Susan G Komen  
I've Been Diagnosed With Breast Cancer Someone I Know Was Diagnosed Share Your Story Join Us And Stay Informed Donate To End Breast Cancer
    Home > Research & Grants > Grants Program > Research Grants > Research Grants Awarded > Abstract

    Research Grants Awarded

    Three-dimensional Computerized Detection and Diagnosis of Breast Cancer with Tomosynthesis

    Study Section:
    Detection, Diagnosis and Prognosis

    Scientific Abstract:
    Background: The accuracy of screening and diagnosis of breast cancer is mainly associated with the nature of imaging and the manner in which screening images get interpreted. Mammography has limitations due to its inherently projective imaging nature. Breast tomosynthesis is an emerging state-of-the-art 3D imaging technology that demonstrates significant early promise in diagnosing breast cancer. However, the large number of tomosynthetic images from each patient to be evaluated by radiologists could aggravate interpretation variability. Moreover, impacts of physical factors of the tomosynthesis system on the interpretation and diagnosis accuracy remain not addressed. Thus automatic interpretation and diagnosis of breast tomosynthesis is paramount for its clinical application. We propose to develop an innovative 3D computerized detection and diagnosis (CDD) system on breast tomosynthesis. Objective/Hypothesis: The purpose is to improve detection and diagnosis accuracy of breast cancer. Our hypothesis is that t he proposed 3D-CDD method will yield significantly superior performance in sensitivity and specificity over two-dimensional (2D) methods in the detection and diagnosis on tomosynthesis. Specific Aims: (1) To develop a 3D-CDD system by developing advanced image/signal analysis algorithms and contemporary pattern recognition techniques on breast tomosynthesis; (2) to evaluate the performance of the 3D-CDD system by statistical analysis. Study Design: The proposed study is composed of development and evaluation of 3D-CDD system, and investigation of impacts of tomosynthesis factors on 3D-CDD system. After suspicious objects extracted from the breast volume using a robust fuzzy logic segmentation algorithm, tomosynthetic feature extraction and selection will be performed in segmented objects, then a multi-stage classification procedure using advanced pattern recognition algorithms will be applied to classify cancerous and benign lesions, and normal breast tissues. The performance of the 3D-CDD system will be validated using statistical evaluation methods. Finally, impacts of key tomosynthesis factors on the 3D-CDD system will be investigated. Potential Outcomes and Benefits of the Research: This study will provide automatic breast cancer interpretation and diagnosis system that will especially benefit women with high breast density. This technique is expected to significantly reduce recall rates of screened women and unnecessary biopsy rate.

    Lay Abstract:
    Mammography has limitations in detecting and diagnosing breast cancer due to its inherently projection imaging nature, although currently it is the most effective method in screening and diagnosing breast cancer. Breast tomosynthesis is a state-of-the-art three-dimensional (3D) imaging technology that demonstrates significant early promise in improving diagnostic accuracy and detecting more subtle lesions over mammography. It acquires images at multiple angles during a short scan and reconstructs them into a series of thin high-resolution slice images for each beast at a total radiation dose approximately equal to that of a single view mammogram. However, this modality requires considerable extra work for the radiologist to evaluate and interpret a single patient. Thus, translation of this novel technology to the benefit of patients requires automatic synthesis and interpretation of a large number of tomosynthetic slices of the breast with the assistance of a computer. In this application we propose to develop an innovative 3D computerized detection and diagnosis (CDD) system to be applied on breast tomosynthesis for detecting and diagnosing breast cancer. To this aim we will develop and integrate advanced analysis methods of the breast tomosynthesis data as well as robust characterization algorithms of tomo-radiographic breast lesions and normal breast tissues, to precisely and automatically detect and diagnose malignant, benign lesions as well as normal breast tissues. Moreover, the impact of physical factors of the tomosynthesis system on the performance of the proposed 3D-CDD system will be investigated. The expected outcome of this study is an automated computerized system to facilitate the application of breast tomosynthesis technology for the early detection of breast cancer. Success in this endeavor is expected to significantly reduce both recall rates of screened women and unnecessary biopsy rates.