> Research & Grants
> Grants Program
> Research Grants
> Research Grants Awarded
Research Grants Awarded
Improving Mammography Performance with Tomosynthesis
Detection, Diagnosis and Prognosis
BACKGROUND: Tomosynthesis is a well known approach to image acquisition that enables partial 3-D reconstruction of multiple levels (slices) of overlying tissue from a series of images acquired at different angles. Tomosynthesis has recently been developed as an add on to Full Field Digital Mammography (FFDM). Ultimately, tomosynthesis can be relatively easily and inexpensively implemented on FFDM systems and it enables conventional and “tomo” images to be acquired during screening examinations. Detection of breast cancer in younger women and those with dense breasts is known to be lower, with some studies claiming as low as 35% sensitivity in this group. This reduction in detection sensitivity is particularly notable in the case of masses. In high risk women, magnetic resonance imaging (MRI) has been recommended as a possible screening tool but it is impractical to use as a general screening modality because of the cost, complexity, and utilization of resources. Women with dense breasts are also being recalled more often for additional procedures. OBJECTIVE/HYPOTHESIS: Our primary hypothesis is that the use of tomosynthesis will improve detection tasks as measured by the area under ROC curves by at least 5 percent, or reduce recall rates by at least 20 percent, or both. Our secondary hypothesis is that classification / diagnostic tasks will improve by at least 5 percent. SPECIFIC AIMS: We will perform a two mode, multi-reader (8), multi-case (approximately 220), ROC type observer performance study to evaluate and compare performance “with” and “without” the availability of tomosynthesis. STUDY DESIGN: An ROC type study that includes both detection and characterization tasks with the conventional viewing and the tomosynthesis generated approach will be performed and analyzed. The results of the different modes will be compared. We have access to the needed set of cases, the advanced display environment and a group of highly experienced and highly trained radiologists who will participate as observers. POTENTIAL OUTCOMES AND BENEFITS OF THE RESEARCH: The use of tomosynthesis could substantially reduce recall rates in women with dense breasts by as much as 30 percent. In addition, if the use of tomosynthesis acquired (and processed) images could increase sensitivity in this group, which constitutes a large fraction of women being screened (as much as 40 percent), it would be a significant step in improving early detection of breast cancer.
BACKGROUND: Detection of breast cancer in younger women and those with dense breasts is known to be lower for masses, with some studies claiming as low as 35% sensitivity in this group. In addition, women with dense breasts are recalled more often for additional procedures. In high risk women, magnetic resonance imaging (MRI) is being investigated as a possible screening tool but it is impractical as a general screening modality because of the cost, complexity, and utilization of resources. A new approach to mammography called tomosynthesis has been developed in recent years. This approach takes multiple views (as many as 11) of the breast and it enables the reconstruction of the breast tissue in a manner that it can be viewed as regular mammograms as well as in a three dimensional form that enable the radiologists to look through multiple levels (slices) of tissue. These levels are as thin as one millimeter. The advantage of this technology is that it can be implemented on Full Field Digital Mammography (FFDM) equipment with a relatively small incremental cost and the procedure of acquiring the images is largely the same as a regular mammogram. Hence, there is no need for an additional procedure. OBJECTIVE/HYPOTHESIS: The hypotheses of this project are that tomosynthesis generated images will reduce the recall rate in women with dense breasts by at least 20 percent and increase sensitivity in detecting breast cancer in the same group of women by 5 percent. In addition, classification / diagnostic tasks can improve by at least 5 percent. SPECIFIC AIMS: Our main study aim is to evaluate and compare performance “with” and “without” the availability of tomosynthesis in an observer performance study with a group of qualified radiologists as readers. STUDY DESIGN: An ROC type observer study that includes both detection (“normal”/“abnormal”) and characterization (benign /cancer) tasks with the conventional viewing approach and the tomosynthesis generated approach will be performed and analyzed. The results of the different reading modes will be evaluated and compared. POTENTIAL OUTCOMES AND BENEFITS OF THE RESEARCH: The use of this relatively inexpensive, easy to implement new technology may prove to be a major step in improving early detection of breast cancer (and /or reducing recall rates) in women with dense breasts.