Improving the softcopy reading of mammograms using CAD

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): As in all other areas of radiology, mammography will become filmless. Full-field digital mammography (FFDM) is necessary for this to happen. There are additional advantages to wide clinical implementation of FFDM. Most important is the possibility of improving the accuracy of mammography through the development of advance techniques such as tomosynthesis and computer-aided diagnosis (CAD). One of the major roadblocks to FFDM is the reluctance of radiologists to employ softcopy reading, principally because the current generation of mammography workstations is cumbersome to use. We believe that softcopy will eventually allow radiologists to become more efficient and thereby more productive. The long-term goal of our research is to improve the utility of softcopy reading so as to facilitate the wide clinical use of full-field digital mammography. The goal of this project is to show that by using a computer-aided detection scheme, radiologists do not have to zoom the image to search for microcalcifications. Our hypothesis is that by employing CAD that radiologists will take less time to read an image on softcopy and that this can be done without a decrease in performance. To test this hypothesis, we will compare two reading conditions: (1) zoom without computer aid, and (2) no zooming with aid available. The specific aims are: 1. Collect 365 cases of mammograms: 275 normal, 70 with a malignant cluster of calcifications, and 20 with a malignant mass or distortion or asymmetry. 2. Modify an existing high-resolution workstation to mimic a clinical mammography workstation. 3. Conduct an observer study to test hypothesis. We believe that FFDM and related technologies have the potential to greatly improve the accuracy of mammography by reducing the miss rate and reducing the callback rate and the benign biopsy rate. Improving the softcopy display is one step to realizing this potential.