An efficient image segmentation and classification of lung lesions in pet and CT image fusion using DTWT incorporated SVM

doi:10.1016/j.micpro.2021.103958

Microprocessors and Microsystems

Volume 82, April 2021, 103958

https://doi.org/10.1016/j.micpro.2021.103958 Get rights and content

Abstract

Automatic segmentation and classification of fused lung Computed Tomography (CT) and Positron Emission Tomography (PET) images is presented. This system consists of four basic stages: 1). Lung image fusion process; 2). Segmentation of fused lung CT/PET images; 3). Post pre-processing; 4). Classification of fused lung images. In the first step, the lung image fusion process is made by deep learning method. At first, the input CT/PET images are decomposed by Dual Tree m-band Wavelet Transform (DTWT). The coefficients of DTWT are fused by deep learning method. This fused image of CT/PET is the input for the following steps. In the next step, the fused CT/PET images are decomposed by DTWT. It produces lower and higher frequency sub-band coefficients. Then the lower frequency components are set to zero. Then higher frequency components are used for reconstruction. Then the clustering-based thresholding method is used for segmentation. In post pre-processing step, the unwanted small regions are removed by morphological operations. Then the lung region is detected. At last, in the classification step, the features are extracted by the intensity and texture-based features. These features are classified by hybrid classifiers like Support Vector Machine (SVM) are used. The performance of the system has a higher classification accuracy of 99% using SVM classifier.

Introduction

The uncontrollable division of cells in the lung region is known as lung cancer. Lung cancer causes the struggle in breathing. This growth of tumor can spread through the nearby regions and other body parts. The different type of lung cancers includes; non-small cell and small cell lung cancer. The symptoms of lung cancer include difficulty in breathing, pain in the chest, coughing with blood and weight loss. The lung cancer is mainly caused by smoking. The other causes include genetics, air pollution, asbestos and radon gas. Lung cancer is one of the leading causes of death. In India 2018, the mortality rate of lung cancer is 8.82%. To reduce the mortality rate, early diagnosis is required for lung cancer.

Section snippets

Literature survey

Classification of lesion margin in CT lung images for feature extraction is described by Riti et al. [1]. Initially, the CT lung images are preprocessed to obtain the Region Of Interest (ROI) area. Segmentation is made by Otsu's thresholding. Some unwanted areas which are obtained after segmentation are removed by connected component labeling. Then the features are extracted by compactness, convexity, circularity and solidity. Multi-Layer Perceptron (MLP) is used for classification.

Methods and materials

An efficient method for lung lesion segmentation and classification method using fused CT/PET images is discussed. Fig. 1 shows the overall workflow of classification and segmentation of CT and PET images. The workflow of this system is as follows: Initially, the CT and PET lung images are separately decomposed by DTWT. Then the coefficients of DTWT are used as input for the fusion process. The deep learning method is used for the fusion of CT/PET images. This fused image of CT/PET is the input

Results and discussion

The performance of fused CT/PET lung image classification and segmentation is discussed in this section. The CT/PET lung images are taken from the Reference Image Database to Evaluate therapy Response (RIDER) lung PET-CT database [29]. To order to reach an initial consensus on how to harmonize data collection and quantitative imaging research used to assessing reactions to drugs and radiation therapy, RIDER is a focused data collecting method. The work flow of segmentation and classification of

Conclusion

An efficient method for classification and segmentation of fused CT/PET lung image is presented. In this study, CNN is used for fusion of CT/PET lung images. The fused lung image is decomposed by DTWT, and the lower frequency components are set to zero. Then inverse transform is applied to reconstruct the image. The reconstructed lung image is used for FCM cluster-based thresholding technique. Morphological operation techniques are used for the removal of unwanted region and detection of a lung

Declaration of Competing Interest

1. All authors have participated in (a) conception and design, or analysis and interpretation of the data; (b) drafting the article or revising it critically for important intellectual content; and (c) approval of the final version.

2. This manuscript has not been submitted to, nor is under review at, another journal or other publishing venue.

3. The authors have no affiliation with any organization with a direct or indirect financial interest in the subject matter discussed in the manuscript.

R. Mohana Priya has received her B.E. Degree in Electronics and Communication Engineering from University of Madras, Tamilnadu in 2004.M.E Degree in Applied Electronics from VMRF in 2010.She is pursuing her Ph.D in the field of Image Processing in Dept.of ECE from Sri Chandrasekharendra Saraswathi ViswaMahavidyalaya, Kanchipuram. Her research areas includes Image Processing, Signal Processing and Embedded Systems.She is life member of IETE since 2008.

