Cancer screening with chest radiography requires shift in medical practice

Despite the prevalence of lung cancer-an estimated 180,000 new cases in the U.S. annually-radiologists have had little luck improving the prospects of patients with the disease. Based on randomized clinical trials conducted in the 1970s, the National Cancer Institute has concluded that chest x-ray screening, a mainstay of the lung cancer fight in the 1960s, has no effect on lung cancer mortality.

Studies indicate that in an estimated seven out of 10 cases, cancer is diagnosed after it has metastasized. The five-year survival rate for those diagnosed with lung cancer today is 14%, a figure that has increased only 2% in the past 30 years, according to the American Cancer Society. But computer-aided detection (CAD) could change that.

When cancer is detected in stage I, survival rates soar to 90%, according to Dr. Michael Yeh, president and CEO of Deus Technologies in Rockville, MD. In July, the company received FDA clearance for its RapidScreen RS-2000, the first device to be approved for lung cancer CAD.

Typically, stage IA detection rates for lung cancer using chest x-rays average 15%, Yeh said. In clinical trials conducted at Georgetown University Medical Center, radiologists using the RapidScreen detected stage IA cancer in nearly 80% of cases.

”The biggest improvement for radiologists using the RapidScreen is in detecting cancer at its earliest stage, in nodules ranging from 9 to 15 mm,” Yeh said.

A 1999 study conducted at the University of Chicago confirms CAD’s potential. The randomized study involved 20 abnormal chest exams, each with a single nodule, and 20 normal radiographs, all of which were digitized with a laser scanner. Detection accuracy was substantially higher for all categories of observers-who included chest radiologists, other radiologists, radiology residents, and nonradiologists-when CAD was used.

Deus’ RapidScreen, which spent three years in development, uses fuzzy logic technology and contains elements of artificial neural networks based on thousands of chest images in the Georgetown University database. The Windows-based system analyzes digitized images of existing PA or AP chest radiographs and uses image enhancement and pattern recognition techniques to locate suspicious areas that may warrant a second review by a radiologist.

In routine use, a film librarian feeds chest x-rays into the RapidScreen scanner deck. The exams are scanned, digitized, and computer-analyzed. The operator receives a paper print of the results with suspicious areas marked. The film and the RapidScreen print-out are then interpreted by a radiologist.

Deus inked a distribution agreement with Marconi Medical Systems in August. Since RapidScreen’s approval in July, the company has sold six of the $169,000 systems, according to Richard Smillie, vice president of sales and marketing. Interest is high, he added, especially in Japan, which has a formal, government-sponsored lung cancer screening program. But marketing a lung cancer CAD product based on chest radiography is challenging.

”Chest x-ray was written off by the medical community for early detection of lung cancer,” Yeh said. “More doctors and hospitals are thinking about low-dose spiral CT for lung cancer.”

Cancer screening has reentered the medical consciousness on the coattails of multislice CT, which has reinvigorated the CT community over the past several years with its advanced capabilities resulting from enhanced speed and resolution. Several vendors are developing CAD algorithms capable of spotting lesions and then reconstructing these lesions volumetrically for interpretation by radiologists. Yeh maintains, however, that lung cancer detection using chest x-rays and CAD is clinically competitive and more cost-effective than the combination of CT and CAD.

”Moreover, concerns have been raised about using CT for lung cancer screening,” he said.

The FDA issued a warning earlier this year that annual screening with low-dose CT could result in unnecessary radiation absorption. There is also reason to believe that CT results are not specific to lung cancer, Yeh said.

”Our experience with CT indicates that it detects a lot of small nodules that are benign, so follow-up is required,” he said. “As a result, the average expense from initial screening through follow-up using CT could be up to $2000.”

That figure dovetails with recently released results from the second year of the Early Lung Cancer Action Project (ELCAP), an ongoing study of 1000 high-risk subjects undergoing annual low-dose CT screening for lung cancer. The program costs about $2500 per life-year saved, according to an analysis of the data presented by Dr. Claudia Henschke, chief of medical imaging at Weill Medical College of Cornell University in New York.

Yet ELCAP has reported encouraging results based on lung cancer screening with CT. In its first report in 2000, the ELCAP research team diagnosed 27 cancers using low-dose CT compared to seven with chest radiography. Twenty-three of the 27 cancers identified with CT were stage I malignant disease; 26 were resectable. Eighty-five percent of the early-stage cases were not detected with radiography.

Whether CAD and chest x-ray can effectively compete with low-dose CT is not known. No published study has yet compared performance of the two technologies, although at least one vendor, R2 Technologies of Los Altos, CA, is working with Cornell to do just that.

Nor is Deus ignoring the potential of CT scanning for lung cancer. The company is fine-tuning its experimental, CT-based CAD package, which will be on display at the RSNA meeting. The RapidDisplay application offers segmentation of anatomic structures and detects regions of interest using volume matching and 3-D display. In addition, the company will showcase a direct digital version of the RapidScreen system that interfaces with both computed radiography and digital radiography devices, Smillie said.

The challenge of breathing new life into chest x-ray-based lung cancer screening is daunting. Another hurdle is educating the public about the health hazards of lung cancer. While lung cancer cases annually total more than breast, prostate, and colon cancer combined, the disease carries a misperception that is tough to shake.

”There is a pervasive attitude among the public that if you have lung cancer, it’s because you smoked,” Smillie said. “In fact, one in three people who get lung cancer have never smoked. The media have done a very good job educating women about breast cancer, and how life-threatening it is. We need to do just as good a job educating people about lung cancer.”