Enhancing ultrasound images for better interpretation of breast cancer

This work was carried out with the aim to help radiologists in ultrasonography in diagnosing breast diseases. In particular, the study was done to improve their efficiency in interpreting results. As the work concerns with identifying breast disease, the focus is on BMode ultrasound that is suitable for breast. Thus, to improve the interpretation of results, a multi-purpose multi-tissue breast phantom is used to capture all the combinations of possible functions in the B-Mode. In the first step, ultrasound images were captured using a Quality multi tissue equivalent Breast Phantom that contains all the needed specifications. All the possibilities for grey scale images were used on some ultrasound machines. In order to check the quality of image, a variety of transducers were used with different frequency responses and the same frequency setting. This work focused on two lesion types, namely, Cyst and Solid, besides other specifications of Phantom. A review of previous research on breast disease has shown that most of them have been done on modalities other than ultrasound,specifically on mammography and digital mammography. Although some work has been done on ultrasound, these are rather limited to very special cases in kidney and abdomen. In order to enhance ultrasound breast image, image factors such as Signal to Noise Ratio (SNR) and Receiver Operating Characteristic (ROC) analysis like sensitivity, specificity, and accuracy were used. Meanwhile, classification of the masses was done based on the features that were extracted from two types of images, namely,phantom images and real human breast images. Then, the proposed designed was applied to all the images and the output data (benign or chance of malignancy) were gathered. After classifying the masses, a set of samples was selected to make the required tests for the current work, such as the ROC analysis. As breast disease is one of the deadliest reasons for death among women in most societies, the aim of the present work was on aiding the radiologists in checking and detecting some of these abnormalities. Findings showed an improved sensitivity of 99% and an enhanced accuracy of 98% for Ultrasound Phantom images. Evaluation of results for breast Ultrasound images also gave 98.5% for sensitivity and 98.2% for accuracy. So a powerful and reliable Computer Assisted Detection framework was introduced. In addition to these results, a full machine performance evaluation was done based on the findings of the proposed routines. Finally, regarding to this thesis findings, Ultrasound can be a good screening modality as the first row image modality in breast imaging.

Preview PDF FPSK(p) 2012 21R.pdf Download (871kB) | Preview

Actions (login required)

View Item