References (32)

S. Makaju et al.
Lung cancer detection using CT scan images
Procedia Comput. Sci
(2018)
S. Uzelaltinbulat et al.
Lung tumor segmentation algorithm
Procedia Comput. Sci.
(2017)
J. John et al.
Multilevel thresholding based segmentation and feature extraction for pulmonary nodule detection
Procedia Technol.
(2016)
G. Kasinathan et al.
Automated 3-D lung tumor detection and classification by an active contour model and CNN classifier
Expert Syst. Appl.
(2019)
S. Sivakumar et al.
Lung nodule segmentation through unsupervised clustering models
Procedia Eng.
(2012)
Y.F. Riti et al.
Feature extraction for lesion margin characteristic classification from CT Scan lungs image
S.C. Park et al.
A multistage approach to improve performance of computer-aided detection of pulmonary embolisms depicted on CT images: preliminary investigation
IEEE Trans. Biomed. Eng.
(2010)
E.M. Van Rikxoort et al.
Automatic segmentation of pulmonary segments from volumetric chest CT scans
IEEE Trans. Med. Imaging
(2009)
S.M.B. Netto et al.
Automatic segmentation of lung nodules with growing neural gas and support vector machine
Comput. Biol. Med.
(2012)
C. Liu et al.
Lung segmentation based on random forest and multi-scale edge detection
IET Image Process.
(2019)

Q. Wang et al.

Multiscale rotation-invariant convolutional neural networks for lung texture classification

IEEE J. Biomed. Health Inform.

(2017)

W. Zuo et al.

Multi-resolution CNN and knowledge transfer for candidate classification in lung nodule detection

IEEE Access

(2019)

N. Emaminejad et al.

Fusion of quantitative image and genomic biomarkers to improve prognosis assessment of early stage lung cancer patients

IEEE Trans. Biomed. Eng.

(2015)

S. Qiu et al.

Lung nodules detection in CT images using Gestalt-based algorithm

Chin. J. Electron.

(2016)

S. Potghan et al.

Multi-layer perceptron based lung tumor classification

R. Manickavasagam et al.

GACM based segmentation method for Lung nodule detection and classification of stages using CT images

Cited by (18)

Enhanced Elman spike Neural network optimized with flamingo search optimization algorithm espoused lung cancer classification from CT images
2023, Biomedical Signal Processing and Control
At present, researchers have been try to enhance the CAD system performance utilizing deep learning techniques in lung cancer screening and computed tomography (CT), but none of them attain the adequate accuracy. Therefore, an Enhanced Elman Spike Neural Network optimized with flamingo search optimization algorithm espoused Lung cancer classification from CT images (EESNN-FSOA-LCC) is proposed in this article. Initially, the input lung CT images are gathered through IQ-OTH/NCCD Lung Cancer Dataset. Then the input lung CT images are pre-processed using anisotropic diffusion Kuwahara filtering. These pre-processed lung CT images are fed to Hesitant Fuzzy Linguistic Bi-objective Clustering process for ROI region of lung cancer. Then the significant features present under ROI region of lung cancer segmentation are clarified through the help of Gray level co-occurrence matrix (GLCM) window adaptive approach. The extracted features are presented to EESNN for categorizing lung CT images image as normal, Benign, and Malignant. In general, EESNN classifier not divulges any adaption of optimization methods for determining the optimum parameters and to assure exact classification of lung cancer. Hence, a flamingo search Optimization Algorithm is proposed to optimize the EESNN classifier, which precisely classifies the lung cancer. The proposed EESNN-FSOA-LCC approach is activated in python. The proposed EESNN-FSOA-LCC approach achieves 38.58%, 25.69% and 43.87%, high accuracy when comparing to the existing CTI-LCC-SVM, CTI-LCC-GoogleNet-DNN and CTI-LCC-CNN respectively.
Segmentation of medical images using an attention embedded lightweight network
2022, Engineering Applications of Artificial Intelligence
Citation Excerpt :
For instance, the threshold-based methods include the Otsu image segmentation (Satapathy et al., 2018), the histogram statistics-based approaches (Cheng et al., 2016), fixed-threshold algorithms (Zhang et al., 2019), among others, and the edge detection-based segmentation methods have also been introduced to detect the peripheries of objects using diverse operators like SIFT (Hannane et al., 2016), Canny (Punarselvam and Suresh, 2019), and Harris (Liu et al., 2019). In particular, the machine learning (ML) methods, such as artificial neural networks (ANNs) (Wang et al., 2013; Chen et al., 2019), support vector machine (SVM) (Priya and Venkatesan, 2021), and K-nearest neighbor (KNN) (Sumathi et al., 2021), have been proposed for image segmentation. Although some impressive results have been reported in the literature, the abovementioned traditional methods have some obvious shortcomings such as the reliance on hand-crafted features, staged computation process, low robustness and stability, etc.
Accurate segmentation of computerized tomography (CT) images is of great significance to clinical diagnosis. However, because of the high similarity of gray values, it is a challenging task for CT image segmentation. The encoder and decoder based CNN architecture has greatly improved the segmentation effect, but it also encounters a bottleneck due to the information loss in the encoding process. In view of this, we proposed an image segmentation model based on a novel network architecture for medical image segmentation. To improve the efficiency and decrease the number of model parameters, we optimized the Inception module by substituting the depth-wise separable convolutions (DWSC) for the standard convolutions. Then, the optimized Inception module paired with the residual network was chosen as the backbone extractor to extract high-quality image features. Further, a hybrid attention mechanism, which consists of channel-wise and spatial attention, was incorporated into the network to realize the maximum reuse of inter-channel relationships and spatial point characteristics. In particular, the attention module was separately embedded into the contracting and expansive paths to enhance the feature extraction capability and detail restoration effects. The experimental indicators were significantly improved on the test dataset, and the intersection over union (IoU) of the proposed method reached no less than 0.9645, 0.6499, and 0.7945 on the Lung, Colon tumor, and DRIVE datasets, respectively, which demonstrated the effectiveness of the proposed method. Our code and data are available at https://github.com/xtu502/medical-image-segmentation/.
Exemplar deep and hand-modeled features based automated and accurate cerebral hemorrhage classification method
2022, Medical Engineering and Physics
Citation Excerpt :
Early detection of TBI causes earlier intervention in patients, leading to a decrease in mortality and morbidity. Computer-aided medical image classification is widely used for different purposes [9–12]. Today, many methods are used to classify CT images.
: Cerebral hemorrhage (CH) is a commonly seen disease, and an accurate diagnosis of the type of CH is a very crucial step in treatment. Therefore, CH requires a prompt and accurate diagnosis. To simplify this process, an accurate CH classification model is presented using a machine learning technique.
: A computed tomography (CT) image dataset was collected retrospectively in this research. This dataset contains 9818 images with five categories. An exemplar fused feature generator is presented to classify these features. This generator uses pre-trained AlexNet, local binary pattern (LBP), and local phase quantization (LPQ). The neighborhood component analysis (NCA) method selects the top features, and the chosen feature vector is classified on the support vector machine.
: Six validation methods are utilized to calculate the performance of the presented exemplar fused features and NCA-based CH classification model. This model attained 97.47%, 96.05%, 95.21%, 93.62%, 91.28% and 96.34% accuracies using five hold-out validations and ten-fold cross-validation respectively.
: The calculated results clearly demonstrate the success and robustness of the introduced exemplar fused feature generation and NCA-based model. Furthermore, this model can be used in emergency services to overcome a prompt diagnosis of CH.
An Efficient Image Classification of Lung Nodule Classification Approach Using CT and PET Fused Images
2024, AIP Conference Proceedings
The effect of harmonization on the variability of PET radiomic features extracted using various segmentation methods
2024, Annals of Nuclear Medicine
Abnormality Classification in PET Images of Prostate Tumour using Neural Network Methods
2024, International Journal of Intelligent Systems and Applications in Engineering

