A SURVEY ON EARLY LUNG CANCER DETECTION AND ITS CLASSIFICATION

A SURVEY ON EARLY LUNG CANCER DETECTION AND ITS CLASSIFICATION USING IMAGE PROCESSING TECHNIQUES

K.Karthika 1 , Dr.G.R.Jothi lakshmi 2

1 Research Scholar, Department of ECE, Vels University, Pallavaram, Chennai, Tamilnadu.

2 Assistant Professor, Department of ECE, Vels University, Pallavaram, Chennai, Tamilnadu.

Abstract

Lung cancer is one of the dangerous and life taking threatening in the world. However, early diagnosis and treatment can save life. There are various types of cancers i.e. lungs cancer, Breast cancer, blood cancer, throat cancer, brain cancer, tongs cancer, mouth cancer etc.

Lung cancer is a disease of abnormal cells multiplying and growing into a tumor. Cancer cells can be carried away from the lungs in blood, or lymph fluid that surrounds lung tissue. Although, CT scan imaging is best imaging technique in medical field, it is difficult for doctors to interpret and identify the cancer from CT scan images.  To predict the lung cancer various features are extracted from the images therefore, pattern recognition based approaches are useful to predict the lung cancer.

Get quality help now

writer-Charlotte

Verified writer

Proficient in: Cancer

4.7 (348)

“ Amazing as always, gave her a week to finish a big assignment and came through way ahead of time. ”

+84 relevant experts are online

Hire writer

Here, a comprehensive review for the prediction of lung cancer by previous researcher using image processing techniques is presented.

Keywords: Lung cancer, CT scan images, prediction, Cancer cell.

1. INTRODUCTION

Lung cancer is one of the causes of cancer deaths. It is difficult to detect because it arises and shows symptoms in final stage. However, mortality rate and probability can be reduced by early detection and treatment of the disease. Best imaging technique CT imaging are reliable for lung cancer diagnosis because it can disclose every suspected and unsuspected lung cancer nodules [1].

However, variance of intensity in CT scan images and anatomical structure misjudgment by doctors and radiologists might cause difficulty in marking the cancerous cell [2]. Recently, to assist radiologists and doctors detect the cancer accurately computer Aided Diagnosis has become supplement and promising tool [3]. There has been many system developed and research going on detection of lung cancer. However, some systems do not have satisfactory accuracy of detection and some systems still has to be improved to achieve highest accuracy tending to 100%. Image processing techniques and machine learning techniques has been implemented to detect and classify the lung cancer.

Lung cancer has the second most fatality rate among other categories of cancer. Even after diagnosis it has smallest survival rate, thereby continuously increasing the death rate yearly. Growth of lung cancer during diagnosis is related to its survival rate. But if the cancer cells been diagnosed in its early stages ones survival rate increases. Cancer cells can borne from the lungs in blood, or lymph fluid surrounding the lung tissue. Lymph floods through lymphatic vessels, which duct into lymph nodes placed in the lungs and in the centre of the chest. The lung cancer generally spread within the middle of the chest due to the usual outflow of lymph from lungs is toward the centre of the chest.

Generally it is grouped into Non-Small Cell Lung Cancer and Small Cell Lung Cancer. Depending on the cellular characteristics these assign the lung cancer types. As the stages, generally there are four stages of lung cancer; I - IV. The Stages are based on size and location of tumor and location of lymph node. Currently, CT are found more efficient than plain chest x-ray in detecting and diagnosing the lung cancer. In this paper, discussing various techniques of image processing to detect the stages of lung cancer more accurately.

2. RELATED WORKS

Ms. Twinkal Patel et.al [4] The core factors of this research are image quality and accuracy. The local energy-based shape histogram (LESH) feature extraction technique was recently intended for lung cancer diagnosis. We extend our work to apply LESH and sensitivity analysis (SA) to detect lung cancer. The JSRT & clinical dataset is selected for research experiments. This process will lead to a more generalized process for all kind of dataset and this approach can give better results than the earlier one.

Tiantian Fang [5] a fast, accurate, and stable system of lung cancer detection based on novel deep learning techniques is proposed. A convolutional neural network (CNN) structure akin to that of GoogLeNet was built using a transfer learning approach. In contrast to previous studies, Median Intensity Projection (MIP) was employed to include multi-view features of three-dimensional computed tomography (CT) scans. The system was evaluated on the LIDC-IDRI public dataset of lung nodule images and 100-fold data augmentation was performed to ensure training efficiency.

De-Ming Wong et.al [6] e focused on the method of lung cancer identification by breath. Lung cancer had occupied the first place in the top ten leading causes of death. When lung cancer patients were diagnosed, most of the patients had lost the opportunity of cure. However, physicians determined the lung cancer cases in complicated steps. Therefore, the purpose of this breath detection system was to help physicians to quickly screen for rapid screening lung cancer. We used KNN and SVM with leave-one-out cross validation to analyze.

Qing Wu et.al [7] propose a novel neural-network based algorithm, which we refer to as entropy degradation method (EDM), to detect small cell lung cancer (SCLC) from computed tomography (CT) images. This research could facilitate early detection of lung cancers. The training data and testing data are high-resolution lung CT scans provided by the National Cancer Institute. We selected 12 lung CT scans from the library, 6 of which are for healthy lungs, and the remaining 6 are scans from patients with SCLC.

