RSNA 2003 

Abstract Archives of the RSNA, 2003


501-p

Effect of a Computer-aided Diagnosis(CAD) System on Radiologists' Performance of Distinction between Benign and Malignant Pulmonary Nodules on Digital Chest Radiographs Using Temporal Subtraction Images

Scientific Posters

Presented on December 3, 2003
Presented as part of L03: Chest Lung Nodules

Participants

Shoko Yamada MD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: To evaluate the effect of a computer-aided diagnosis (CAD) system on radiologists' performance of distinction between benign and malignant pulmonary nodules on digital chest radiographs using temporal subtraction images. Methods and Materials: The temporal subtraction image can enhance pulmonary nodules on a current image by subtracting a previous image. Therefore, we used the temporal subtraction image as well as the original image to obtain image features for distinction between benign and malignant nodules. First, nodule candidates were extracted by multi-level thresholding of the temporal subtraction image. Then, twenty one image features were determined, which included three features based on the shape and eighteen features obtained from the histogram analysis of pixel values for inside region of the nodule candidate. Finally, the discriminant analysis with Mahalanobis distance was used for nodule detection and determination of likelihood of malignancy. Forty digital chest radiographs (20 primary lung cancers and 20 benign nodules) were used in the observer performance test. Chest radiographs with pulmonary nodules larger than 3 cm were excluded and no nodules had evidence of calcification on CT images. Twelve radiologists participated in the observer test and evaluated pulmonary nodules without and with a computer output. The observers' performance for distinction between benign and malignant nodules was evaluatede by the Az value, the area under a receiver operating characteristic (ROC) curve. Results: The observer performance with the computer output was significantly improved (p<0.001) in comparison with that without the computer output. Conclusion: This CAD system can assist radiologists to make distinction between benign and malignant pulmonary nodules on digital chest radiographs, and has a potential to reduce the number of unnecessary CT examinations.       Questions about this event email: yy-shoko@med.uoeh-u.ac.jp

Cite This Abstract

Yamada MD, S, Effect of a Computer-aided Diagnosis(CAD) System on Radiologists' Performance of Distinction between Benign and Malignant Pulmonary Nodules on Digital Chest Radiographs Using Temporal Subtraction Images.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3101242.html