Mario Mustra

A Survey of Image Processing Algorithms in Digital Mammography

Mammography is at present the best available technique for early detection of breast cancer. The most common breast abnormalities that may indicate breast cancer are masses and calcifications. In some cases, subtle signs that can also lead to a breast cancer diagnosis, such as architectural distortion and bilateral asymmetry, are present. Breast abnormalities are defined with wide range of features and may be easily missed or misinterpreted by radiologists while reading large amount of mammographic images provided in screening programs. To help radiologists provide an accurate diagnosis, a computer-aided detection (CADe) and computer-aided diagnosis (CADx) algorithms are being developed. CADe and CADx algorithms help reducing the number of false positives and they assist radiologists in deciding between follow up and biopsy. This chapter gives a survey of image processing algorithms that have been developed for detection of masses and calcifications. An overview of algorithms in each step (segmentation step, feature extraction step, feature selection step, classification step) of the mass detection algorithms is given. Wavelet detection methods and other recently proposed methods for calcification detection are presented. An overview of contrast enhancement and noise equalization methods is given as well as an overview of calcification classification algorithms.

Robust automatic breast and pectoral muscle segmentation from scanned mammograms

Breast skin–air interface and pectoral muscle segmentation are usually first steps in all CAD applications on scanned as well as digital mammograms. Breast skin–air interface segmentation is much more difficult task when performed on scanned mammograms than on digital mammograms. In case of pectoral muscle segmentation, segmentation difficulty of analog and digital mammograms is usually similar. In this paper we present adaptive contrast enhancement method for breast skin–air interface detection which combines usage of adaptive histogram equalization method on small region of interest which contains actual edge and edge detection operators. Pectoral muscle detection method uses combination of contrast enhancement using adaptive histogram equalization and polynomial curvature estimation on selected region of interest. This method makes segmentation of very low contrast pectoral muscle areas possible because of estimation used to segment areas which have lower contrast difference than detection threshold.

Breast border extraction and pectoral muscle detection using wavelet decomposition

Digital mammography is used more and more each day in comparison with screen film mammography (SFM). Main advantage of digital mammography for image processing is the use of images with few or no artifacts that can occur on SFM images. Finding breast border contour is therefore easier and gives more precise results. On the other hand, detection of pectoral muscle and breast abnormalities has almost the same results in both cases. The presence of pectoral muscle can affect results of lesion detection algorithms so it is recommended to have it removed from the image. Detection and segmentation of pectoral muscle can also help in image registration for further analysis of breast abnormalities such as bilateral asymmetry. Algorithm presented in this paper uses hybrid method for the pectoral muscle detection. Proposed method uses bit depth reduction and wavelet decomposition for finding pectoral muscle border. Algorithm has been tested on the set of 40 digital mammography images.

Breast Density Classification Using Multiple Feature Selection

Mammography as an x-ray method usually gives good results for lower density breasts while higher breast tissue densities significantly reduce the overall detection sensitivity and can lead to false negative results. In automatic detection algorithms knowledge about breast density can be useful for setting an appropriate decision threshold in order to produce more accurate detection. Because the overall intensity of mammograms is not directly correlated with the breast density we have decided to observe breast density as a texture classification problem. In this paper we propose breast density classification using feature selection process for different classifiers based on grayscale features of first and second order. In feature selection process different selection methods were used and obtained results show the improvement on overall classification by choosing the appropriate method and classifier. The classification accuracy has been tested on the mini-MIAS database and KBD-FER digital mammography database with different number of categories for each database. Obtained accuracy stretches between 97.2 % and 76.4 % for different number of categories.

Review of recent advances in segmentation of the breast boundary and the pectoral muscle in mammograms

This paper presents a review of recent advances in the development of methods for segmentation of the breast boundary and the pectoral muscle in mammograms. Regardless of improvement of imaging technology, accurate segmentation of the breast boundary and detection of the pectoral muscle are still challenging tasks for image processing algorithms. In this paper, we discuss problems related to mammographic image preprocessing and accurate segmentation. We review specific methods that were commonly used in most of the techniques proposed for segmentation of mammograms and discuss their advantages and disadvantages. Comparative analysis of the methods reported on is made difficult by variations in the datasets and procedures of evaluation used by the authors. We attempt to overcome some of these limitations by trying to compare methods which used the same dataset and have some similarities in approaches to the breast boundary segmentation and detection of the pectoral muscle. In this paper, we will address the most often used methods for segmentation such as thresholding, morphology, region growing, active contours, and wavelet filtering. These methods, or their combinations, are the ones most used in the last decade by the majority of work published in this image processing domain.

