Histopathological images are widely used to diagnose diseases including skin cancer. As digital histopathological images are typically of very large size, in the order of several billion pixels, automated identification of all abnormal cell nuclei and their distribution within multiple tissue sections would assist
rapid comprehensive diagnostic assessment. In this paper, we propose a technique, using deep learning algorithms, to segment the cell nuclei in Hematoxylin and Eosin (H&E) stained images and detect the abnormal melanocytes within histopathological images. The Nuclear segmentation is done by using a Convolutional Neural Network (CNN) and hand-crafted features are extracted for each nucleus. The segmented nuclei are then classified into normal and abnormal nuclei using a Support Vector Machine classifier. Experimental results show that the CNN can segment the nuclei with more than 90% accuracy. The proposed technique has a low computational complexity.
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...ijcseit
The research work presented in this paper is to achieve the tissue classification and automatically
diagnosis the abnormal tumor region present in Computed Tomography (CT) images using the wavelet
based statistical texture analysis method. Comparative studies of texture analysis method are performed
for the proposed wavelet based texture analysis method and Spatial Gray Level Dependence Method
(SGLDM). Our proposed system consists of four phases i) Discrete Wavelet Decomposition (ii)
Feature extraction (iii) Feature selection (iv) Analysis of extracted texture features by classifier. A
wavelet based statistical texture feature set is derived from normal and tumor regions. Genetic Algorithm
(GA) is used to select the optimal texture features from the set of extracted texture features. We construct
the Support Vector Machine (SVM) based classifier and evaluate the performance of classifier by
comparing the classification results of the SVM based classifier with the Back Propagation Neural network
classifier(BPN). The results of Support Vector Machine (SVM), BPN classifiers for the texture analysis
methods are evaluated using Receiver Operating Characteristic (ROC) analysis. Experimental results
show that the classification accuracy of SVM is 96% for 10 fold cross validation method. The system
has been tested with a number of real Computed Tomography brain images and has achieved satisfactory
results.
Intelligent algorithms for cell tracking and image segmentationijcsit
Sensitive and accurate cell tracking system is important to cell motility studies. Recently, researchers have
developed several methods for detecting and tracking the living cells. To improve the living cells tracking
systems performance and accuracy, we focused on developing a novel technique for image processing. The
algorithm we propose presents novel image segmentation and tracking system technique to incorporate the
advantages of both Topological Alignments and snakes for more accurate tracking approach. The results
demonstrate that the proposed algorithm achieves accurate tracking for detecting and analyzing the
mobility of the living cells. The RMSE between the manual and the computed displacement was less than
12% on average. Where the Active Contour method gave a velocity RMSE of less than 11%, improves to
less than 8% by using the novel Algorithm. We have achieved better tracking and detecting for the cells,
also the ability of the system to improve the low contrast, under and over segmentation which is the most
cell tracking challenge problems and responsible for lacking accuracy in cell tracking techniques.
Microarray Data Classification Using Support Vector MachineCSCJournals
DNA microarrays allow biologist to measure the expression of thousands of genes simultaneously on a small chip. These microarrays generate huge amount of data and new methods are needed to analyse them. In this paper, a new classification method based on support vector machine is proposed. The proposed method is used to classify gene expression data recorded on DNA microarrays. It is found that the proposed method is faster than neural network and the classification performance is not less than neural network.
Efficient Small Template Iris Recognition System Using Wavelet TransformCSCJournals
Iris recognition is known as an inherently reliable biometric technique for human identification. Feature extraction is a crucial step in iris recognition, and the trend nowadays is to reduce the size of the extracted features. Special efforts have been applied in order to obtain low templates size and fast verification algorithms. These efforts are intended to enable a human authentication in small embedded systems, such as an Integrated Circuit smart card. In this paper, an effective eyelids removing method, based on masking the iris, has been applied. Moreover, an efficient iris recognition encoding algorithm has been employed. Different combination of wavelet coefficients which quantized with multiple quantization levels are used and the best wavelet coefficients and quantization levels are determined. The system is based on an empirical analysis of CASIA iris database images. Experimental results show that this algorithm is efficient and gives promising results of False Accept Ratio (FAR) = 0% and False Reject Ratio (FRR) = 1% with a template size of only 364 bits.
Retinal Macular Edema Detection Using Optical Coherence Tomography ImagesIOSRJVSP
Macular Edema affects around 20 million people of the world each year. Optical Coherence Tomography (OCT), a non-invasive eye-imaging modality, is capable of detecting Macular Edema both in its early and advanced stages. In this paper, an algorithm which detects Macular Edema from OCT images has been presented. Initially the images are filtered to de-noise them. Then, the retinal layers - Inner Limiting Membrane (ILM) and Retinal Pigment Epithelium (RPE) are segmented using Graph Theory method. Region splitting is performed on the OCT scan and the thickness between the two layers in the different regions are determined. Area enclosed between the two layers is also estimated. Support Vector Machine, a binary classifier is used to draw a classification between normal and abnormal OCT scans. Region-wise thickness, a few Haralick’s features, area between ILM and RPE and a few wavelet features are used to train the classifier. The classifier yielded an accuracy of 95% and a sensitivity of 100%. Thus, this algorithm can be used by ophthalmologists in early detection of Macular Edema.
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...ijcseit
The research work presented in this paper is to achieve the tissue classification and automatically
diagnosis the abnormal tumor region present in Computed Tomography (CT) images using the wavelet
based statistical texture analysis method. Comparative studies of texture analysis method are performed
for the proposed wavelet based texture analysis method and Spatial Gray Level Dependence Method
(SGLDM). Our proposed system consists of four phases i) Discrete Wavelet Decomposition (ii)
Feature extraction (iii) Feature selection (iv) Analysis of extracted texture features by classifier. A
wavelet based statistical texture feature set is derived from normal and tumor regions. Genetic Algorithm
(GA) is used to select the optimal texture features from the set of extracted texture features. We construct
the Support Vector Machine (SVM) based classifier and evaluate the performance of classifier by
comparing the classification results of the SVM based classifier with the Back Propagation Neural network
classifier(BPN). The results of Support Vector Machine (SVM), BPN classifiers for the texture analysis
methods are evaluated using Receiver Operating Characteristic (ROC) analysis. Experimental results
show that the classification accuracy of SVM is 96% for 10 fold cross validation method. The system
has been tested with a number of real Computed Tomography brain images and has achieved satisfactory
results.
Intelligent algorithms for cell tracking and image segmentationijcsit
Sensitive and accurate cell tracking system is important to cell motility studies. Recently, researchers have
developed several methods for detecting and tracking the living cells. To improve the living cells tracking
systems performance and accuracy, we focused on developing a novel technique for image processing. The
algorithm we propose presents novel image segmentation and tracking system technique to incorporate the
advantages of both Topological Alignments and snakes for more accurate tracking approach. The results
demonstrate that the proposed algorithm achieves accurate tracking for detecting and analyzing the
mobility of the living cells. The RMSE between the manual and the computed displacement was less than
12% on average. Where the Active Contour method gave a velocity RMSE of less than 11%, improves to
less than 8% by using the novel Algorithm. We have achieved better tracking and detecting for the cells,
also the ability of the system to improve the low contrast, under and over segmentation which is the most
cell tracking challenge problems and responsible for lacking accuracy in cell tracking techniques.
Microarray Data Classification Using Support Vector MachineCSCJournals
DNA microarrays allow biologist to measure the expression of thousands of genes simultaneously on a small chip. These microarrays generate huge amount of data and new methods are needed to analyse them. In this paper, a new classification method based on support vector machine is proposed. The proposed method is used to classify gene expression data recorded on DNA microarrays. It is found that the proposed method is faster than neural network and the classification performance is not less than neural network.
Efficient Small Template Iris Recognition System Using Wavelet TransformCSCJournals
Iris recognition is known as an inherently reliable biometric technique for human identification. Feature extraction is a crucial step in iris recognition, and the trend nowadays is to reduce the size of the extracted features. Special efforts have been applied in order to obtain low templates size and fast verification algorithms. These efforts are intended to enable a human authentication in small embedded systems, such as an Integrated Circuit smart card. In this paper, an effective eyelids removing method, based on masking the iris, has been applied. Moreover, an efficient iris recognition encoding algorithm has been employed. Different combination of wavelet coefficients which quantized with multiple quantization levels are used and the best wavelet coefficients and quantization levels are determined. The system is based on an empirical analysis of CASIA iris database images. Experimental results show that this algorithm is efficient and gives promising results of False Accept Ratio (FAR) = 0% and False Reject Ratio (FRR) = 1% with a template size of only 364 bits.
