New lung product propels CAD into mainstream CT

Chest CTs are rich with more data than meet the eye. R2's ImageChecker CT Lung Version 2.0 CAD system, armed with its AutoPoint temporal comparison algorithm, highlights abnormalities, then compares new and past images to demonstrate changes that have occurred over time.

The goal is to improve accuracy and efficiency in the early identification of lung nodules, an objective partially achieved through the use of software to back up the radiologist.

"Experienced chest radiologists can miss a significant number of nodules under 1 cm in size, and some of those are very relevant," said Dr. David Mendelson, an associate professor of radiology at Mount Sinai School of Medicine in New York City. "So the issue is to use ImageChecker to find any nodules that you didn't and to quickly exclude false positives."

The diagnostic process advances further using AutoPoint, which compares and tracks a patient's lung nodule over time. The tool examines two data sets acquired at different times.

"It tells you how the nodule has changed in terms of volume, making comparisons over time fairly easy," Mendelson said.

This temporal tool, available as an option on Version 2.0, also provides a tabular report featuring key images and information on nodule volume and density changes. It predicts the time needed for the nodule to double in size, which is an important indicator of whether the nodule is malignant.

"Changes in volume are not strictly used to indicate whether the patient goes to surgery, but on occasion the availability of volumetric comparisons has helped my decision making," he said.

These and other CAD results obtained with Version 2.0 can be sent to a PACS using standard formats for enterprise-wide accessibility.

"The lung nodule CAD is definitely ready for prime time," he said. "This is a very useful product."

Of potentially greater clinical significance is the pulmonary artery patency tool, which helps in the diagnosis of pulmonary embolism by detecting obstructions in the pulmonary artery.

R2's PE tool isolates the relevant CT data and automatically calculates vessel diameter, percent occlusion, and size of the detected obstruction.

"Radiologists typically review hundreds of images for a multislice CT scan obtained for patients with suspected pulmonary embolism," said Dr. Michael B. Gotway, vice chair of radiology at the University of California, San Francisco. "With CAD, this process should become more accurate and efficient, thus improving patient outcomes."

Clinical data presented at the 2004 European Congress of Radiology indicate that R2's PE tool achieved a sensitivity of 88% in the detection of segmental pulmonary embolism and 78% for detection of subsegmental pulmonary emboli, with an average of four false CAD marks per normal case.

The problem with R2's PE tool, however, is a lack of specificity, Mendelson said.

"It finds PEs, but it also finds things that pretty clearly are not PEs," he said. "This is an issue that some people in our group are working through because they don't want to miss a PE, while other people say it is not worth the trouble of filtering out the false positives."

Both the lung nodule and PE tools are equipped with powerful and intuitive interfaces. Ease of use is a key consideration. The lung nodule assessment includes a lung map algorithm that displays CAD-detected nodules, indicated by a green circle on an axial view alongside an automatically generated 3D volume rendering.

"It helps orient you," he said.

R2's emphasis on clinical utility and operability distinguishes the company and its product, according to Mendelson.

"The company understands that the way you display information affects utilization, and they are looking for the best way to display that information for radiologists," he said.