Monitoring the nodule prognosis from lung x-ray data

In this study a system that automatically detects the prognosis of nodules from chest X-ray data is proposed. Lung nodules are commonly seen as an illness. Nodules can be precisely monitored with a more riskful to human health methods such as CT or more expensive methods such as MR. However, X-ray imaging is a low cost method as well as its widespread usage. Monitoring the nodules in short intervals by X-ray method, gives benefits in many aspects. X-ray data taken at two different times must be overlapped to perform the assessment. By using feature matching methods homography matrix is calculated and then image registration is performed. To assess the registration success, the distance between the identical points which are selected manually, are compared. In addition, a method based on closed contour information is proposed to measure the success. According to nodule position, matched nodules are determined after image registration. Unmatched nodules on the first image are evaluated as lost while those on the second image are evaluated as newly appeared. Nodules are considered as closed contours consisting of pixel set where closed contour area is calculated after nodule matching process. In this way growth and shrink states are determined numerically. As a result, satisfactory results are obtained, both for the image registration and the process of nodule matching and prognosis.