Retinal Macular Edema Detection Using Optical Coherence Tomography ImagesIOSRJVSP
Macular Edema affects around 20 million people of the world each year. Optical Coherence Tomography (OCT), a non-invasive eye-imaging modality, is capable of detecting Macular Edema both in its early and advanced stages. In this paper, an algorithm which detects Macular Edema from OCT images has been presented. Initially the images are filtered to de-noise them. Then, the retinal layers - Inner Limiting Membrane (ILM) and Retinal Pigment Epithelium (RPE) are segmented using Graph Theory method. Region splitting is performed on the OCT scan and the thickness between the two layers in the different regions are determined. Area enclosed between the two layers is also estimated. Support Vector Machine, a binary classifier is used to draw a classification between normal and abnormal OCT scans. Region-wise thickness, a few Haralick’s features, area between ILM and RPE and a few wavelet features are used to train the classifier. The classifier yielded an accuracy of 95% and a sensitivity of 100%. Thus, this algorithm can be used by ophthalmologists in early detection of Macular Edema.
CLASSIFICATION OF OCT IMAGES FOR DETECTING DIABETIC RETINOPATHY DISEASE USING...sipij
Optical Coherence Tomography (OCT) imaging aids in retinal abnormality detection by showing the
tomographic retinal layers. OCT images are a useful tool for detecting Diabetic Retinopathy (DR) disease
because of their capability to capture micrometer-resolution. An automated technique was introduced to
differentiate DR images from normal ones. 214 images were subjected to the experiment, of which 160
images were used for classifiers’ training, and 54 images were used for testing. Different features were
extracted to feed our classifiers, including statistical features and local binary pattern (LBP) features. The
experimental results demonstrated that our classifiers were able to discriminate DR retina from the normal
retina with Area Under the Receiver Operating Characteristic (ROC) Curve (AUC) of 100%. The retinal
OCT images have common texture patterns and using a powerful tool for pattern analysis like LBP
features has a significant impact on the achieved results. The result has better performance than previously
proposed methods in the literature.
During past few years, brain tumor segmentation in CT has become an emergent research area in the field of medical imaging system. Brain tumor detection helps in finding the exact size and location of tumor. An efficient algorithm is proposed in this project for tumor detection based on segmentation and morphological operators. Firstly quality of scanned image is enhanced and then morphological operators are applied to detect the tumor in the scanned image. The problem with biopsy is that the patient has to be hospitalized and also the results (around 15%) give false negative. Scan images are read by radiologist but it's a subjective analysis which requires more experience. In the proposed work we segment the renal region and then classify the tumors as benign or malignant by using ANFIS, which is a non-invasive automated process. This approach reduces the waiting time of the patient.
A COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS FOR EEG SIGNAL CLASSIFICATIONsipij
In this paper, different machine learning algorithms such as Linear Discriminant Analysis, Support vector
machine (SVM), Multi-layer perceptron, Random forest, K-nearest neighbour, and Autoencoder with SVM
have been compared. This comparison was conducted to seek a robust method that would produce good
classification accuracy. To this end, a robust method of classifying raw Electroencephalography (EEG)
signals associated with imagined movement of the right hand and relaxation state, namely Autoencoder
with SVM has been proposed. The EEG dataset used in this research was created by the University of
Tubingen, Germany. The best classification accuracy achieved was 70.4% with SVM through feature
engineering. However, our prosed method of autoencoder in combination with SVM produced a similar
accuracy of 65% without using any feature engineering technique. This research shows that this system of
classification of motor movements can be used in a Brain-Computer Interface system (BCI) to mentally
control a robotic device or an exoskeleton.
A Wavelet Based Automatic Segmentation of Brain Tumor in CT Images Using Opti...CSCJournals
This paper presents an automated segmentation of brain tumors in computed tomography images (CT) using combination of Wavelet Statistical Texture features (WST) obtained from 2-level Discrete Wavelet Transformed (DWT) low and high frequency sub bands and Wavelet Co-occurrence Texture features (WCT) obtained from two level Discrete Wavelet Transformed (DWT) high frequency sub bands. In the proposed method, the wavelet based optimal texture features that distinguish between the brain tissue, benign tumor and malignant tumor tissue is found. Comparative studies of texture analysis is performed for the proposed combined wavelet based texture analysis method and Spatial Gray Level Dependence Method (SGLDM). Our proposed system consists of four phases i) Discrete Wavelet Decomposition (ii) Feature extraction (iii) Feature selection (iv) Classification and evaluation. The combined Wavelet Statistical Texture feature set (WST) and Wavelet Co-occurrence Texture feature (WCT) sets are derived from normal and tumor regions. Feature selection is performed by Genetic Algorithm (GA). These optimal features are used to segment the tumor. An Probabilistic Neural Network (PNN) classifier is employed to evaluate the performance of these features and by comparing the classification results of the PNN classifier with the Feed Forward Neural Network classifier(FFNN).The results of the Probabilistic Neural Network, FFNN classifiers for the texture analysis methods are evaluated using Receiver Operating Characteristic (ROC) analysis. The performance of the algorithm is evaluated on a series of brain tumor images. The results illustrate that the proposed method outperforms the existing methods.
Segmentation of Brain MR Images for Tumor Extraction by Combining Kmeans Clus...CSCJournals
Segmentation of images holds an important position in the area of image processing. It becomes more important while typically dealing with medical images where pre-surgery and post surgery decisions are required for the purpose of initiating and speeding up the recovery process [5] Computer aided detection of abnormal growth of tissues is primarily motivated by the necessity of achieving maximum possible accuracy. Manual segmentation of these abnormal tissues cannot be compared with modern day’s high speed computing machines which enable us to visually observe the volume and location of unwanted tissues. A well known segmentation problem within MRI is the task of labeling voxels according to their tissue type which include White Matter (WM), Grey Matter (GM) , Cerebrospinal Fluid (CSF) and sometimes pathological tissues like tumor etc. This paper describes an efficient method for automatic brain tumor segmentation for the extraction of tumor tissues from MR images. It combines Perona and Malik anisotropic diffusion model for image enhancement and Kmeans clustering technique for grouping tissues belonging to a specific group. The proposed method uses T1, T2 and PD weighted gray level intensity images. The proposed technique produced appreciative results
Brain tumor classification using artificial neural network on mri imageseSAT Journals
Abstract
In this paper, an attempt has been made to summarize the multi-resolution transformation and the different classifiers useful to
analyze the brain tumor using MRI. X-ray, MRI, Ultrasound etc. are different techniques used to scan brain tumor images.
Radiologist prefers MRI to get detail information about tumor to help him diagnoses. In this paper we have used MRI of brain
tumor for analysis. We have used Digital image processing tool for detection of the tumor. The identification, detection and
classification of brain tumor have been done by extracting features from MRI with the help of wavelet transformation. The MRI of
brain tumor is classified into two categories normal and abnormal brain. In this work Digital image processing has been used as
a tool for getting clear and exact details about tumor in earlier stages. This helps the physicians and practitioners for diagnoses.
Key word – Brain tumor, Wavelet transform, segmentation.
Automatic Detection of Malaria Parasites for Estimating ParasitemiaCSCJournals
Malaria parasitemia is a measurement of the amount of Malaria parasites in the patient's blood and an indicator for the degree of infection. In this paper an automatic technique is proposed for Malaria parasites detection from blood images by extracting red blood cells (RBCs) from blood image and classifying as normal or parasite infected. Manual counting of parasitemia is tedious and time consuming and need experts. Proposed automatic approach is used Otsu thresholding on gray image and green channel of the blood image for cell segmentation, watershed transform is used for separation of touching cells, color and statistical features are extracted from segmented cells and SVM binary classifier is used for classification of normal and parasite infected cells.
