Middle-East Journal of Scientific Research 23 (3): 436-441, 2015
ISSN 1990-9233
© IDOSI Publications, 2015
DOI: 10.5829/idosi.mejsr.2015.23.03.22088
Corresponding Author: Bottu Balagangadhar, Department of ECE Sathyabama University, Chennai, India.
436
A Survey on PET-CT Lung Tumor Delineation
Bottu Balagangadhar and K. Srilatha
Department of ECE Sathyabama University, Chennai, India
Abstract: Accurate parenchymal lung tumor delineation with PET-CT can be problematic given the inherent
tumor heterogeneity and proximity / involvement of extra-parenchymal tissue. In this proposed method, we
propose a tumor delineation approach that is based on new tumor–Ground truth method models for a given
image. The ground truth method give more efficient result compared to other methods. Then this method
extends to segmentation, region based graph methods (n-cut and a-cut) and another segmentation model is
k-means algorithm is used. After the tumor detection part move on to classification part, this classification done
by neural network based implementation has been used. The result computation is very fast the more efficiency.
In background likelihood method not efficient compare other methods. The above model is processed in
MATLAB tool and achieve high efficiency in the detection and classification model compared to previous one.
Key words: CT Dicom Ground truth PET Segmentation
INTRODUCTION applying conventional Thresholding methods and
Lung cancer is the one of the most leading cause of Currently, the most effective image technique for early
cancer death in men and the second leading cause of detection of lung cancers is PET-CT imaging technique
women. Every year, nearly more than 90,000 men and and it is the most reliable tool [2].
79,000 women are diagnosed because of cancer. Now
coming to cancer statistics about 12.7 millions cancer Lung Tumor: Cancer is a group of diseases that are
cases and 7.6 million cancer deaths are estimated to have characterized by an abnormal and unregulated growth of
occurred in 2008. This result tells us one fourth of the cells. Those cells are we can see in Fig. 1. The main
population will be diagnosed with the cancer in their differences are that beginning tumor grows slowly and it
lifetime and one fifth of the cases will go to die from the will usually not come back if it is surgically removed.
lung tumor disease [1]. There are two groups of lung cancer, those are Small cell
Most of them notice their disease when it’s too late Lung cancer (SCLC) and the other one is Non-Small Cell
and the surgery will be tough to do. In 2011, lung cancer Lung Cancer (NSCLC), which covers more than 85% of all
has been the main cause of death from cancers by 15%. cases. CT scans (Computed Tomography), PET scans
Today, several types of techniques are used diagnosis of (Positron Emission Tomography) and MRI (Magnetic
the disease, such as Computerized Tomography, Positron Resonance Imaging) is the diagnosing methods are
Emission Tomography, Chest Radiograph (x-ray), available to detect Lung cancer; [3].
Magnetic Resonance Imaging (MRI scans). We can calculate the lung tumor based on tumor size
These techniques can detect only in its advanced and we can give which stage it is also calculated. Now we
stages, but we need the technique to diagnose in its early will see stages of lung tumor.
stages. Different software tools are use diagnose the lung This is an early stage of lung tumor; some symptoms
cancer, such as 3D-Doctor, Workstation, analyzes, gave clarity about lung cancer. Those symptoms are
3dviewnix, 3D slicer. Some of these software tools as weight loss, chest infections, coughing up blood, difficult
image support some visual formats like Dicom and to take a breath, tired, lack of energy and shoulder pain.
analyze. Some tools provide features such as processing These are very early symptoms of lung cancer. There
and visualization for instant registration, noise many reasons caused by to get a lung cancer. Those
suppression, analyzing images for diagnostic purposes, reasons we can see in Figure 3 [4].
representing 2D and 3D data using visualization methods.