Efficient presentation of DICOM mammography images using Matlab

DICOM has become a standard for medical imaging. Its purpose is to standardize digital medical imaging and data for easy access and sharing. There are many commercial viewers that support DICOM image format and can read metadata, but image displaying is not always optimal. Since mammography images are mostly 12-bit grayscale images, image viewer has to be adapted for correct displaying when different number of bits has to be allocated. Another problem occurs when only a small portion of amplitudes is of interest to display. In this paper a method for efficient presentation of DICOM mammography images in Matlab is introduced. Algorithm that is developed does automatic contrast and brightness adjustments, regardless of bit dept of an image, number of amplitudes higher than zero and the number of allocated bits.

Comparison of Calculated and Measured Travel Times with PND Prediction

Correct estimation of the travel time is a rather difficult task because it relies on many factors which act in an unpredictable way. To verify predicted travel time results obtained by calculation from historic data, we have tried to verify those results by measurement and comparison with industrial personal navigation devices available on the market. With the limited time and resources we were able to cover only a finite number of routes with different parameters. Comparison of results show that there are many factors which influence the travel time on short routes and results suffer from higher inaccuracy. On longer routes or on combination of more short routes results show much smaller deviation and are better comparable to the predicted ones.

Comparison of Dirac and H.264/AVC Coding Quality Using Objective Video Quality Measures

Digital video braodcasting requires high image quality, keeping the bitrate as low as possible at the same time. In the digital photography, new image compression standard named JPEG2000 proves to be more efficient than JPEG in subjective evaluation. JPEG2000 uses discrete wavelet transform (DWT) while JPEG uses discrete cosine transform (DCT). In video coding there are two most popular multi-purpose coding formats, MPEG-2 and H.264/AVC also known under the name MPEG-4 Part 10. Both of these formats use DCT. This paper present comparison of the coding quality of H.264/AVC and newly developed Dirac codec. Dirac codec unlike H.264/AVC uses DWT. For the quality comparison of two standards, PSNR, VQM and SSIM methods are used.

Point-based estimation of calculated speed profiles for road segments

Route planning for freight and personal navigation in urban areas is rather difficult because optimal route is not only distance-but also time-dependent. This task often depends on varying traffic conditions which can be influenced by different factors. Optimal route planning is therefore dependent on the time of the day and of the day in the week for different urban areas. One of the usual solutions to this problem is calculation of speed profiles from GPS traces gathered from different vehicles. Speed profiles are calculated estimates of average speed which can be achieved on a specific road segment during one day. To calculate speed profiles we used gathered GPS data collected during five years and 4,000 vehicles, mostly in urban areas in Croatia. There are different approaches considering time and distance averaging, and final results depend of the chosen values. Actual speed estimation can be made using radar measurements which are point-based but can give a good estimate for the chosen segment. In this paper we compare measurements obtained using Frequency Modulated Continuous Wave (FMCW) radars with calculated speed profiles and present difference for the four road segments.

Comparison of segmentation accuracy for different LUTs applied to digital mammograms

In this paper we try to compare usage of different image processing techniques applied to mammograms. Different imaging devices produce images with different properties, which can be determined by histogram comparison. Digital imaging devices store images according to the DICOM standard, but to view images properly, images need to be processed for displaying on a desired display. Linear grayscale transformation does not provide a good solution and therefore grayscale values are being converted using look-up-tables (LUTs). Once converted for optimal displaying, images suffer from a loss of details and sometimes information which could be useful for computer-aided detection (CAD) algorithms. In this paper we will compare segmentation accuracy of the breast tissue and nipple detection accuracy when using different image preprocessing techniques.