DETECTING BRAIN TUMOUR FROM MRI IMAGE USING MATLAB GUI PROGRAMMEIJCSES Journal
Engineers have been actively developing tools to detect tumors and to process medical images. Medical image segmentation is a powerful tool that is often used to detect tumors. Many scientists and researchers are working to develop and add more features to this tool. This project is about detecting Brain tumors from MRI images using an interface of GUI in Matlab. Using the GUI, this program can use various combinations of segmentation, filters, and other image processing algorithms to achieve the best results.
We start with filtering the image using Prewitt horizontal edge-emphasizing filter. The next step for detecting tumor is "watershed pixels." The most important part of this project is that all the Matlab programs work with GUI “Matlab guide”. This allows us to use various combinations of filters, and other
image processing techniques to arrive at the best result that can help us detect brain tumors in their early stages.
Segmentation and Automatic Counting of Red Blood Cells Using Hough TransformIJARBEST JOURNAL
Authors:- V. Antony Asir Daniel1, J. Surendiran2, K. Kalaiselvi3
Abstract- Red blood cells are specialized as oxygen carrier RBC plays a crucial role in
medical diagnosis and pathological study. The blood samples are collected using the smear
glass slide. These samples are taken under the test using the image of the blood. Filtering
process are carries out to remove the noise. Morphological operation are applied on the
blood image and using Hough transform method the RBC are counted which is the
effective segmentation process.
Abstract:
A technique for exudate detectionin fundus image is been presented in this paper. Due to diabetic retinopathy an abnormality is caused known as exudates.The loss of vision can be prevented by detecting the exudates as early as possible. The work mainly aims at detecting exudates which is present in the green channel of the RGB image by applying few preprocessing steps, DWT and feature extraction. The extracted features are fed to 3 different classifiers such as KNN, SVM and NN. Based on the classifier result if an exudate is present the extraction of exudate ROI is done based on canny edge detection followed by morphological operations. The severity of the exudates is established on the area of the detected exudate.
Keywords:Exudates, Fundus image, Diabetic retinopathy, DWT, KNN, SVM, NN, Canny edge detection, Morphological operations.
International Journal of Computational Engineering Research(IJCER) is an intentional online Journal in English monthly publishing journal. This Journal publish original research work that contributes significantly to further the scientific knowledge in engineering and Technology
This paper proposes the development of a software that performs the pre-diagnosis of malignant melanoma, spincellular carcinoma and basal-cell carcinoma. The software is divided into five modules, these being: digital imaging, analysis and processing, storage, feature extraction and classification by means of an Artificial Neural Network (ANN). The results shown the performance of the software for two different combination of activation functions in the network. With the use of spectroscopic techniques for the acquisition of images and the combination of non-linear and linear activation functions in the ANN, the software shows an effectiveness greater than 80%, concluding that it can be an effective tool as an aid in the diagnosis of cancer of skin.
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...Editor IJCATR
Manual classification of brain tumor is time devastating and bestows ambiguous results. Automatic image classification is
emergent thriving research area in medical field. In the proposed methodology, features are extracted from raw images which are then
fed to ANFIS (Artificial neural fuzzy inference system).ANFIS being neuro-fuzzy system harness power of both hence it proves to be
a sophisticated framework for multiobject classification. A comprehensive feature set and fuzzy rules are selected to classify an
abnormal image to the corresponding tumor type. This proposed technique is fast in execution, efficient in classification and easy in
implementation.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology.
SEGMENTATION OF LUNG GLANDULAR CELLS USING MULTIPLE COLOR SPACESIJCSEA Journal
Early detection of lung cancer is a challenging problem, the world faces today. Prior to classify glandular cells as malignant or benign a reliable segmentation technique is required. In this paper we present a novel lung glandular cell segmentation technique. The technique uses a combination of multiple color spaces and various clustering algorithms to automatically find the best possible segmentation result. Unsupervised clustering methods of K-means and Fuzzy C-means were used on multiple color spaces such as HSV, LAB, LUV, xyY. Experimental results of segmentation using various color spaces are provided to show the performance of the proposed system.
CONVOLUTIONAL NEURAL NETWORK BASED RETINAL VESSEL SEGMENTATIONCSEIJJournal
In human eye, the state of the blood vessel is a crucial diagnostic factor. The segmentation of blood vessel
from the fundus image is difficult due to the spatial complexity, adjacency, overlapping and variability of
blood vessel. The detection of ophthalmic pathologies like hypertensive disorders, diabetic retinopathy and
cardiovascular diseases are remain challenging task due to the wide-ranging distribution of blood vessels.
In this paper, Stacked Autoencoder and CNN (Convolutional Neural Network) technique is proposed to
extract the blood vessel from the fundus image. Based on the experiments conducted using the Stacked
Autoencoder and Convolutional Neural Network gives 90% & 95% accuracy for segmentation.
CLASSIFICATION OF OCT IMAGES FOR DETECTING DIABETIC RETINOPATHY DISEASE USING...sipij
Optical Coherence Tomography (OCT) imaging aids in retinal abnormality detection by showing the
tomographic retinal layers. OCT images are a useful tool for detecting Diabetic Retinopathy (DR) disease
because of their capability to capture micrometer-resolution. An automated technique was introduced to
differentiate DR images from normal ones. 214 images were subjected to the experiment, of which 160
images were used for classifiers’ training, and 54 images were used for testing. Different features were
extracted to feed our classifiers, including statistical features and local binary pattern (LBP) features. The
experimental results demonstrated that our classifiers were able to discriminate DR retina from the normal
retina with Area Under the Receiver Operating Characteristic (ROC) Curve (AUC) of 100%. The retinal
OCT images have common texture patterns and using a powerful tool for pattern analysis like LBP
features has a significant impact on the achieved results. The result has better performance than previously
proposed methods in the literature.
During past few years, brain tumor segmentation in CT has become an emergent research area in the field of medical imaging system. Brain tumor detection helps in finding the exact size and location of tumor. An efficient algorithm is proposed in this project for tumor detection based on segmentation and morphological operators. Firstly quality of scanned image is enhanced and then morphological operators are applied to detect the tumor in the scanned image. The problem with biopsy is that the patient has to be hospitalized and also the results (around 15%) give false negative. Scan images are read by radiologist but it's a subjective analysis which requires more experience. In the proposed work we segment the renal region and then classify the tumors as benign or malignant by using ANFIS, which is a non-invasive automated process. This approach reduces the waiting time of the patient.
A COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS FOR EEG SIGNAL CLASSIFICATIONsipij
In this paper, different machine learning algorithms such as Linear Discriminant Analysis, Support vector
machine (SVM), Multi-layer perceptron, Random forest, K-nearest neighbour, and Autoencoder with SVM
have been compared. This comparison was conducted to seek a robust method that would produce good
classification accuracy. To this end, a robust method of classifying raw Electroencephalography (EEG)
signals associated with imagined movement of the right hand and relaxation state, namely Autoencoder
with SVM has been proposed. The EEG dataset used in this research was created by the University of
Tubingen, Germany. The best classification accuracy achieved was 70.4% with SVM through feature
engineering. However, our prosed method of autoencoder in combination with SVM produced a similar
accuracy of 65% without using any feature engineering technique. This research shows that this system of
classification of motor movements can be used in a Brain-Computer Interface system (BCI) to mentally
control a robotic device or an exoskeleton.
A Wavelet Based Automatic Segmentation of Brain Tumor in CT Images Using Opti...CSCJournals
This paper presents an automated segmentation of brain tumors in computed tomography images (CT) using combination of Wavelet Statistical Texture features (WST) obtained from 2-level Discrete Wavelet Transformed (DWT) low and high frequency sub bands and Wavelet Co-occurrence Texture features (WCT) obtained from two level Discrete Wavelet Transformed (DWT) high frequency sub bands. In the proposed method, the wavelet based optimal texture features that distinguish between the brain tissue, benign tumor and malignant tumor tissue is found. Comparative studies of texture analysis is performed for the proposed combined wavelet based texture analysis method and Spatial Gray Level Dependence Method (SGLDM). Our proposed system consists of four phases i) Discrete Wavelet Decomposition (ii) Feature extraction (iii) Feature selection (iv) Classification and evaluation. The combined Wavelet Statistical Texture feature set (WST) and Wavelet Co-occurrence Texture feature (WCT) sets are derived from normal and tumor regions. Feature selection is performed by Genetic Algorithm (GA). These optimal features are used to segment the tumor. An Probabilistic Neural Network (PNN) classifier is employed to evaluate the performance of these features and by comparing the classification results of the PNN classifier with the Feed Forward Neural Network classifier(FFNN).The results of the Probabilistic Neural Network, FFNN classifiers for the texture analysis methods are evaluated using Receiver Operating Characteristic (ROC) analysis. The performance of the algorithm is evaluated on a series of brain tumor images. The results illustrate that the proposed method outperforms the existing methods.
