Automated thresholding of lung CT scan for Artificial Neural Network based classification of nodules

Author

Akram, Sheeraz ; Javed, Muhammad Younus ; Hussain, Ayyaz

Author_Institution

Dept. of Comput. Eng. (DCE), Nat. Univ. of Sci. & Technol. (NUST), Islamabad, Pakistan

fYear

2015

fDate

June 28 2015-July 1 2015

Firstpage

335

Lastpage

340

Abstract

In this paper, the threshold value of overlapped circular region is calculated. The lung volume is segmented by thresholding, lung lobe extraction, hole filling and contour corrected. The regions of interest are segmented from extracted lung volume. The candidate nodules are selected from the ROIs. The features of candidate nodules are extracted. Artificial Neural Network classifier is trained and tested on the dataset. The proposed methodology produces sensitivity of 96.55% with accuracy of 91.87% and 0.40 FP/scan.

Keywords

computerised tomography; feature extraction; image classification; image segmentation; lung; medical image processing; neural nets; ROIs; artificial neural network based classification; candidate nodule extraction; contour correction; hole filling; lung CT scan automated thresholding; lung lobe extraction; overlapped circular region threshold value; region of interest segmentation; Artificial neural networks; Cancer; Computed tomography; Feature extraction; Lungs; Sensitivity; Three-dimensional displays; Classification; Computed Tomography; Geometric Features; Segmentation; Statistical Features;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer and Information Science (ICIS), 2015 IEEE/ACIS 14th International Conference on

Conference_Location

Las Vegas, NV

Type

conf

DOI

10.1109/ICIS.2015.7166616

Filename

7166616