Janee Alam et.al [8] Recognition and prediction of lung cancer in the earliest reference point stage can be very useful to improve the survival rate of patients. But diagnosis of cancer is one the major challenging task for radiologist. For detecting, predicting and diagnosing lung cancer, an intelligent computer-aided diagnosis system can be very much useful for radiologist. This paper proposed an efficient lung cancer detection and prediction algorithm using multi-class SVM (Support Vector Machine) classifier. Multi-stage classification was used for the detection of cancer. This system can also predict the probability of lung cancer. In every stage of classification image enhancement and segmentation have been done separately.

N. Hadavi et al. [9] presented a technique for automatic detection of lung cancer by using cellular learning automata. Image enhancement was performed using Gabor filter. Thresholding technique was used for image segmentation because of its advantages such as fast processing and easy influence. Features are extracted as nodule size, shape, contrast and the region for analysis. The new technique used Cellular automata is a mathematical Model. It is composed of lattice of cells where each cell has a set of stats and local rules governing them.

M. Mirah Kasturi et.al [10] proposes a method to locate and detect the cancerous cells effectively from the 2D and 3D CT scan images by reducing the detection error made by the physicians' naked eye. Early stage of cancer is diagnosed by using image enhancement and Sobel edge segmentation techniques .

Amutha et al. [11] has proposed level set Active shape model for the identification of lung tumor. This system depended on part capacity having the base mean square error value. At that point second request components were computed which depended on the histogram of the noise free image. The classification between the normal and abnormal lung image was made on these components.

Hongmei et al. [12] provides an automatic segmentation method on PET images taking into account the random walks (RW) algorithm and an expansion of the random walks structure to coordinate a tumor development data, which is the anticipated tumor area coming about because of a model for lung tumor development and reaction to radiotherapy. The region of interest (ROI) and named seeds are consequently created.

Yuhua et al. [13] has proposed a single click ensemble segmentation (SCES) methodology in view of a current ""Click & Grow"" algorithm. The SCES methodology requires stand out administrator chose seed point as contrasted and different administrator inputs. Additionally proposed a accurate, accurate and programmed single tick group division algorithm in this paper. The critical part of this work is to decrease the human communications while sore depiction stays exact and predictable as a consequence of ensemble segmentation. Despite the fact that the calculation time was expanded for every case subsequent to various ""Click&Grow"" algorithms were connected. The tumor segmentation ought not contrast much with distinctive manual seeds gave by different readers.

Dandil et al. [14] has proposed a programmed Computer Aided Diagnosis (CAD) framework that effectively separates the lung nodules as benign or malignant on CT images. The proposed CAD framework is an incorporated structure since it incorporates pre-preparing, segmentation, feature extraction, feature selection and classification steps. SOM strategy incorporated into CAD framework permits fruitful identification of lung nodules in ahead of schedule stages. ANN was favored high exactness rates in classification.

Talebpour et al. [15] provides a new computer-aided detection (CAD) framework exhibited that distinguish little size nodules (bigger 3 mm) in High Resolution CT (HRCT) images. In the initial step, the lung district is separated, and then with a sort of 3D filtering nodule assumed cases is established. In the last step, a neural system is utilized for false positive reduction. For filtering nodule cases from different items in images, it's utilized a cylindrical filter. The discovery execution was assessed tentatively utilizing lung LIDC (Lung Image Database Consortium) image database. Suitable results demonstrate that the utilization of the 3D model and the features analysis based Feature-based (FPs) decrease can precisely identify nodules in HRCT images. The technique has generally great speed and can be promising in clinical application. The outcomes demonstrate that the strategy is strong to different nodule shapes, simple to utilize, does not require any client activity and can be connected for other imaging modalities.

Anam et al. [16] has proposed a modernized framework for lung nodule recognition in CT scan images. The framework comprises of two stages i.e. lung segmentation and enhancement, feature extraction and classification. Threshold segmentation is connected to evacuate foundation and concentrates the nodules from an image. A feature vector for conceivable abnormal regions is ascertained and areas are arranged utilizing neuro fuzzy classifier. Framework encourages the location of little nodules which prompt early analysis of lung tumor.

Hui Cui et al. [17] provides an improved RW model that fully utilizes the prior knowledge on PET for lung tumor segmentation from low- contrast CT. The impact of the tumor certainty area and the strolling extent were utilized and the forefront and foundation seeds were acquired in light of the PET automatically.

A.Kulkarni et al. [18] proposed a system on lung cancer detection using CT images in DCOM format. Image smoothing was done by Median filter. It reduces blurring of edges. The advantage of using median filter in the system is that it is not affected by individual noise spike, eliminates impulsive noise quite well and it does not blur edges much and can be applied iteratively. Gabor filter is used for enhancement purpose as it gives better result compared to Fast Fourier Transform and auto enhancement. Image presentation based on Gabor function constitutes an excellent local and multi-scale decomposition in terms of logons that are simultaneously localization in space and frequency domain.

Fatma Taher et.al [19] present a feature extraction process followed by a rule based classification technique to classify the sputum cell into cancerous or normal cell. We used 100 sputum color images to test the rule based method. The performance criteria such as sensitivity, precision, specificity and accuracy were used to evaluate the proposed techniques. The evaluation demonstrated the advantages of the new technique.

BIPIN NAIR B J et.al [20] propose an efficient tool to analyze the possibility of getting affected by Non-Small Cell Lung Cancer (NSCLC) by comparing Lung Cancer microRNAs (LC-miRNAs) structures. Here we use global optimal alignment and TargetScan for target comparison and binding location detection. A previous research showed that major lung cancer genes are targeted by 8 type miRNAs. These 8 LC-miRNAs (let-7a-1, miR-7-1, miR-17, miR-21, miR-96, miR-125a-5p, miR-128b, and miR-145) were used for this analysis for accuracy in research.