Segmentation of Brain MR Images for Tumor Extraction by Combining Kmeans Clus...CSCJournals
Segmentation of images holds an important position in the area of image processing. It becomes more important while typically dealing with medical images where pre-surgery and post surgery decisions are required for the purpose of initiating and speeding up the recovery process [5] Computer aided detection of abnormal growth of tissues is primarily motivated by the necessity of achieving maximum possible accuracy. Manual segmentation of these abnormal tissues cannot be compared with modern day’s high speed computing machines which enable us to visually observe the volume and location of unwanted tissues. A well known segmentation problem within MRI is the task of labeling voxels according to their tissue type which include White Matter (WM), Grey Matter (GM) , Cerebrospinal Fluid (CSF) and sometimes pathological tissues like tumor etc. This paper describes an efficient method for automatic brain tumor segmentation for the extraction of tumor tissues from MR images. It combines Perona and Malik anisotropic diffusion model for image enhancement and Kmeans clustering technique for grouping tissues belonging to a specific group. The proposed method uses T1, T2 and PD weighted gray level intensity images. The proposed technique produced appreciative results
Brain tumor classification using artificial neural network on mri imageseSAT Journals
Abstract
In this paper, an attempt has been made to summarize the multi-resolution transformation and the different classifiers useful to
analyze the brain tumor using MRI. X-ray, MRI, Ultrasound etc. are different techniques used to scan brain tumor images.
Radiologist prefers MRI to get detail information about tumor to help him diagnoses. In this paper we have used MRI of brain
tumor for analysis. We have used Digital image processing tool for detection of the tumor. The identification, detection and
classification of brain tumor have been done by extracting features from MRI with the help of wavelet transformation. The MRI of
brain tumor is classified into two categories normal and abnormal brain. In this work Digital image processing has been used as
a tool for getting clear and exact details about tumor in earlier stages. This helps the physicians and practitioners for diagnoses.
Key word – Brain tumor, Wavelet transform, segmentation.
Automatic Detection of Malaria Parasites for Estimating ParasitemiaCSCJournals
Malaria parasitemia is a measurement of the amount of Malaria parasites in the patient's blood and an indicator for the degree of infection. In this paper an automatic technique is proposed for Malaria parasites detection from blood images by extracting red blood cells (RBCs) from blood image and classifying as normal or parasite infected. Manual counting of parasitemia is tedious and time consuming and need experts. Proposed automatic approach is used Otsu thresholding on gray image and green channel of the blood image for cell segmentation, watershed transform is used for separation of touching cells, color and statistical features are extracted from segmented cells and SVM binary classifier is used for classification of normal and parasite infected cells.
DETECTING BRAIN TUMOUR FROM MRI IMAGE USING MATLAB GUI PROGRAMMEIJCSES Journal
Engineers have been actively developing tools to detect tumors and to process medical images. Medical image segmentation is a powerful tool that is often used to detect tumors. Many scientists and researchers are working to develop and add more features to this tool. This project is about detecting Brain tumors from MRI images using an interface of GUI in Matlab. Using the GUI, this program can use various combinations of segmentation, filters, and other image processing algorithms to achieve the best results.
We start with filtering the image using Prewitt horizontal edge-emphasizing filter. The next step for detecting tumor is "watershed pixels." The most important part of this project is that all the Matlab programs work with GUI “Matlab guide”. This allows us to use various combinations of filters, and other
image processing techniques to arrive at the best result that can help us detect brain tumors in their early stages.
Segmentation and Automatic Counting of Red Blood Cells Using Hough TransformIJARBEST JOURNAL
Authors:- V. Antony Asir Daniel1, J. Surendiran2, K. Kalaiselvi3
Abstract- Red blood cells are specialized as oxygen carrier RBC plays a crucial role in
medical diagnosis and pathological study. The blood samples are collected using the smear
glass slide. These samples are taken under the test using the image of the blood. Filtering
process are carries out to remove the noise. Morphological operation are applied on the
blood image and using Hough transform method the RBC are counted which is the
effective segmentation process.
Abstract:
A technique for exudate detectionin fundus image is been presented in this paper. Due to diabetic retinopathy an abnormality is caused known as exudates.The loss of vision can be prevented by detecting the exudates as early as possible. The work mainly aims at detecting exudates which is present in the green channel of the RGB image by applying few preprocessing steps, DWT and feature extraction. The extracted features are fed to 3 different classifiers such as KNN, SVM and NN. Based on the classifier result if an exudate is present the extraction of exudate ROI is done based on canny edge detection followed by morphological operations. The severity of the exudates is established on the area of the detected exudate.
Keywords:Exudates, Fundus image, Diabetic retinopathy, DWT, KNN, SVM, NN, Canny edge detection, Morphological operations.
International Journal of Computational Engineering Research(IJCER) is an intentional online Journal in English monthly publishing journal. This Journal publish original research work that contributes significantly to further the scientific knowledge in engineering and Technology
This paper proposes the development of a software that performs the pre-diagnosis of malignant melanoma, spincellular carcinoma and basal-cell carcinoma. The software is divided into five modules, these being: digital imaging, analysis and processing, storage, feature extraction and classification by means of an Artificial Neural Network (ANN). The results shown the performance of the software for two different combination of activation functions in the network. With the use of spectroscopic techniques for the acquisition of images and the combination of non-linear and linear activation functions in the ANN, the software shows an effectiveness greater than 80%, concluding that it can be an effective tool as an aid in the diagnosis of cancer of skin.
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...Editor IJCATR
Manual classification of brain tumor is time devastating and bestows ambiguous results. Automatic image classification is
emergent thriving research area in medical field. In the proposed methodology, features are extracted from raw images which are then
fed to ANFIS (Artificial neural fuzzy inference system).ANFIS being neuro-fuzzy system harness power of both hence it proves to be
a sophisticated framework for multiobject classification. A comprehensive feature set and fuzzy rules are selected to classify an
abnormal image to the corresponding tumor type. This proposed technique is fast in execution, efficient in classification and easy in
implementation.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology.
SEGMENTATION OF LUNG GLANDULAR CELLS USING MULTIPLE COLOR SPACESIJCSEA Journal
Early detection of lung cancer is a challenging problem, the world faces today. Prior to classify glandular cells as malignant or benign a reliable segmentation technique is required. In this paper we present a novel lung glandular cell segmentation technique. The technique uses a combination of multiple color spaces and various clustering algorithms to automatically find the best possible segmentation result. Unsupervised clustering methods of K-means and Fuzzy C-means were used on multiple color spaces such as HSV, LAB, LUV, xyY. Experimental results of segmentation using various color spaces are provided to show the performance of the proposed system.
CONVOLUTIONAL NEURAL NETWORK BASED RETINAL VESSEL SEGMENTATIONCSEIJJournal
In human eye, the state of the blood vessel is a crucial diagnostic factor. The segmentation of blood vessel
from the fundus image is difficult due to the spatial complexity, adjacency, overlapping and variability of
blood vessel. The detection of ophthalmic pathologies like hypertensive disorders, diabetic retinopathy and
cardiovascular diseases are remain challenging task due to the wide-ranging distribution of blood vessels.
In this paper, Stacked Autoencoder and CNN (Convolutional Neural Network) technique is proposed to
extract the blood vessel from the fundus image. Based on the experiments conducted using the Stacked
Autoencoder and Convolutional Neural Network gives 90% & 95% accuracy for segmentation.
Convolutional Neural Network based Retinal Vessel SegmentationCSEIJJournal
In human eye, the state of the blood vessel is a crucial diagnostic factor. The segmentation of blood vessel
from the fundus image is difficult due to the spatial complexity, adjacency, overlapping and variability of
blood vessel. The detection of ophthalmic pathologies like hypertensive disorders, diabetic retinopathy and
cardiovascular diseases are remain challenging task due to the wide-ranging distribution of blood vessels.
