CAD helps detect early stage lung cancer
BOSTON--A computer-aided detection (CAD) method may help radiologists identify cancerous lung nodules at an early stage, according to a study presented today at the 2009 annual meeting of the American Roentgen Ray Society (ARRS).

Researchers from the University of Maryland Medical Center in Baltimore identified 3,000 patients who were diagnosed with lung cancer from an institutional registry between 1995 and 2006. Two radiologists retrospectively reviewed all chest radiographs prior to the discovery of the lung cancer to determine if a lesion was present.

"In our study, we identified 88 nodules that were not detected at the time of interpretation but visible in retrospect and were subsequently determined to be lung cancers," said the study's lead author Joseph Jen-Sho Chen, MD.

A total of 112 radiographs with missed lung cancer were found among 88 patients (mean age of 65 years, 79 men). Lesions were a mean size of 1.8 cm, and were most common on the right (55 percent) lung, in the periphery (75 percent) and in the upper lobes (79 percent).

"CAD was applied to the overlooked nodules and we found that 45-55 percent of the missed nodules were found using the CAD software. The implication of our study is that it is possible that at least some of the nodules representing lung cancer might have been diagnosed at an earlier stage, resulting in early treatment and perhaps a better outcome," Chen said.

The researchers reported CAD identified 45 percent, 50 percent and 55 percent of the nondetected lesions on a per-film basis, respectively; 44 percent, 48 percent and 54 percent on a per-patient basis. The false-positive rate per chest radiograph was 2.6, 3.2 and 3.6, respectively, Chen and colleagues said. There was a significant difference found in the location of the nodules detected by CAD and those that were not detected. However, no statistical difference was found in nodule size.

"The complexity of the structures in the chest including the ribs, mediastinum and pulmonary vessels can make it difficult to identify separate pulmonary nodules that may represent an early lung cancer from normal anatomy," he said. "CAD is a method that can be used to assist the radiologist in the search for lung cancer. The software highlights abnormalities that may be overlooked by the radiologist on an initial search."

"Lung cancer accounts for more than 150,000 deaths annually in the U.S. alone. Overall, only about 15 percent of patients survive five years or more, but with early detection, survival increases to greater than 70 percent," Chen said. "The use of CAD may be particularly valuable in early lung cancer, where the findings are often subtle," he said.

"We hope ultimately that studies such as ours will determine whether CAD should be adopted as part of the standard armamentarium for evaluating lung nodules," said Chen.