Volume 3, Issue 1

Multi-Threshold Image Segmentation for Lung Cancer Image Mining

Author

Samatha G.*1, Manjula D.V.1, and Madhu K.2

Abstract

In the medical image mining, the mining pattern is directly depending upon the features, which are extracting from images at the pre-processing stage. Therefore accurate and robust segmentation is essential to estimate these parameters for quantitative assessment of medical images in order to achieve correct clinical diagnosis. Global threshold determination for a lung CT image based on histogram much simpler and less time consuming process, but it fails to segment the image more effectively. In order to overcome this limitation, a new method, based on an iterative-Otsu optimization approach is proposed for multi-thresholds determination. The objective of this work is to estimate the lung nodule size more accurately by utilizing “between class-variance” criterion of Otsu method in an iterative way. The algorithm runs through n-number of iterations and exits when reaches a preset threshold difference value. We applied this method on few lung CT images from the National Lung Screening Trail database to verify the proposed methodology against single threshold results. The extracted features are presented for the comparison. The preliminary results show that the proposed algorithm is better performing in estimate the nodule size efficiently for Lung CT images.

DOI

https://doi.org/10.62226/ijarst20140186

PAGES : 17-22 | 31 VIEWS | 78 DOWNLOADS

Download Full Article