In this paper, Stacked Autoencoder and CNN (Convolutional Neural Network) technique is proposed to
extract the blood vessel from the fundus image. Based on the experiments conducted using the Stacked
Autoencoder and Convolutional Neural Network gives 90% & 95% accuracy for segmentation.
Statistical Feature-based Neural Network Approach for the Detection of Lung C...CSCJournals
Lung cancer, if successfully detected at early stages, enables many treatment options, reduced risk of invasive surgery and increased survival rate. This paper presents a novel approach to detect lung cancer from raw chest X-ray images. At the first stage, we use a pipeline of image processing routines to remove noise and segment the lung from other anatomical structures in the chest X-ray and extract regions that exhibit shape characteristics of lung nodules. Subsequently, first and second order statistical texture features are considered as the inputs to train a neural network to verify whether a region extracted in the first stage is a nodule or not . The proposed approach detected nodules in the diseased area of the lung with an accuracy of 96% using the pixel-based technique while the feature-based technique produced an accuracy of 88%.
8 A Cellular Neural Network based system for cell counting in culture of biol...Cristian Randieri PhD
A Cellular Neural Network based system for cell counting in culture of biological cells - Proceedings of the 1998 IEEE International Conference on Control Applications, Trieste (Italy) 1-4 September 1998, Vol. 1, pp. 341-345.
di L. Bertucco, G. Nunnari, C. Randieri
Abstract
Cell counting methods are important tools in molecular biology as well as clinical medicine. It is not always technically possible to measure quantitatively the events of cellular growth and fission. When it can be done, the procedures are neither so simple nor without excessive tedium as to lend themselves practically to the necessary replication of observations with large number of individual cells. In this paper, we describe a CNN based system that uses a CNN simulator for counting cells. The performances of the proposed system are illustrated by a simple cell counting experiment using a Petroff- Hauser based counter system.
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...INFOGAIN PUBLICATION
Image fusion is the process of combining important information from two or more images into a single image. The resulting image will be more enhanced than any of the input pictures. The idea of combining multiple image modalities to furnish a single, more enhanced image is well established, special fusion methods have been proposed in literature. This paper is based on image fusion using laplacian pyramid and Discreet Wavelet Transform (DWT) methods. This system uses an easy and effective algorithm for multi-focus image fusion which uses fusion rules to create fused image. Subsequently, the fused image is obtained by applying inverse discreet wavelet transform. After fused image is obtained, watershed segmentation algorithm is applied to detect the tumor part in fused image.
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptxR&R Consult
CFD analysis is incredibly effective at solving mysteries and improving the performance of complex systems!
Here's a great example: At a large natural gas-fired power plant, where they use waste heat to generate steam and energy, they were puzzled that their boiler wasn't producing as much steam as expected.
R&R and Tetra Engineering Group Inc. were asked to solve the issue with reduced steam production.
An inspection had shown that a significant amount of hot flue gas was bypassing the boiler tubes, where the heat was supposed to be transferred.
R&R Consult conducted a CFD analysis, which revealed that 6.3% of the flue gas was bypassing the boiler tubes without transferring heat. The analysis also showed that the flue gas was instead being directed along the sides of the boiler and between the modules that were supposed to capture the heat. This was the cause of the reduced performance.
Based on our results, Tetra Engineering installed covering plates to reduce the bypass flow. This improved the boiler's performance and increased electricity production.
It is always satisfying when we can help solve complex challenges like this. Do your systems also need a check-up or optimization? Give us a call!
Work done in cooperation with James Malloy and David Moelling from Tetra Engineering.
More examples of our work https://www.r-r-consult.dk/en/cases-en/
Explore the innovative world of trenchless pipe repair with our comprehensive guide, "The Benefits and Techniques of Trenchless Pipe Repair." This document delves into the modern methods of repairing underground pipes without the need for extensive excavation, highlighting the numerous advantages and the latest techniques used in the industry.
Learn about the cost savings, reduced environmental impact, and minimal disruption associated with trenchless technology. Discover detailed explanations of popular techniques such as pipe bursting, cured-in-place pipe (CIPP) lining, and directional drilling. Understand how these methods can be applied to various types of infrastructure, from residential plumbing to large-scale municipal systems.
Ideal for homeowners, contractors, engineers, and anyone interested in modern plumbing solutions, this guide provides valuable insights into why trenchless pipe repair is becoming the preferred choice for pipe rehabilitation. Stay informed about the latest advancements and best practices in the field.
Immunizing Image Classifiers Against Localized Adversary Attacksgerogepatton
This paper addresses the vulnerability of deep learning models, particularly convolutional neural networks
(CNN)s, to adversarial attacks and presents a proactive training technique designed to counter them. We
introduce a novel volumization algorithm, which transforms 2D images into 3D volumetric representations.
When combined with 3D convolution and deep curriculum learning optimization (CLO), itsignificantly improves
the immunity of models against localized universal attacks by up to 40%. We evaluate our proposed approach
using contemporary CNN architectures and the modified Canadian Institute for Advanced Research (CIFAR-10
and CIFAR-100) and ImageNet Large Scale Visual Recognition Challenge (ILSVRC12) datasets, showcasing
accuracy improvements over previous techniques. The results indicate that the combination of the volumetric
input and curriculum learning holds significant promise for mitigating adversarial attacks without necessitating
adversary training.
Water scarcity is the lack of fresh water resources to meet the standard water demand. There are two type of water scarcity. One is physical. The other is economic water scarcity.
NO1 Uk best vashikaran specialist in delhi vashikaran baba near me online vas...Amil Baba Dawood bangali
Contact with Dawood Bhai Just call on +92322-6382012 and we'll help you. We'll solve all your problems within 12 to 24 hours and with 101% guarantee and with astrology systematic. If you want to take any personal or professional advice then also you can call us on +92322-6382012 , ONLINE LOVE PROBLEM & Other all types of Daily Life Problem's.Then CALL or WHATSAPP us on +92322-6382012 and Get all these problems solutions here by Amil Baba DAWOOD BANGALI
#vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore#blackmagicformarriage #aamilbaba #kalajadu #kalailam #taweez #wazifaexpert #jadumantar #vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore #blackmagicforlove #blackmagicformarriage #aamilbaba #kalajadu #kalailam #taweez #wazifaexpert #jadumantar #vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore #Amilbabainuk #amilbabainspain #amilbabaindubai #Amilbabainnorway #amilbabainkrachi #amilbabainlahore #amilbabaingujranwalan #amilbabainislamabad
Final project report on grocery store management system..pdfKamal Acharya
In today’s fast-changing business environment, it’s extremely important to be able to respond to client needs in the most effective and timely manner. If your customers wish to see your business online and have instant access to your products or services.
Online Grocery Store is an e-commerce website, which retails various grocery products. This project allows viewing various products available enables registered users to purchase desired products instantly using Paytm, UPI payment processor (Instant Pay) and also can place order by using Cash on Delivery (Pay Later) option. This project provides an easy access to Administrators and Managers to view orders placed using Pay Later and Instant Pay options.
In order to develop an e-commerce website, a number of Technologies must be studied and understood. These include multi-tiered architecture, server and client-side scripting techniques, implementation technologies, programming language (such as PHP, HTML, CSS, JavaScript) and MySQL relational databases. This is a project with the objective to develop a basic website where a consumer is provided with a shopping cart website and also to know about the technologies used to develop such a website.
This document will discuss each of the underlying technologies to create and implement an e- commerce website.
