Lung cancer is one of the leading causes of death in the United States, surpassing breast, prostate, colon, and cervical cancers combined. One of the keys to improving the prognosis of lung cancer is early detection of small solitary lung nodules in chest radiographs, a task highly limited by the presence of anatomical variations in the image and by perceptual visualization processes. This project proposes a new method, Bi-plane Correlation Imaging (BC1), for improved detection of subtle lung nodules. The method involves the utilization of angular information in conjunction with digital acquisition and computer-assisted diagnosis (CAD) to cost-effectively reduce the degrading influence of anatomical variations with little or no increase in the patient dose. In BCI, two digital images of the chest are acquired within a short time interval from slightly different posterior projections. The image data are incorporated into an enhanced CAD algorithm in which nodules present in the thoracic cavity are detected by examining the geometrical correlation of the detected signals in the two views. Angular information minimizes the undesirable influence of anatomical noise by identifying and positively reinforcing the nodule signals, while CAD provides a complete search of the image data. The expected high sensitivity/specificity of the method has the potential to change the current state of practice, perhaps leading to a preventive lung cancer screening program for high-risk populations. The project focuses on implementation and clinical evaluation of BCI. Initially, steps will be taken to streamline the optimum acquisition and processing methods for BCI using phantom images. A dedicated imaging system will then be developed capable of high-speed bi-plane imaging of the chest. The system will be used to acquire bi-plane paired digital radiographs from 150 high-risk human subjects with confirmed lung nodules and 75 normals. The data will be processed according to an optimized BCI scheme based on the spatial correlation of CAD-identified lesions in the two views. A graphical user interface will be developed for the diagnostic evaluation of the method. A method and a graphical user interface will also be developed for stereoscopic presentation of bi-plane images. The comparative performance of single image and dual image detection of lung nodules in the collected database with and without the BCI and stereoscopic display will be evaluated. The findings will be used to substantiate the capability of BCI for improving the early detection of lung nodules and to optimize its performance on actual clinical cases.
