Sheng Chen, Presenter: Nothing to Disclose

Kenji Suzuki PhD, Abstract Co-Author: Nothing to Disclose

Heber Macmahon MD, Abstract Co-Author: Consultant, Riverain Medical Speaker, Konica Minolta Group Stockholder, Hologic, Inc

Major challenges in current computer-aided detection (CADe) of nodules in chest radiographs (CXRs) are to detect nodules that overlap with ribs and to reduce the frequent false positives (FPs) caused by ribs. Our purpose was to develop a CADe scheme with improved sensitivity and specificity by using “virtual dual-energy” (VDE) CXRs where ribs are suppressed with a massive-training artificial neural network (MTANN).

To reduce rib-induced FPs and to detect nodules overlapping with ribs, we incorporated VDE technology in our CADe scheme. VDE technology suppresses ribs in CXRs while maintaining soft-tissue opacity by using an MTANN that had been trained with real dual energy imaging. Our CADe scheme detected nodule candidates on VDE images by using a morphologic filtering technique. Sixty-four morphologic and gray-level-based features were extracted from each candidate from both original and VDE CXRs. A nonlinear support vector classifier was employed for classification of the nodule candidates. A publicly available database containing 126 nodules in 126 CXRs was used for testing. All nodules were confirmed by CT examinations. The average size of the nodules was 24.0 mm. Twenty nine percent (36/126) of the nodules were rated “extremely subtle” or “very subtle” by a radiologist.

With the original CADe scheme, a sensitivity of 63.5% (80/126) with 1.8 (227/126) FPs per image was achieved. By use of VDE images, more nodules overlapping with ribs were detected and the sensitivity was improved substantially to 73.0% (92/126) at the same FP rate (1.8 per image) in a leave-one-out cross-validation test (the improvement was statistically significant; p value < .05), whereas the literature shows that other CADe schemes achieved sensitivities of 66.0% and 72.0% with 5.0 FPs per image. For “very subtle” and “extremely subtle” nodules, a sensitivity of 52.8% (19/36) was achieved with 1.8 FPs per image, whereas an observer performance study in the literature shows that radiologists detected 44.0% of these nodules.

By using VDE images, the sensitivity and specificity of our CADe scheme for nodule detection in CXRs were improved substantially.

CADe with a high sensitivity for subtle nodules and a low FP rate would be useful for radiologists in detecting such lung nodules in CXRs.

Chen, S, Suzuki, K, Macmahon, H, Improved Computerized Detection of Lung Nodules in Chest Radiographs by Means of “Virtual Dual-Energy” Radiography.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9002380.html