Spatial image variability analysis

Author

Pattichis, M.S. ; Cacoullos, T. ; Soliz, P.

Author_Institution

Dept. of Electr. Eng. & Comput. Eng., New Mexico Univ., Albuquerque, NM, USA

Volume

1

fYear

2002

fDate

4-7 Aug. 2002

Abstract

In our analysis, each image is partitioned into a number of non-overlapping, spatial regions. Each spatial region is viewed as a Region of Interest (ROI). A grade is assigned to each ROI from a set of independent raters. The model is applied in lung image analysis, where a grade of 0/1 is assigned to each ROI. Here, I represents success, while 0 represents failure to detect a hypothesized pattern in the region. Methods for establishing spatial symmetry and spatial growth analysis are presented. In addition, a Bayes and a novel summation classifier are used for classifying the entire lung based on their regional grades. When compared against the readers, the two classifiers have a total misclassification error of the order of inter-rater variability error.

Keywords

Bayes methods; image classification; lung; medical image processing; Bayes classifier; ROI; Region of Interest; hypothesized pattern; independent raters; lung image analysis; misclassification error; nonoverlapping spatial regions; spatial growth analysis; spatial image variability analysis; spatial symmetry; summation classifier; Biomedical imaging; Humans; Image analysis; Image processing; Lungs; Mathematics; Pixel; Radiography; Random variables; Shape;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 2002. MWSCAS-2002. The 2002 45th Midwest Symposium on

Print_ISBN

0-7803-7523-8

Type

conf

DOI

10.1109/MWSCAS.2002.1187197

Filename

1187197