Detection of small peripheral lung cancer by digital chest radiography. Performance of unprocessed versus unsharp mask-processed images.

PURPOSE

To clarify whether processed digital chest radiography can improve the detection rate for small peripheral lung cancer.

MATERIAL AND METHODS

Five radiologists independently interpreted 54 digitized chest radiographs of 18 patients with small peripheral lung cancers measuring less than 20 mm, which were displayed following 3 types of digital processing: 1) an original version; 2) unsharp mask processing with a type 1 filter (very low-frequency-enhancing, mid-frequency-suppressing, and high-frequency-enhancing filter); and 3) unsharp mask processing with a type 2 filter (very low- and high-frequency-enhancing filter). A total of 1,620 pooled observations were evaluated by receiver operating characteristic (ROC) analysis.

RESULTS

The mean area under the ROC curves was 0.68 for the type 1 filter, 0.68 for the type 2 filter, and 0.65 for the unprocessed (original) image. There were no statistically significant differences among these 3 kinds of image processing (p>0.05). In all types of images, the small lung cancer with an alveolar lining tumor growth was less visible than a solid tumor growth (p<0.01); the sensitivity increased with tumor size when the 3 groups of cancers, those measuring less than 10 mm, 11-15 mm, and 16-20 mm, were compared (p<0.01).

CONCLUSION

Unsharp mask-image processing of digital chest radiography will not improve the detection rate of small peripheral lung cancer, probably due to a substantial drawback: the limited conspicuity of cancer lesions in the surrounding lung and superposition of structures.