detection dramatically removes the amount of information in the image because all non-structural data is removed, leaving all information about the edges. Edge information in an image is found by looking at the relationship a pixel has with its neighbor.
If a pixel has a neighbor with a very different gray level, it may represent an edge point. If a pixel's gray level value is similar to that of its surrounding neighbor, there is probably no edge at the point. Statistical pattern recognition assumes that the image may contain one or more objects and that the image may contain one or more objects that each object belongs to one of several predefined types, categories, or classes.
Pattern recognition systems typically consider a feature space onto which the observation vector is first mapped. The feature vector is then used to determine the class to which the observation vector belongs based on the measured objects.
FEATURE EXTRACTION
Depending on morphological criteria for the mass, similar hood of malignancy can be established. The general rule is that the larger, round or oval calcifications uniform in size are more likely to be associated with a benign process. They conclude that the combination of these three measures is better than just using one or two.
However, according to Woods [5], the characterization of benign and malignant lesions represents a very complex problem due to the small size of renal calcification. Number of calcifications that made up a cluster was used as an indicator of benignity and malignancy. While the actual number itself is arbitrary, radiologists tend to agree that the minimum number of calcifications is four, five, or six of significance.
A number of calcifications of less than four by itself rarely leads to the detection of breast cancer. Most radiologists consider calcifications 0.5 mm or smaller to have a high probability of being associated with cancer; and calcification of 2.0 mm or more are characteristic of a benign process. Perimeter measurement to measure the perimeter of an object to determine whether the boundary of the object is a polygon with a vertex at the center of each boundary pixel.
You can also measure the range by centering the distance between adjacent pixels on the border. To calculate the total volume, the total area of the bwperim is divided by the number of calcifications. Eccentricity is the ratio between the distance between the foci of an ellipse and the length of its major axis.
It is the portion of pixels in the convex hull that are also in the region. Mathematically, the area of a circle is calculated as pi *r2, while the circumference is calculated as 2 *pi *r2. • By calculating the equivDiameter, the diameter of an object can be obtained.
CHAPTER3
PROJECT DEVELOPMENT
The median filter, on the other hand, replaces a pixel's value with the median of the gray level near the pixel, and is better at reducing random noise without reducing the sharpness of the image. However, the sharp transition in gray level also consists of edges, which are the beneficial features of an image, but averaging filters have the unwanted side effects that can blur edges. The Top Hat transformation is defined by the difference between an image and its open version.
Its strength lies in its ability to enhance detail in an image that would otherwise be provided by shadows. The Top Hat masking filter with a disc-shaped structural element is applied to the image to remove uneven background illumination, and since the result of the operation is dark, the author applies stretchlim which calculates Morphology is defined as "the study of the shape of the thing"; which is the object segmentation structure.
The morphological technique often used to improve the appearance of an image is thresholding and is used to understand the structure or shape of an image by identifying objects or boundaries within an image. In this step, expansion and erosion operations are performed. These two operations are used to eliminate small object features such as noise noises and broken edges.
THE TEST
- DETECTED CLUSTER
- MAMMOGRAM DATABASE
- RESULT AND DISCUSSION
- CONCLUSION
- RECOMMENDATION
An image analysis can be performed to conclude whether the case is malignant or benign. For circumference measurement, most radiologists consider calcifications 0.5 mm or less to have a high probability of association with cancer; and calcification of 2.0 mm or greater are typical of benign tumor. Some of the results did not meet the author's expectations, which means that the object did not show the real result.
Microcalcification is usually concentrated or distributed in a specific area of the breast, so it is advantageous for the user to crop the image to a specific area suspected of containing microcalcification. File Edlt IJiew Insert Too~ Desl The median filter is effective in removing the impulse noise that usually occurs during image digitization. The resulting image is sharpened with the features in the image increased in contrast. This filter ensures that edges and fine details in the image become sharper. This approach is called high-boost filtering. After using different image neighborhood types and values, 12-neighborhood disk types are found to be suitable. The image is enhanced by contrast stretching, which adjusts the histogram of the image so that there is greater separation between the grayscale distribution of the foreground and background. Mammography has proven to be one of the most reliable methods for early detection of breast cancer. Computers can help doctors identify abnormal areas on a mammogram by acting as a second set of "eyes." For standard mammograms, the film is fed into a machine, which converts the image into a digital signal that is then analyzed by a computer. For this project the report consists of two main steps which are research, preprocessing and segmentation and feature extraction Due to the author there is only one semester to complete the research and come up with the product so one part has not met the objective. To improve image feature extraction, apart from using shape and area measurement, one can use different techniques like wavelet transform or fuzzy logic to have more accurate value. With this implementation we can have the image that is needed and it is easy when it comes to the programming part. MIAS LICENSE AGREEMENT
APPENDICES