Final project report on grocery store management system..pdf
MELANOMA CELL DETECTION IN LYMPH NODES HISTOPATHOLOGICAL IMAGES USING DEEP LEARNING
1. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
DOI: 10.5121/sipij.2020.11401 1
MELANOMA CELL DETECTION IN LYMPH
NODES HISTOPATHOLOGICAL IMAGES
USING DEEP LEARNING
Salah Alheejawi, Richard Berendt, Naresh Jha and Mrinal Mandal
University of Alberta, Edmonton, Alberta, Canada
ABSTRACT
Histopathological images are widely used to diagnose diseases including skin cancer. As digital
histopathological images are typically of very large size, in the order of several billion pixels, automated
identification of all abnormal cell nuclei and their distribution within multiple tissue sections would assist
rapid comprehensive diagnostic assessment. In this paper, we propose a technique, using deep learning
algorithms, to segment the cell nuclei in Hematoxylin and Eosin (H&E) stained images and detect the
abnormal melanocytes within histopathological images. The Nuclear segmentation is done by using a
Convolutional Neural Network (CNN) and hand-crafted features are extracted for each nucleus. The
segmented nuclei are then classified into normal and abnormal nuclei using a Support Vector Machine
classifier. Experimental results show that the CNN can segment the nuclei with more than 90% accuracy.
The proposed technique has a low computational complexity.
KEYWORDS
Histopathological image analysis, Nuclear segmentation, Melanoma Detection, Deep learning.
1. INTRODUCTION
Primary cutaneous melanomas include a range of potentially lethal melanocytic neoplasms which
most often present in skin as an archetypical macular growth phase lesion comprised of an in-situ
component plus a papillary dermal component of similar cytomorphology. The diagnosis depends
upon the histomorphological identification of abnormal melanocytes forming radial and vertical
growth phase neoplastic cellular infiltrates which by invasion and widespread metastasis can
secondarily involve regional lymph nodes and ultimately any other part of the body. As per
recent statistics, it is estimated that about 100,350 new cases of melanoma cancer will be
diagnosed in United States alone in 2020, which will result in about 6,850 deaths [1]. The early
diagnosis of melanoma is very important as it helps to increase the chances of successful
treatment and the survival rate. The Computer-aided diagnosis (CAD) techniques can effectively
help doctors to diagnose and detect the melanoma in early stages [2]. The digitized
histopathological slides, which are typically obtained by staining and scanning the biopsy slides
of the skin tissue, can provide the cell morphological features with a high resolution. The
digitized slides are known as Whole Slide Images (WSIs) and with help of CAD techniques that
will permit the pathologist for precise diagnosis [3]. Pathologists usually use H&E stained
images, because the morphological features of the melanocytes and other cells become vividly
clear. In H&E stained image, the cell nuclei contain chromatin and that can be observed in blue
shade while the cytoplasm and other connective tissues are observed with varying shades of pink.
Fig. 1 shows an H&E stained histopathological image of metastatic melanoma within regional
lymph nodes (amidst adjacent salivary gland). The Lymph node cross-sections in the image are
2. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
2
contoured with black color, and melanoma metastases are contoured in green color. Note that in
this image the metastases are contoured manually by morphology. In the zoomed patch, it is
observed that the abnormal melanocytes appear with irregularity in shape and color intensity
[4,5].
Several techniques have been proposed to segment the cell nuclei in histopathological images [6-
11]. Xu et al. [6] proposed an automated technique (henceforth referred to as the
Watershed+Voting technique) to segment the cell nuclei in H&E stained images. The technique
detects the nuclei seeds by using voting areas and segments the nuclei cells using marked
watershed algorithm. The technique provides a good performance with high computational
complexity due the seed detection algorithm. Xu et al. [11] also proposed cell nuclei
segmentation technique (henceforth referred to as the gLoG+mRLS technique) using generalized
Laplacian of Gaussian (gLoG) filters to detect the seeds nuclei and multiple Radial Lines
Scanning (mRLS) algorithm to segment the cells. The mRLS uses high gradient pixel locations
and shape information to accurately segment the cell nuclei.
Figure 1. Example of an H&E stained lymph node tissue image. Lymph nodal tissue is contoured in black,
and metastatic melanoma deposits are contoured in green.
The techniques [6-11] mentioned above are generally based on extracted hand-crafted features
that require significant time to calculate. The deep learning algorithms using CNN have been
recently been used successfully in medical image analysis. The CNN models can train the feature
extraction process to provide high performance with low computational complexity in many
different tasks (e.g. classification, detection or segmentation [12-13]). Basrinarayanan et al. [14]
proposed the SegNet architecture for object segmentation. The architecture uses a number of
sampling and upsampling layers for extracting the features in hierarchical levels. Ronneberger et
al. [15] proposed the U-Net architecture for biomedical image segmentation. The U-Net
architecture has encoder and decoder sides with number of sampling and upsampling layers,
respectively. The upsampling layer outputs are enhanced by concatenating them with features
from the encoder side.
In this paper, we propose an automated technique to segment the cell nuclei and differentiate
melanoma from non-melanoma cells using a Support Vector Machine (SVM) classifier. The
technique uses a CNN architecture to segment the cell nuclei on H&E stained images. The
proposed CNN used several convolutional layers with different size of filters. Experimental
3. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
3
results demonstrate high accuracy and low computational complexity of the proposed technique
compared to the state-of-the-art techniques.
The organization of the paper is as follows. Section 2 describes the dataset used to train and
evaluate the proposed technique. Section 3 describes the proposed technique in detail. Section 4
presents the performance evaluation, followed by the conclusion in Section 5.
2. DATA DESCRIPTION
In this section, we present the details of the training and testing dataset to evaluate the
performance of the proposed nuclei segmentation and cell classification technique. The digitized
biopsies were collected at the Cross Cancer Institute, University of Alberta, Edmonton, Canada in
accordance with the protocol for the examination of specimens with skin melanoma. Standard
Neutral Buffered formalin-fixed paraffin-embedded tissue blocks of these biopsies were cut into
thin slices (e.g., 4μm for light microscope). These slices were then mounted to glass slides and
stained using H&E stain [5]. The WSIs were obtained by scanning the H&E slides using aperio
scanscope scanning system under 40X magnification. The size of a WSI is typically around
40,000×60,000 pixels (in color) and each WSI contains thousands of cell nuclei. The image
dataset consists of 9 WSIs of lymph node tissue.
3. PROPOSED TECHNIQUE
The schematic of proposed technique is shown in Fig. 2 which consists of two modules: CNN-
based nuclei segmentation and nuclei classification. The details of each module are presented in the
following.
Figure 2. Schematic of the proposed melanoma detection technique.
3.1. CNN-based Nuclei Segmentation
In this module, the input H&E stained images are segmented into cell nuclei and background.
The nuclei segmentation is done by using the proposed CNN architecture, henceforth referred to
as the NS-Net architecture (Nuclei Segmentation Net). The NS-Net architecture, shown in Fig. 3,
consists of five convolutional layers (shown in gray color) and one softmax (shown in pink)
followed by the pixel classification layer (shown in blue). The convolutional layer in the NS-
architecture consists of three operations: convolution, batch normalization [16], and activation
[17]. A brief description of each operation is presented in the following:
(i) Convolution: Let 1lf denote the (3D) feature map generated in the convolutional layer l-1.
In the convolution layer l, the feature map 1lf is convolved with a (3D) filter jF :
, 1 , 1,2,.., (1)l j l jR f F j N
where N is the number of filters (which is also known as the depth of the layer l), ,l jR is the
(2D) output corresponding to the jth convolution filter. Note that for the first convolution
layer (l=1), the input image is considered as 0f .
4. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
4
(ii) Batch normalization: During the CNN training, the convolution outputs ,l jR corresponding
to all images in a mini-batch (of S images) are considered. In this work, we have used S=8.
The ,l jR is made zero mean with unit variance as follows:
,
,ˆ (2)l j
R
l j
R
where and are the mean and standard deviation of all ,l jR in a mini-batch.
The normalized ,
ˆ
l jR is scaled with , and a bias is added as follows:
, ,
ˆ (3)l j l jy R
Note that Eqs. 2 and 3 are applied in both training and testing modes. Eq. (2) is applied with
S=8 (the mini-batch size) and S=1, in training and testing mode, respectively. The and
are trainable parameters, and are updated iteratively during the backpropagation.
(iii) Activation: In this step, an activation function is applied on the batch normalized output jy .
In this work, the Rectified Linear Unit (ReLU) activation is used. The output of the ReLU
activation function ,l jf can be expressed as follows.
