Workshop Organizer: Jasjit S. Suri, PhD. Idaho State University, Idaho, USA

Speakers:
Raj M Rangayyan, PhD., University of Calgary, Calgary, Alberta, CANADA
Ruey-Feng Chang, PhD., National Chung Cheng University, Chiayi, Taiwan
Nico Lanconelli, PhD., University of Bologna, Bologna, Italy
Matteo Masotti, PhD., University of Bologna, Bologna, Italy
Matteo Roffilli, MS,University of Bologna, Bologna, Italy
Renato Campanini, PhD., University of Bologna, Bologna, Italy

Abstract

Mammograms and sonograms are difficult images to interpret, especially in the screening and diagnosing context. Objective methods for the analysis of mammographic and sonographic features are needed for the development of computer-aided diagnosis (CAD) methods to assist radiologists in the evaluation of ambiguous features.  But that alone is not enough. The integration of multiple modalities to detect early breast cancer, such as the fusion FFDM with Ultrasonography resulting in FFDMUS could also lead to improved diagnosis.

This workshop is organized by leading researchers in the field of mammography imaging, breast ultrasound imaging, digital image processing and computer vision. The workshop will be in three parts: Part-I will cover integrated FFDM and ultrasound Imaging, fast CAD and micro-calcification enhancement strategies, skin-line estimation techniques and multi-modality breast registration algorithms. Part-II will cover the latest CAD techniques for early breast cancer detection including methods for detection of bilateral asymmetry, architectural distortion. These CAD tools will be based on the application of digital image processing, computer vision, pattern recognition, content-based retrieval, data mining, and indexed atlas techniques to analyze mammograms for CAD of breast cancer. Part-III will cover the latest CAD techniques for 2-D and 3-D breast ultrasound. The texture-based and shape-based features are discussed for 2-D breast ultrasound. Also, the segmentation methods such as snake and level set are discussed for the breast ultrasound tumors. Classification techniques for breast US CAD, the recently developed 3-D ultrasound and its CAD will be also discussed.

The first part of the workshop will include discussions on the importance of fusion imaging of ultrasonography and FFDM. It will cover the state-of-the-art method in breast imaging. It will also present data analysis and comparisons of hand-held ultrasound with FFDMUS as imaging devices. The talk will mainly focus on:

FFDMUS Design based on Slot Scanning Methodology.
Fast Skin-line Extraction algorithms.
Fast Mirocalcification Extraction algorithms.
Application of Level Sets in Mammography.

The second part of the workshop will present an overview of several image processing techniques for the following applications:

Detection of masses and tumors
Shape analysis of tumors
Texture flow-field analysis of masses
Texture analysis of tumors
Analysis of bilateral asymmetry
Detection of architectural distortion
Pattern classification and computer-aided diagnosis.

The seminar will present general descriptions and examples of the techniques listed above. The latest works on the application of Gabor filters, phase portraits, and oriented texture analysis for the detection of architectural distortion will be described in detail.

The third part of the workshop will present an overview of several computer-aided diagnosis systems for 2-D/3-D breast ultrasound:

Texture-based computer-aided diagnosis for 2-D breast ultrasound
Shape-based computer-aided diagnosis for 2-D Breast ultrasound
Level-set tumor segmentation
Classification techniques for breast US CAD, such as neural network, self-organizing map, support vector machine, and decision tree.
Data acquisition and equipment for 3-D ultrasound
Texture-based computer-aided diagnosis for 3-D breast ultrasound
Speculation detection method for 3-D breast ultrasound

The last part of the workshop will focus on the use of advanced Pattern Recognition techniques, such as Support Vector Machines and Genetic Algorithms to a Computer Aided Detection (CAD) problem in FFDM. A brief theory preamble and an overview of the advantages of these techniques over other similar techniques will be given. Finally, a survey of the application of these methods to digital mammography CAD will be provided. The following points will be discussed:

Support Vector Machines classifier in FFDM CAD
Featureless approach to CAD with SVM
Genetic Algorithm for parameters optimization
Genetic Algorithms for feature selection