Ikuo Kawashita PhD, Presenter: Nothing to Disclose

Ryuji Akita BS, Abstract Co-Author: Nothing to Disclose

Yasuhiko Okura PhD, Abstract Co-Author: Nothing to Disclose

Takayuki Ishida PhD, Abstract Co-Author: Nothing to Disclose

In the development of a computer-aided diagnosis / detection (CAD) system a large database containing training samples is of major importance. However, it is not easy to obtain database with enough size and variety. Therefore, the purpose of this study is to develop an implantation method of digital simulated disease similar to real image into digital radiograph. In this study we investigate the three diseases which include lung nodules and pneumoconiosis on chest radiographs and mass on mammograms.

Normal digital chest radiographs and mammograms were used for background of simulated disease. This method consists of the following two steps. First, three dimensional disease image data is simulated based on Iden model. Secondly, two dimensional projection image is obtained by ray summation with gradient and blur correction. Figure illustrates the example of simulated disease image. To evaluate the similarity with real images, receiver operating characteristic curve (ROC) curves attending two radiologists and five radiological technologists were compared.

As the results, area under the ROC curve (AUC) of lung nodule and pneumoconiosis and mass were 0.55, 0.58, 0.62, respectively. These AUC values were almost 0.5, the results show that a simulated images were similar to real images. The accuracy was 0.47, 0.48, 0.49, respectively. The sensitivity was 0.40, 0.40, 0.40, respectively. The specificity were 0.58, 0.63, 0.68, respectively.

We have developed an implantation method of digital simulated diseases into digital radiographs. This method would be useful for improvement of the performance of CAD systems, and training for regident and medical students.  

Simulated disese images of lung nodules and pneumoconiosis into chest radiographs and mass into mammograms can improve the performance of a computer aided detection system (CAD system).

Kawashita, I, Akita, R, Okura, Y, Ishida, T, Development of the Implantation Method of Digital Simulated Diseases into Digital Radiographs.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12043929.html