, ,
max(0, ) (4)l j l j
f y
The overall output of convolutional layer l, which will be passed on to the next layer is as
follows:
,
,
0,1,...l l j
f f j N
In this architecture, the features are extracted in hierarchical levels by using convolutional filters
of different sizes. The change on the convolutional filters can precisely locate the object
boundaries that need to be segmented. Most existing CNN architectures include pooling layers. In
our experiment, it has been found that the pooling leads to loss of the spatial information that
carries important texture and shape features of the nuclei. Therefore, the pooling layer has been
omitted in the proposed architecture. Table 1 shows the number and the size of filters in each
layer of the NS-Net architecture. The NS-Net architecture is trained and evaluated using a dataset
of 24 high resolution H&E stained image patches (each with 1920×2500 color pixels) obtained
from the WSI dataset described in Section 2. Each image patch is divided into overlapping blocks
of 64×64 color pixels to obtain 458 block-images. The total number of block-images will be
10,992 (i.e. 24×458) and it is divided into 70% for training, 15% for validation and 15% for
testing. The entropy loss function with the stochastic gradient descent with momentum (SGDM)
optimizer is used to train the NS-Net architecture [18]. Fig. 4 (a) shows an input H&E stained
image and (b) shows the masked nuclei image obtained using the NS-Net architecture.
Figure 3. The proposed NS-Net architecture for nuclei segmentation.
5. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
5
Table 1. Details of the NS-Net architecture with 5 convolutional layers.
Input image size: M×N pixels (color). Number of classes: C.
Number of
Filters
Number of
Channels
Output Image
size
Filter
size
Layer-1 64 3 M×N×64 3×3
Layer-2 64 64 M×N×64 5×5
Layer-3 64 64 M×N×64 7×7
Layer-4 64 64 M×N×64 9×9
Layer-5 C 64 M×N× C 11×11
Softmax layer - C M×N×C
(a) (b)
Figure 4. Segmentation results. (a) An input image patch (b) the segmented image patch obtained using the
NS-Net architecture.
3.2. Nuclei Classification
In this module, the segmented nuclei obtained using the NS-Net architecture is classified into two
classes based on hand-crafted features. The feature vector consists of 18 first-order features, 9
Histogram of Oriented Gradient features, 24 Haralick texture features and 3 Morphological
features. The features are extracted for each pre-segmented cell nuclei and described briefly as
follows:
(i) First-order features: it includes six histogram-based features: mean, standard deviation,
third moment, smoothness, entropy, and uniformity for 3-channels (R, G and B) to
obtain 18 features (6*3).
(ii) Histogram of Oriented Gradient (HOG) features: it measures the gradient of 9
orientations in localized portions of the segmented nuclei image. A 1x9 HOG Feature
Vector (HOGFVb) is computed for each non-overlapping block of 8x8 pixels (from the
64 gradient values). Each segmented cell, depending on the cell size, may contain
several HOGFVs. The HOGFVb’s from different blocks corresponding to a cell are
summed and an overall 1x9 HOGFVn is obtained for a cell nucleus. Fig. 5 shows an
example of HOG feature extraction of a cell nucleus. Note that melanoma cells having
larger size (compared to other cells) typically contain more HOGFVb’s than other cells,
and this can result in a large magnitude of HOGFVn’s. Also, the HOGFVn’s
corresponding to the melanoma cells tend to have non-uniform distribution compared to
other cells which typically have uniform distribution.
(iii) Haralick texture features: it calculated from a Gray Level Co-occurrence Matrix,
(GLCM). It includes GLCM features such as the correlation, energy, homogeneity,
6. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
6
contrast, entropy, and dissimilarity in 4 directions (i.e., 0°, 45°, 90° and 135°) to obtain
24 features (6*4).
(iv) Morphological features: it includes the eccentricity, solidity, and the ratio of major and
minor axes of the cell nuclei to obtain (3) features.
(a) (b) (c)
Figure 5. Example of HOG feature extraction for a cell nucleus. (a) An image patch with segmented cell
nuclei and overlapped gradient orientation. (b) Blown-up of image of two nuclei (melanoma and other
cells). (c) The 1x9 HOGFVn’s of the cell nuclei.
The extracted feature vectors (with dimension 1×54) of each cell nuclei are then classified into
normal and melanoma using an SVM classifier [19-20]. The SVM is very efficient supervised
classifier that can handle even a non-linearly separable features and create hyperplane to separate
abnormal features from the normal ones. In the proposed technique, the SVM model is trained
and tested on 1,388 cell nuclei (70% for training and 30% for testing) obtained from 9 H&E
stained WSI of lymph nodes.
4. RESULTS AND DISCUSSIONS
In this section, we present the performance of the proposed nuclei segmentation and classification
technique. The performance of the segmentation technique is presented first followed by the
performance of the classification technique.
4.1. Segmentation performance
The details of the NS-Net training were presented in Section 3.1. The segmentation performance
is evaluated using 1,649 H&E stained lymph node block images (each with 64×64 color pixels).
7. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
7
The segmentation performance of the proposed technique is evaluated and compared with
handcrafted feature-based algorithms: gLoG+mRLS and Voting+Watershed techniques as well as
CNN-based techniques using SegNet and U-Net architectures. The segmentation performance is
evaluated using Precision, Accuracy and BF-score [21] measures defined as follows:
100%
TP
TP
Precisi
FP
on
100%
TP TN
Accuracy
TP FP FN TN
100%
2
2
TP TN
TP TN
B c
TN FP TN FN
F S ore
where TP, TN, FN and FP denote the number of true positives, true negatives, false negatives and
false positives, respectively. Table 2 shows the segmentation performance of different
techniques. It is observed that the deep learning algorithms provide excellent performance
compared to the classical feature-based algorithms. This is because the classical features are less
sensitive to the diversity of the cell nuclei in the skin tissue. For example, the melanoma cells
tend to have light and inhomogeneous color (see Fig. 1) and that causes miss detection of the
melanoma cells in the gLoG+mRLS and Voting+Watershed techniques.
Table 2. Segmentation Performance of the deep learning algorithms and the classical feature-based
algorithms.
In this work, the NS-Net, SegNet and U-Net architectures are trained with the same number of
training images. The NS-Net architecture is also evaluated with CNN architecture in terms of the
required parameters need to be train as shown in Table 3.
Figs. 6 (b)-(f) show the subjective segmentation performance of Voting+Watershed [6],
gLoG+mRLS [11], SegNet [14], U-Net [15] and the proposed NS-Net architecture, respectively.
It is observed that the NS-Net architecture provides excellent nuclei segmentation, whereas
gLoG+mRLS, Voting+Watershed techniques miss a few cell nuclei due to the inhomogeneity in
the cell nuclei color. It is also observed that the U-Net architecture does not perform well
compared to the other techniques because the overfitting due the large number of the filters that
are used in the cell nuclei segmentation.
Table 3. Properties of CNN architectures used in performance evaluation.
CNN
Architecture
Convolutional
layers
No. of Trained
parameters
Filter size No. of Filters
SegNet [14] 8 225,542 3×3 64
U-Net [15] 11 905,472 3×3 (64, 128, 256)
NS-Net 5 150,336 (3×3)- (11×11) 64
Technique: Precision Accuracy BF-Score Execution
time (in s)
Voting+Watershed [6] 78.24 83.64 81.31 143.71
gLoG+mRLS [11] 79.27 76.67 68.46 128.57
SegNet [14] 84.16 87.84 85.81 15.37
U-Net [15] 87.41 78.79 69.63 20.82
NS-Net 87.20 90.21 88.52 14.27
8. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
8
Figure 6. Subjective comparison of cell nuclei segmentation results (contoured in blue color) (a) original
test image, (b)-(f) Segmentation results for Voting+Watershed [6], gLoG+mRLS [11], SegNet [14], U-Net
[15] and NS-Net techniques, respectively.
4.2. Classification performance
The classification performance is evaluated using 240 H&E stained lymph node nuclei. As
explained in Section 3.2, a 1×54 size feature vector is obtained for each nucleus. The obtained
1×54 feature vectors include 18 first-order features, 9 Histogram of Oriented Gradient features,
24 Haralick texture features and 3 Morphological features. A nucleus is classified based on its
feature vector using the SVM model.
Figure 7. Three principal components (PC-1, PC-2, and PC-3) of the 1×54 features.
