Dual Energy Chest Radiography CAD
DESCRIPTION (provided by applicant):
The long-term objective of our proposed research is to improve the early detection of lung cancer with chest radiography. Because chest radiography is used for such a wide variety of medical conditions, it is, by far, the most widely used diagnostic imaging examination with over 30 million exams per year. Radiologists scan these images for all indications of disease, including lung cancer nodules, not just for the initial reason for the examination. Unfortunately, the fraction of nodules missed is quite high, over 30% in many studies. Our specific aim is to develop a computer-assisted detection (CAD) system to reduce the miss rate. The system uses low-noise, dual energy subtraction images, which have significant advantages for CAD yet there has been little research to exploit their unique characteristics. This is the opportunity for technological innovation addressed by our research. Dual energy provides images that eliminate ribs, a major contributor to errors in conventional CAD. It also provides images that may be used to measure nodule calcification, an important factor in malignant vs. benign diagnosis. Recently introduced digital, flat-panel x-ray systems provide dual energy images with lower noise than previous approaches. These lower noise images may improve detection of smaller, early-stage cancers. In Phase I, we showed the feasibility of our approach by developing new methods to utilize dual energy information in CAD. We developed a method for detection of potential nodules with high sensitivity and much lower extraneous response. We also developed a method for locating the lung fields in a chest image with higher accuracy and sensitivity than previous methods. Another innovation was a method of characterizing nodules based on a statistical technique called eigenimages that derives information directly from nodule images and whose accuracy increases as more nodule images become available. In Phase II, we will build on these methods to develop algorithms for the other components of the CAD system. We will study methods to incorporate these algorithms in a network application prototype to provide CAD services to multiple digital x-ray systems connected to a modern hospital information network. We will use the prototype to conduct a small-scale observer study to compare the performance of radiologists using our CAD system with unaided radiologists.
Small Business Information at Submission:
APREND TECHNOLOGY 2369 LAURA LN MOUNTAIN VIEW, CA 94043
Number of Employees: