- Image Interpolation, Medical Image Segmentation, Medical Image Processing, Medical Image Analysis, Liver Segmentation, Level Set Based Image Segmenation, and 10 moreNeural Network Based Segmentation, Image Analysis, Digital image processing, medical image processing, Level Set Methods, Digital Pathology, Pattern Recognition, K-means, Image Classification, Deep Learning, and Machine Learningedit
Skin diseases are considered as major public health problems. Early diagnosis of skin diseases is crucial for patients to get the treatment on time. However, several factors make difficult to access medical care. For instance, physical... more
Skin diseases are considered as major public health problems. Early diagnosis of skin diseases is crucial for patients to get the treatment on time. However, several factors make difficult to access medical care. For instance, physical disability, physiological problems, old age, distance, limited medical expertise in rural areas, climate conditions and employment. Also, long diagnosing time and high cost can discourage people from receiving dermatological care. The increasing availability and easy to use of smartphone applications has allowed significant growth of smartphones. Various smartphone applications have been introduced in different areas of medicine. They are becoming important particularly in dermatology since dermatological diseases are usually visible by human eyes and diagnosis is mainly based on visual screening of the lesions and pattern recognition. Therefore, patients can have active roles in their health-management using these applications. However, is it possible to diagnose skin diseases automatically with the advancements in mobile technologies and deep learning-based methodologies? To answer this question, in this paper, recent smartphone applications have been reviewed.
Research Interests:
The aim of this work is to classify Hemangioma, Rosacea and Acne Vulgaris diseases from digital colored photographs automatically. To determine the most appropriate deep neural network for this multi-class classification, network... more
The aim of this work is to classify Hemangioma, Rosacea and Acne Vulgaris diseases from digital colored photographs automatically. To determine the most appropriate deep neural network for this multi-class classification, network architectures have been examined. To perform a meaningful comparison of deep networks, they should be (i) implemented with the same parameters, (ii) applied with the same activation, loss and optimization functions, (iii) trained and tested with the same datasets, (iv) run on computers having the same hardware configurations. Therefore, in this work, five deep networks, which are applied widely in image classification, have been used to compare their performances by considering these factors. Those networks are VGG16, VGG19, GoogleNet, InceptionV3 and ResNet101. Comparative evaluations of the results obtained from these networks have been performed in terms of accuracy, precision and specificity. F1 score and Matthew's correlation coefficient values have also been computed. Experimental results indicated that ResNet101 architecture can classify images used in this study with higher accuracy (77.72%) than the others.
Research Interests:
Brain tumors are abnormally growing brain tissues as a result of uncontrolled multiplication of cells under certain conditions. They affect normal functions of the brain by increasing the pressure in the brain, size of it and causing... more
Brain tumors are abnormally growing brain tissues as a result of uncontrolled multiplication of cells under certain conditions. They affect normal functions of the brain by increasing the pressure in the brain, size of it and causing swelling. Therefore, they significantly impact quality of life of a patient and can be deadly. Various types of tumors, such as glioma, pituitary, meningioma and ependymoma, can occur in the brain. Accurate classification of these tumors has a vital role in early diagnosis. However, visual evaluation of many magnetic resonance images that are produced routinely is a difficult task. In the literature, computer-assisted brain tumor classification techniques have been proposed to overcome this issue. These techniques have several drawbacks or limitations. Capsule based neural networks are new approaches that can preserve spatial relationships of learned features using dynamic routing algorithm. By this way, not only performance of tumor recognition increases but also sampling efficiency and generalization capability improves. Therefore, in this work, a Capsule Network (CapsNet) is used to achieve fully automated classification of tumors from brain magnetic resonance images. In this work, prevalent three types of tumors (pituitary, glioma and meningioma) have been handled. The main contributions in this paper are as follows: 1) A comprehensive review on CapsNet based methods is presented. 2) A new CapsNet topology is designed by using a Sobolev gradient-based optimization, expectation-maximization based dynamic routing and tumor boundary information. 3) The network topology is applied to categorize three types of brain tumors. 4) Comparative evaluations of the results obtained by other methods are performed. According to the experimental results, the proposed CapsNet based technique can achieve extraction of desired features from image data sets and provides tumor classification automatically with 92.65% accuracy.
Research Interests:
Deep convolutional neural networks have been implemented for image classification tasks and achieved promising results in recent years. Particularly, ResNETs have been used since they can eliminate vanishing gradient problem in very deep... more
Deep convolutional neural networks have been implemented for image classification tasks and achieved promising results in recent years. Particularly, ResNETs have been used since they can eliminate vanishing gradient problem in very deep networks. However, ResNET architectures with different activation functions, batch sizes, number of images in the testing and training stages can cause different results. Therefore, the effect of residual connections and activation functions image classification is still unclear. Also, in the literature, ResNET based models have been trained and tested with data sets having different characteristics. However, to make meaningful evaluations of the results obtained from different ResNET models, the same data sets should be used. Therefore, in this work, four network models have been implemented to analyze the effect of two activation functions (ReLU and SELU) and residual learning for image classification using the same data sets. To evaluate performances of these models, a real world issue, which is automated skin disease classification from colored digital images, has been handled. Experimental results and comparative analyses indicated that the ResNET with SELU and without residual block yields in the highest validation accuracy (97.01%) for image classification.
Research Interests:
This work aims to study performance of different deep learning based approaches to classify skin diseases automatically from colored digital photo-graphs. We applied recent network models, which are U-Net, Inception Version-3... more
This work aims to study performance of different deep learning based approaches to classify skin diseases automatically from colored digital photo-graphs. We applied recent network models, which are U-Net, Inception Version-3 (InceptionV3), Inception and Residual Network (InceptionResNetV2), VGGNet, and Residual Network (ResNet). Comparative evaluations of the results obtained by these network models indicated that automated diagnosis from digital photographs is possible with accuracy between 74% (by U-net) and 80% (by ResNet). Therefore, further studies are still required in this area to design and develop a new model by combining advantages of different network models and also to obtain higher accu-racy. In addition, testing of the model should be performed with more data including more diversity to see reliability of the model.
Research Interests:
Image segmentation has a key role in computer vision and image processing. Superiority of deep learning based segmentation techniques has been shown in various studies in the literature. However, there are challenging issues affecting... more
Image segmentation has a key role in computer vision and image processing. Superiority of deep learning based segmentation techniques has been shown in various studies in the literature. However, there are challenging issues affecting performances of these methods. Therefore, in this paper, these challenges that are mostly related to architecture and training of deep neural networks are explained. In addition, the state-of-the-art solutions applied in the literature are presented to help researchers to design proper network architectures according to their problems and to be aware of possible challenging issues and recent solutions.
Research Interests:
Alzheimer's disease is a neuropsychiatric, progressive, also an irreversible disease. There is not an effective cure for the disease. However, early diagnosis has an important role for treatment planning to delay its progression since the... more
Alzheimer's disease is a neuropsychiatric, progressive, also an irreversible disease.
There is not an effective cure for the disease. However, early diagnosis has an important role for treatment planning to delay its progression since the treatments have the most impact at the early stage of the disease. Neuroimages obtained by different imaging techniques (for example, diffusion tensor‐based and magnetic resonance‐based imaging) provide powerful information and help to diagnose the disease. In this work, a deeply supervised and robust method has been developed using three dimensional features to provide objective and accurate diagnosis from magnetic resonance images. The main contributions are (a) a new three dimensional convolutional neural network topology; (b) a new Sobolev gradient‐based optimization with weight values for each decision parameters; (c) application of the proposed topology and optimizer to diagnose Alzheimer's disease; (d) comparisons of the results obtained from the recent techniques that have been implemented for
Alzheimer's disease diagnosis. Experimental results and quantitative evaluations indicated that the proposed network model is able to achieve to extract desired features from images and provides automated diagnosis with 98.06% accuracy.
There is not an effective cure for the disease. However, early diagnosis has an important role for treatment planning to delay its progression since the treatments have the most impact at the early stage of the disease. Neuroimages obtained by different imaging techniques (for example, diffusion tensor‐based and magnetic resonance‐based imaging) provide powerful information and help to diagnose the disease. In this work, a deeply supervised and robust method has been developed using three dimensional features to provide objective and accurate diagnosis from magnetic resonance images. The main contributions are (a) a new three dimensional convolutional neural network topology; (b) a new Sobolev gradient‐based optimization with weight values for each decision parameters; (c) application of the proposed topology and optimizer to diagnose Alzheimer's disease; (d) comparisons of the results obtained from the recent techniques that have been implemented for
Alzheimer's disease diagnosis. Experimental results and quantitative evaluations indicated that the proposed network model is able to achieve to extract desired features from images and provides automated diagnosis with 98.06% accuracy.