View all citing articles on Scopus

DR. P. Venkatesan, B.E.,(ECE).,ME-Power Electronics & drives, PhD, Associate Professor, Department of Electronics Communication Eenginering, Sri Chandrasekharendra Saraswathi Viswamahavidyalaya, Kanchipuram. His Area Of Interest Lies In Signal Processing, Digital Image Processing, Embedded Systems, Neural Networks. He has a more than 20 year of teaching and research experience.He has participated in International and national conferences. He has a good number of publications in reputed journals. He has guided many Ph.D Scholars and also guiding many good numbers of Ph.D scholars

View full text

An efficient image segmentation and classification of lung lesions in pet and CT image fusion using DTWT incorporated SVM

Abstract

Introduction

Section snippets

Literature survey

Methods and materials

Results and discussion

Conclusion

Declaration of Competing Interest

Procedia Comput. Sci

Procedia Comput. Sci.

Procedia Technol.

Expert Syst. Appl.

Procedia Eng.

Feature extraction for lesion margin characteristic classification from CT Scan lungs image

A multistage approach to improve performance of computer-aided detection of pulmonary embolisms depicted on CT images: preliminary investigation

IEEE Trans. Biomed. Eng.

Automatic segmentation of pulmonary segments from volumetric chest CT scans

IEEE Trans. Med. Imaging

Automatic segmentation of lung nodules with growing neural gas and support vector machine

Comput. Biol. Med.

Lung segmentation based on random forest and multi-scale edge detection

IET Image Process.

Multiscale rotation-invariant convolutional neural networks for lung texture classification

IEEE J. Biomed. Health Inform.

Multi-resolution CNN and knowledge transfer for candidate classification in lung nodule detection

IEEE Access

Fusion of quantitative image and genomic biomarkers to improve prognosis assessment of early stage lung cancer patients

IEEE Trans. Biomed. Eng.

Lung nodules detection in CT images using Gestalt-based algorithm

Chin. J. Electron.

Multi-layer perceptron based lung tumor classification

GACM based segmentation method for Lung nodule detection and classification of stages using CT images