Junji Shiraishi, Presenter: Nothing to Disclose

Katsutoshi Sugimoto, Abstract Co-Author: Nothing to Disclose

Fuminori Moriyasu MD, PhD, Abstract Co-Author: Nothing to Disclose

Kunio Doi, Abstract Co-Author: Nothing to Disclose

We have been developing a computer-aided diagnosis (CAD) scheme for differential diagnosis of focal liver lesions (FLLs) on contrast-enhanced ultrasonography (CEUS). In this study we have performed an observer study for classification of FLLs with and without cues provided by the CAD.

Informed consent and institutional review board approval were obtained. Our data set contained 107 FLLs from 106 cases, consisted of 27 metastases, 27 hemangiomas, and 53 hepatocellular carcinomas (HCCs): 17 well-differentiated, 26 moderately differentiated, and 10 poorly differentiated. Pathologies of all cases except for hemangiomas were determined based on biopsy or surgical specimens. In the observer study, the CEUS dynamic images obtained with a micro-flow imaging (defined as <60 seconds after injection) was displayed together with static B-mode and Kupffer phase (defined as 10 minutes after injection) images. Seven radiologists were asked first to give confidence ratings regarding benign or malignant; second, if malignant, whether metastasis or HCCs; finally to select one of 5 classifications matched to the unknown FLL. Each radiologist read cases first without and immediately thereafter with CAD. Observers' rating data were analyzed with the multi-reader multi-case receiver operating characteristic (ROC) analysis.

The overall sensitivity of the CAD for differentiation of 5 types of 107 FLLs was 76.6% (82/107). For distinction between benign and malignant FLLs, the average area under the ROC curve (AUC) value for 7 observers was improved significantly from 0.861±0.068 to 0.942±0.045 by use of the CAD (p<0.001). In addition, as to metastasis vs. HCC, the mean AUC value was also improved significantly from 0.707±0.09 to 0.776±0.075 by use of the CAD (p=0.035). The overall sensitivity of all observers was also improved from 52.6% to 66.5%.

The use of the CAD can significantly improve the diagnostic performance of radiologists for differential diagnosis of the FLLs on CEUS.

The use of CAD for classification of focal liver lesions on contrast-enhanced ultrasonography could improve radiologists' diagnostic accuracy for differential diagnosis of focal liver lesions.

Shiraishi, J, Sugimoto, K, Moriyasu, F, Doi, K, Computer-aided Diagnosis for Differential Diagnosis of Focal Liver Lesions in Contrast-enhanced Ultrasonography: An ROC Study.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9010492.html