Namkug Kim PhD, Abstract Co-Author: Nothing to Disclose

Joon Beom Seo MD, PhD, Abstract Co-Author: Speaker, Siemens AG

June-Goo Lee, Abstract Co-Author: Nothing to Disclose

Young Kyung Lee, Abstract Co-Author: Nothing to Disclose

Sujin Park, Presenter: Nothing to Disclose

Taekjin Jang, Abstract Co-Author: Nothing to Disclose

Youn Joo Lee MS, Abstract Co-Author: Nothing to Disclose

To develop and evaluate computer aided differential diagnosis (CADD) between usual interstitial pneumonia (UIP) and nonspecific interstitial pneumonia (NSIP) at HRCT. 

A CADD system was developed with double layered support vector machine (SVM) classifiers. The first layer is texture- and shape-based analysis of local HRCT image characteristics to classify whole lung parenchyma voxels into six local disease patterns (normal; ground-glass opacity; reticular opacity; honeycombing; emphysema; and consolidation). The second layer is the differential diagnosis algorithm based on the features of the extent and distribution of regional disease patterns including area fraction (AF), directional probabilistic density function (pdf) (dPDF: mean, SD, skewness of pdf per 3 directions: superior-inferior, anterior-posterior, central-peripheral), regional cluster distribution pattern (RDP: number, mean, SD of clusters, mean, SD of center of clusters). Left and right lungs were spatially normalized and evaluated separately. In addtition, asymmetric index (AI) (P_left – P_right/P_left) of all above parameters and disease division index (DDI) on every combination of AFs were evaluated. To evaluate the CADD system, fifty-four HRCT data sets in patients with pathologically diagnosed UIP (n=26) and NSIP (n=28) were used. The CADD system was optimized with sequential forward floating feature selection (SFFS). For comparison, two thoracic radiologists reviewed the whole HRCT images and diagnosed each case either as UIP or NSIP, independently. 

The accuracies of radiologists’ decision were 0.75 and 0.87, respectively. The accuracies of the CADD system using the features of AF, dPDF, RDP, DDI and AI’s of AF, dPDF and RDP were 0.70, 0.79, 0.77, 0.80, 0.78, and 0.81, respectively. The accuracy of the CADD with the selected features by SFFS was 0.91.

We developed the CADD system to differentiate between UIP and NSIP using automated assessment of the extent and distribution of regional disease patterns at HRCT.

CADD of various diffuse interstitial diseases at HRCT may be useful in supporting the radiologists’ decision. CADD may provide objective and reproducible differential diagnosis.

Kim, N, Seo, J, Lee, J, Lee, Y, Park, S, Jang, T, Lee, Y, A Computer-aided Differential Diagnosis between Usual Interstitial Pneumonia and Nonspecific Interstitial Pneumonia Using Assessment of the Extent and Distribution of Regional Disease Patterns at HRCT: Comparison with the Radiologist Decision.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034347.html