Comparison of medical images is frequently needed for diagnosis or evaluation of a progressive disease. The comparison is performed by alignment and registration of images or warping them with a transformation function. It is possible to... more
Comparison of medical images is frequently needed for diagnosis or evaluation of a progressive disease. The comparison is performed by alignment and registration of images or warping them with a transformation function. It is possible to compare images of the same patient, which have been taken at different time periods to detect or quantify the changes that might have taken place in-between acquisitions. Also, a comparison can be performed using images from different subjects. In a Magnetic Resonance Image (MRI), intensity values do not only depend on the underlying tissue type. They also depend on developmental processes, scanner-related intensity artifacts and disease progression. Therefore, spatial normalization, which brings an image into the coordinate system of a template using a coordinate transformation to make meaningful comparisons of spatially varying data, is required. In this work, an intensity normalization method based on spatially varying distribution matching is proposed. The efficiency of the proposed method has been shown on brain MRIs.
Research Interests:
Deep Learning (DL) methods are a set of algorithms in Machine Learning (ML), which provides an effective way to analysis medical images automatically for diagnosis/assessment of a disease. DL enables higher level of abstraction and... more
Deep Learning (DL) methods are a set of algorithms in Machine Learning (ML), which provides an effective way to
analysis medical images automatically for diagnosis/assessment of a disease. DL enables higher level of abstraction and
provides better prediction from datasets. Therefore, DL has a great impact and become popular in recent years. In this
work, we present advances and future researches on DL based medical image analysis.
analysis medical images automatically for diagnosis/assessment of a disease. DL enables higher level of abstraction and
provides better prediction from datasets. Therefore, DL has a great impact and become popular in recent years. In this
work, we present advances and future researches on DL based medical image analysis.
Research Interests:
Living donated liver transplantation is an important task since a person (healthy donor) donates some part of her/his liver to a person in this surgery operation. The success of this operation mainly depends on the sufficiency of vessels... more
Living donated liver transplantation is an important task since a person (healthy donor) donates some part of her/his liver to a person in this surgery operation. The success of this operation mainly depends on the sufficiency of vessels and volume of the liver. Accurate labeling of portal and hepatic veins of donors reduces the incidence of complications during and after transplantation. Therefore, prior to the hepatic surgery, automatic analysis and labeling of vasculature structures in the liver are vital to see whether liver is suitable or not for transplantation. However, automatic labeling of veins in the liver is challenging because of partial volume effects, noise and image resolution, which causes wrong connections between vessels. The goal of this paper is to propose an automatic labeling approach for vessels. The proposed automated labeling method is based on gray-level values in the MR images and anatomical information. In this work, detection and segmentation of vascular structures in the liver is performed automatically with clustering-based segmentation and refinement stages. The accuracy of the automatic labeling approach is 85 %. Required processing time for the proposed method (average 6 s) is shorter than manual approach (average 295 s) for labeling of hepatic and portal veins from segmented vessels. The proposed approach is efficient in terms of both computational cost and accuracy of labeling and segmentation of hepatic and portal veins.
Research Interests: Pattern Recognition, Magnetic Resonance Imaging, Image Classification, Liver, Liver Transplantation, and 10 moreK means Clustering, Automated Detection, LIVER TRANSPLANT, Portal vein, Automated Vessel Segmentation, Liver Vessel, Vessel Labeling, Hepatic Vein, Portal Vein Labeling, and Liver Segments
Research Interests:
Research Interests:
Research Interests:
Quantitative analysis and precise measurements on the liver have vital importance for pre-evaluation of surgical operations and require high accuracy in liver segmentation from all slices in a data set. However, automated liver... more
Quantitative analysis and precise measurements on the liver have vital importance for pre-evaluation of surgical operations and require high accuracy in liver segmentation from all slices in a data set. However, automated liver segmentation from medical image data sets is more challenging than segmentation of any other organ due to various reasons such as vascular structures in the liver, high variability of liver shapes, similar intensity values, and unclear edges between liver and its adjacent organs. In this study, a variational level set-based segmentation approach is proposed to be efficient in terms of processing time and accuracy. The efficiency of this method is achieved by (1) automated initialization of a large initial contour, (2) using an adaptive signed pressure force function, and also (3) evolution of the level set with Sobolev gradient. Experimental results show that the proposed fully automated segmentation technique avoids local minima and stops evolution of the active contour at desired liver boundaries with high speed and accuracy. Copyright © 2016 John Wiley & Sons, Ltd.
