Regional contrast enhancement and four-directional thresholding techniques for pulmonary nodule extraction and discrimination

• Main Author: Heidari, Saleheh
• Format: Thesis
• Language: English
• Published: 2015
• Online Access: http://psasir.upm.edu.my/id/eprint/57115/1/FSKTM%202015%2010RR.pdf; http://psasir.upm.edu.my/id/eprint/57115/

Automated pulmonary nodules extraction and lung disease diagnosis by Computer Aided Diagnosis (CAD) systems is a challenging task. Generally, the CAD system utilizes the Computed-Tomography (CT) images to diagnose tumor and observe its condition during the treatment process. Due to extensive similarity between pulmonary vessels, bronchus and arteries in lung region and the low contrast of Computed-Tomography (CT) images the accuracy of lung tumor diagnosis is highly dependent on image’s contrast and the precision of segmentation. Contrast enhancement and image segmentation are the most prominent image preprocessing techniques that are utilized as a primary and essential steps of almost every pathological applications. Thus, a particular contrast enhancement and image thresholding techniques are required to enhance the contrast of lung CT image by refining their pixels’ intensity value and overcome the difficulties of precise segmentation as well as facilitating the accurate pulmonary nodule extraction. Accordingly, in this research Regional Contrast Enhancement (RCE) and Four-Directional Thresholding (FDT) techniques are introduced followed by nodule extraction and their discrimination based on their respective size and circularity measurements. Regional Contrast Enhancement (RCE) technique aims to improve the CT image’s visual quality by boosting the contrast of lung CT images and modifying the image histogram by implementing the proposed algorithm on every individual pixel based on their intensity value and their regional variations. The proposed FDT technique also aims to augment the precision of lung CT image’s segmentation by implementing a specific thresholding approach from four different directions in which the determination of pixels’ value as being either on foreground or background is highly dependent on its adjacent pixel’s intensity value and the final decision is made based on all four directions’ thresholding results. Finally, pulmonary nodules are extracted from thresholded CT images by several morphological techniques and then extracted candidates are discriminated based on their eccentricity and corresponding size as benign and malignant nodules. To demonstrate the superiority of proposed RCE technique the minimum Absolute Mean Brightness Error (AMBE), the highest Peak Signal to Noise Ratio and structural Similarity Measurement Index obtained by RCE technique are compared with the other advanced contrast enhancement by histogram equalization methods. The effectiveness and high exactitude of proposed FDT method also has been evaluated on different CT images by correlation and regional non-uniformity measurement criteria. Ultimately, the performance of nodule extraction and discrimination were evaluated and 93.33% of sensitivity, 93.90% of accuracy and 94.59% of specificity have been obtained.