Figure 1. High-resolution CT scan images displayed with advanced 3D image processing techniques show a small (approximately 4 mm) pulmonary nodule. The two scans were obtained 189 days apart and the growth rates are automatically determined.

The Early Lung Cancer Action Project (ELCAP) has been performing research on computed tomography (CT) screening for lung cancer for the past 10 years. During this time, there have been tremendous technological advances in CT scanners. When screening first began, each CT scan image was 10 mm thick in order to allow a scan of the entire chest in a single breath-hold, and all images were printed out on film. Now, a chest scan on a single breath-hold can be done with slices only 1 mm thick, and all images are stored digitally for display and processing. The latest CT scanners can take 300 images in a breath-hold, compared with 30 for older machines.

These technological advances, especially the digital image acquisition, allow the use of advanced computer image processing techniques. The clinical group at the Weill Cornell campus of NewYork-Presbyterian Hospital and the computer group at the Cornell University College of Engineering have collaborated to lead the development of these image processing techniques. The first of the techniques developed at the Weill Cornell campus involves the accurate measurement of pulmonary nodules observed on high-resolution CT scans. The image processing creates three-dimensional reconstructions so that volumetric determinations can be made; with these accurate volumes measurements, we now can compare nodules on scans obtained at different times to determine accurate growth rates (Fig. 1). This approach is particularly well suited to the evaluation of small nodules observed during CT screens, and the whole team worked to convert it into a product in the General Electric

Figure 2. A scout view from a CT scan (A) shows where nodule candidates are automatically selected for review. They are outlined with the squares. The CT image (B) shows one of the selected candidates presented to the radiologist for review.

Figure 3. Advanced image processing shows the extent of emphysema. Normal lung tissue appears as pink and emphysema appears black. Patient (A) has no emphysema, while patient (B) has advanced emphysema prominent in the upper lobes.

Advanced Lung Application Software. The collaboration with General Electric is extending to the development of new software that actually can detect nodules on a CT scan and act as a second read for radiologists (Fig. 2). This application, which is undergoing FDA review, could greatly facilitate radiologic interpretations, as the number of images that each radiologist must review continues to increase.

Several other techniques for computer-aided diagnostics are being developed. For example, now automated quantification of emphysema on CT scan images can be done (Fig. 3). This information has been found to be particularly helpful in providing patients an incentive to quit smoking; screening patients have been strongly influenced to quit when shown an image of the extent of emphysema in their lungs. The groups at the Weill Cornell and Columbia campuses have been investigating this added benefit of CT scanning. Another application of computer-aided techniques being investigated here is the processing of chest x-rays to help identify small cancers with these images as well.

Editor’s Note (About the Authors):

Claudia Henschke, MD, PhD, is Professor of Radiology at the Weill Medical College of Cornell University, and Chief of the Chest Imaging Division and an Attending Radiologist at NewYork-Presbyterian Hospital.

David Yankelevitz, MD, is Professor of Radiology at the Weill Medical College of Cornell University and Attending Radiologist at NewYork-Presbyterian Hospital.

James P. Smith, MD, is Clinical Professor of Medicine at the Weill Medical College of Cornell University and Attending Physician in Pulmonary and Critical Care at NewYork-Presbyterian Hospital.

Daniel Libby, MD, is Clinical Professor of Medicine at the Weill Medical College of Cornell University and Attending Physician at NewYork-Presbyterian Hospital.

Mark Pasmantier, MD, is Clinical Professor of Medicine in the Medical Oncology/Solid Tumor Program at Weill Medical College of Cornell University.

William J. Kostis, PhD, is Assistant Professor of Electrical and Computer Engineering in Radiology at Weill Medical College of Cornell University.

Anthony Reeves, PhD, is Associate Professor of Electrical Engineering at the School of Electrical and Computer Engineering at the Cornell University College of Engineering.