Literature DB >> 11274847

Computerized classification of interstitial lung abnormalities on chest radiographs with normalized radiographic index and normalized fractal dimension.

S Kido1, S Tamura.   

Abstract

OBJECTIVE: To evaluate the performance of two kinds of physical measures, the normalized radiographic index (R) and the normalized fractal dimension (F), for computerized classification of interstitial lung abnormalities on chest radiographs. METHODS AND MATERIAL: The values of R were obtained as the normalized percent area of extracted opacities in selected regions of interest (ROIs). The values of F were calculated with a box-counting algorithm and then normalized. To extract linear opacities on chest radiographs selectively, we processed ROIs by four-directional Laplacian-Gaussian filtering and binarization (4LG/B), linear opacity judgment (LOJ), and linear opacity subtraction (LOS). We used the ROIs of 50 mild and 50 severe interstitial lung abnormalities. In both groups, all cases were divided into H (n=21, honeycombing opacities were found to be dominant) and Non-H (n=79, abnormal opacities were found, but these were excluded from H). We obtained three types of normalized physical measures of R and F in one ROI from 4LG/B, LOJ, and LOS images, and the combined indices, R(COM) and F(COM) were calculated.
RESULTS: The values of F(LOJ) could differentiate H from Non-H in the mild-and the severe-abnormality group. However, all Rs could not differentiate H from Non-H in the severe-abnormality group. The combined indices of both R and F could differentiate H from Non-H in the mild-abnormality group; however, these could not differentiate H from Non-H in the severe-abnormality group.
CONCLUSION: The values of F(LOJ) seem to be useful in the classification of interstitial lung abnormalities on chest radiographs.

Entities:  

Mesh:

Year:  2001        PMID: 11274847     DOI: 10.1016/s0720-048x(00)00296-5

Source DB:  PubMed          Journal:  Eur J Radiol        ISSN: 0720-048X            Impact factor:   3.528


  1 in total

1.  Remote-sensing image classification based on an improved probabilistic neural network.

Authors:  Yudong Zhang; Lenan Wu; Nabil Neggaz; Shuihua Wang; Geng Wei
Journal:  Sensors (Basel)       Date:  2009-09-23       Impact factor: 3.576

  1 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.