Radiology: Image processing software aids in lung cancer detection
New bone suppression (BS) image processing software, used with a standard x-ray, can improve radiologists’ accuracy in detecting small lung cancers on chest films, according to a study published Sept. 22 online in Radiology.

Feng Li, MD, PhD, of the department of radiology at the University of Chicago, and colleagues wanted to determine whether the use of BS imaging could improve radiologists’ performance in detecting small lung cancers. The researchers also wanted to compare the results of BS imaging with the accuracy rates achieved using dual-energy subtraction (DES) radiography.

The authors observed a significant increase in accuracy using the new techniques, and wrote that “while DES radiography provides the highest accuracy, use of BS software in combination with a standard radiograph can significantly improve detection of small nodular cancers compared with use of standard radiographs alone.”

DES techniques are able to distinguish bone from soft tissue and have been shown in previous observer performance studies to improve the ability to identify lung nodules, according to the researchers.

BS image processing uses software to suppress the conspicuity of bones, and while the authors reported the BS image appeared very similar to DES soft-tissue images, the bone subtraction was less complete in the apical areas and local enhancement was more pronounced.

The downside of DES is that it requires special equipment and a small increase in radiation dose, making alternatives like the software-only BS image processing worth investigating, offered Li and colleagues.

Ten observers, including six experienced radiologists and four radiology residents, assessed a number of images and rated their confidence levels in detecting the presence or absence of a lung cancer first by using standard imaging, then a BS image and lastly with a DES image. Included in the study were radiographs featuring 50 patients with a total of 55 confirmed primary nodular cancers and 30 patients without cancer. Receiver operating characteristic (ROC) analysis was then used to evaluate the radiologists’ performance.

Results showed the average area under the ROC curve for all observers improved from 0.807 to 0.867 with BS imaging and to 0.916 with DES. The six experienced radiologists were able to achieve higher evaluations, improving to 0.894 with BS images and to 0.945 with DES images.

To emphasize the importance of finding an effective way of minimizing bone obstruction in lung cancer testing, the authors cited a previous study that showed the vast majority of missed lung cancers were missed due to the presence of superimposed bones. Of the 40 missed cancers investigated by Shah et al, 38 were partly obscured by overlying ribs, including 26 that were obscured by multiple bones. The authors found similar results in a previous study of their own.

The researchers wrote that they never intended for BS imaging to completely replace standard radiograph in other applications as it is optimized for nodular opacities. They also don’t believe that BS images will become the preferred method over DES in situations where DES is used currently. “However, for the large majority of digital chest radiographs that are obtained without energy subtraction, including bedside (portable) examinations, it seems likely that BS could be beneficial,” wrote the authors.