To demonstrate the discrimination ability of the features, the obtained features are analysed using
Principal Component Analysis (PCA) to capture the data and features variations. The PCA
technique is applied on the training cells nuclei (972 cell nuclei: 486 Melanoma and 486 other
nuclei). Fig. 7 visualizes the feature variance of melanoma and other cells with respect to three
principal components that are extracted from the nuclei features. Analyzing the feature vectors
and the principal components, it has been found that the first principal component (PC1) has high
positive association with some of the first order and Haralick features, whereas the second and
third principal components have positive association with some of the histogram and
morphological features. In this experiment, the first, second and third principal components
capture 82% of feature variance. The remaining (51) dimensions add 18% of feature variance. In
this work, all 54 features are used for the SVM classification.
9. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
9
The classification performance is evaluated in terms of the Precision, Recall and Accuracy
defined in Section 4.1. The results are shown in Table 4. The SVM classifier has been evaluated
with linear, polynomial and Gaussian kernels [18]. It is observed that the SVM classifier with
Gaussian kernel provides a superior performance over the other two kernels. Fig. 8 shows the
nuclei classification results obtained using the Gaussian kernel, where the melanoma and non-
melanoma nuclei are contoured in red and blue colors, respectively.
Table 4. Performance of the nuclei classification using different SVM kernels.
Evaluation
Measures
SVM Kernel
Linear Polynomial Gaussian
Precision 77.43 57.28 80.04
Recall 93.14 96.65 97.42
Accuracy 80.52 57.28 85.72
Figure 8. Example of classification results. (a) NS-Net input image (b) NS-Net segmented output image (c)
Classified image obtained using SVM, where melanocytes and other cell nuclei are contoured with red and
blue color, respectively.
Figure 9. Visual example of nuclei segmentation and classification obtained by the proposed technique).
The classified melanoma cells are shown in red color, whereas the non-melanoma
cells are shown in blue color.
Fig. 9 shows the segmentation and classification results of the proposed technique for the lymph
nodes in Fig.1. It is observed that the melanoma cells (in red color) are dense on the melanoma
metastasises, whereas the other cells are dense on the normal tissue of the lymph nodes. The
10. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
10
dense areas of the melanoma cells can be used to mask the melanoma regions on the lymph node
tissue. It is also noticed that there are some areas out of the lymph node tissue contains melanoma
cells and that can be processed by applying an initial lymph node segmentation technique [23].
5. CONCLUSIONS
This paper proposes an automated technique to detect melanoma nuclei in lymph node
histopathological images. The technique segments the cell nuclei in H&E stained image using a
deep learning NS-Net architecture. The NS-Net architecture segments the image into background
and cell nuclei regions. The segmented nuclei are then classified into melanoma and other cell
nuclei using an SVM classifier with Gaussian kernel. The proposed CNN architecture provides
an excellent segmentation performance with a low computational complexity. The future work
includes identifying the region of interest (ROI) with clusters of melanoma nuclei and derive
prognostic information such as proliferation index.
REFERENCES
[1] Rebecca L. Siegel, Kimberly D. Miller & Ahmedin Jemal, (2020), "Cancer Statistics, 2020," CA: A
Cancer Journal for Clinicians, vol. 70, No. 1, pp.7–30.
[2] Brochez, Lieve, Verhaeghe, Evelien, Grosshans, Edouard, et al, (2002), “Inter-observer variation in
the histopathological diagnosis of clinically suspicious pigmented skin lesions”. J. Pathol. 196(4),
pp.459–466.
[3] Cheng Lu, Muhammad Mahmood, Naresh Jha & Mrinal Mandal, (2013), “Automated segmentation
of the melanocytes in skin histopathological images”. IEEE J. Biomed. Health Inf. 17(2), pp. 284–
296.
[4] Marcial G. Rojo, Ana M. Castro, & Luis Gonçalves, (2011), “COST action “eurotelepath”: digital
pathology integration in electronic health record, including primary care centres,” Diagnostic
Pathology 6, S6 (2011), pp. S1–6.
[5] Ronald S. Weinstein, Anna R. Graham, Lynne C. Richter, et al. (2009), “Overview of telepathology,
virtual microscopy, and whole slide imaging: prospects for the future,” Human Pathology, vol. 40,
no. 8, pp. 1057–1069.
[6] Hongming Xu, Cheng Lu & Mrinal Mandal (2013), “An efficient technique for nuclei segmentation
based on ellipse descriptor analysis and improved seed detection algorithm”. IEEE J. Biomed. Health
Inf. 18(5), pp. 1729–1741.
[7] Olcay Sertel, (2012), “Image analysis for computer-aided histopathology,” Ph.D. dissertation, The
Ohio State University.
[8] Juliana M Haggerty, Xiao N Wang, Anne Dickinson, Chris J O’Malley & Elaine B Martin, (2014),
“Segmentation of epidermal tissue with histopathological damage in images of haematoxylin and
eosin stained human skin”. BMC Med. Imaging. 14(1), pp. 7.
[9] Jierong Cheng, Jagath C. Rajapakse, (2009), “Segmentation of clustered nuclei with shape markers
and marking function,” IEEE Transactions on Biomedical Engineering, vol. 56, no. 3, pp. 741–748.
[10] Bahram Parvin, Qing Yang, Ju Han, Hang Chang, Bjorn Rydberg, Mary Helen & Barcellos-Hoff,
(2007), “Iterative voting for inference of structural saliency and characterization of subcellular
events,” IEEE Transactions on Image Processing, vol. 16, no. 3, pp. 615–623.
[11] Hongming Xu, Cheng Lu, Richard Berendt, Naresh Jha & Mrinal Mandal, (2017), “Automatic nuclei
detection based on generalized Laplacian of Gaussian filters,” IEEE Journal of Biomedical and
Health Informatics (JBHI), vol.21, no.3, pp.826-837.
[12] Jonathan Long, Evan Shelhamer & Trevor Darrell, (2015), “Fully convolutional networks for
semantic segmentation,” in CVPR, pp. 3431–3440.
[13] David Eigen, & Rob Fergus, (2015) “Predicting depth, surface normals and semantic labels with a
common multi-scale convolutional architecture”, in Proceedings of the International Conference on
Computer Vision, pp. 2650–2658.
[14] Vijay Badrinarayanan, Alex Kendall, & Roberto Cipolla (2015). "SegNet: A Deep Convolutional
Encoder-Decoder Architecture for Image Segmentation." arXiv. Preprint arXiv: 1511.0051.
11. Signal & Image Processing: An International Journal (SIPIJ) Vol.11, No.4, August 2020
11
[15] Olaf Ronneberger, Philipp Fischer, & Thomas Brox (2015). "U-Net: Convolutional Networks for
Biomedical Image Segmentation." Medical Image Computing and Computer-Assisted Intervention
(MICCAI). Vol. 9351, pp. 234–241.
[16] Sergey Ioffe, & Christian Szegedy, (2015), “Batch normalization: accelerating deep network training
by reducing internal covariate shift,” arXiv preprint arXiv:1502.03167.
[17] Vinod Nair, & Geoffrey E. Hinton (2010), “Rectified linear units improve restricted Boltzmann
machines,” in Proceedings of the 27th international conference on machine learning, pp. 807–814.
[18] Christian Robert, (2014) “Machine learning, a probabilistic perspective,” CHANCE, vol. 27, no. 2,
pp. 62–63.
[19] Corinna Cortes & Vladimir Vapnik, (1995), "Support-vector networks". Machine Learning. 20 (3),
pp. 273–297.
[20]. Asa Ben-Hur, David Horn, Hava T. Siegelmann, & Vladimir N. Vapnik, (2001) "Support vector
clustering," Journal of Machine Learning Research, vol. 2, pp. 125-137.
[21] Gabriela Csurka, Diane Larlus, & Florent Perronnin (2013), "What is a good evaluation measure for
semantic segmentation?" Proc. of the British Machine Vision Conference, pp. 32.1-32.11.
[22] Hervé Abdi & Lynne J. Williams (2010), "Principal component analysis". Wiley Interdisciplinary
Reviews: Computational Statistics. 2 (4): 433–459.
[23] Salah Alheejawi, Hongming Xu & Mrinal Mandal (2018), “Novel lymph node segmentation and
proliferation index measurement for skin melanoma biopsy images” Computerized Medical Imaging
and Graphics CMIG.