Research Interests:
ABSTRACT
Research Interests:
Research Interests:
AbstractPurpose: Histogram Analysis in Radiation Therapy (HART) is an efficient dose-volume histogram (DVH) computational tool in radiation therapy and radiation oncology research. Various applications of the software were also presented... more
AbstractPurpose: Histogram Analysis in Radiation Therapy (HART) is an efficient dose-volume histogram (DVH) computational tool in radiation therapy and radiation oncology research. Various applications of the software were also presented and published earlier in different journals (Med Phys 37(6), p.3217, 2010; J Appl Clin Med Phys 11(1), p.3013, 2010). The main objective of this work was to review the important applications of the program. Method and Materials:MATLAB based codes were primarily designed to read and write a simpler HART format of the DVH statistics, from the standard RTOG data formats exported from the Pinnacle3 treatment planning system (TPS; Philips Healthcare, Best, Netherlands). Various applications such as conventional DVH (cDVH) analysis, and spatial DVH (sDVH; x-,y-, and z-DVHs respectively) analysis, universal-plan indices (UPI) evaluation, biological modeling based outcome analyses (BMOA), radiobiological dose-response modeling (DRM), and physical parameterization (PP) modules have been incorporated in the program. The fundamental results obtained in these applications, were thoroughly validated using the primary data derived from the DVH statistics extracted from the Pinnacle3 system. The program also comprises the simple computational mechanism, the graphical simulations, and the flexible interactive modules. Results: HART offers cDVH and sDVH computational modules, UPI evaluations, BMOA features, DRM simulations, and PP modules respectively for the radiotherapy plans. The cDVH and BMOA were the most applicable features among the HART users in the past year. Nearly 50% of the users (N=91) have found the program useful around the globe. The program is also available freely online. Conclusions: Several applications have been upgraded into a simpler, user-friendly, and automated software package, HART. The program is useful to the medical physics and radiation oncology communities. We further expect to develop HART for various applications in radiotherapy research, and its expansion to other TPSs that utilize DICOM-RT objects.
Research Interests:
ABSTRACT
Research Interests: Medical Image Processing, Neural Networks, K-means, Segmentation, Level Set Based Image Segmenation, and 8 moreLiver Transplantation, Medical Image Segmentation, GMM Analysis, Liver Segmentation, Computer Aided Detection and Diagnosis for Medical Images, Image Processing, Active Contours, Segmentation, MRI image processing, and Automated Segmentation
ABSTRACT
Research Interests:
Accurate liver segmentation is an important component of surgery planning for liver transplantation, which enables patients with liver disease a chance to survive. Spectral pre-saturation inversion recovery (SPIR) image sequences are... more
Accurate liver segmentation is an important component of surgery planning for liver transplantation, which enables patients with liver disease a chance to survive. Spectral pre-saturation inversion recovery (SPIR) image sequences are useful for liver vessel segmentation because vascular structures in the liver are clearly visible in these sequences. Although level-set based segmentation techniques are frequently used in liver segmentation due to their flexibility to adapt to different problems by incorporating prior knowledge, the need to initialize the contours on each slice is a common drawback of such techniques. In this paper, we present a fully automated variational level set approach for liver segmentation from SPIR image sequences. Our approach is designed to be efficient while achieving high accuracy. The efficiency is achieved by (1) automatically defining an initial contour for each slice, and (2) automatically computing weight values of each term in the applied energy functional at each iteration during evolution. Automated detection and exclusion of spurious structures (e.g. cysts and other bright white regions on the skin) in the pre-processing stage increases the accuracy and robustness. We also present a novel approach to reduce computational cost by employing binary regularization of level set function. A signed pressure force function controls the evolution of the active contour. The method was applied to ten data sets. In each image, the performance of the algorithm was measured using the receiver operating characteristics method in terms of accuracy, sensitivity and specificity. The accuracy of the proposed method was 96%. Quantitative analyses of results indicate that the proposed method can accurately, efficiently and consistently segment liver images.
Research Interests:
... Performance evaluation of nonrigid registration methods is very difficult since ... IV. SELECT INTERPOLATION METHODS OF THE LAST TEN YEARS Energy variation based interpolation approaches have been proposed. An ideal image should take... more
... Performance evaluation of nonrigid registration methods is very difficult since ... IV. SELECT INTERPOLATION METHODS OF THE LAST TEN YEARS Energy variation based interpolation approaches have been proposed. An ideal image should take minimum energy. ...
Research Interests:
Robust kidney segmentation from MR images is a very difficult task due to the especially gray level similarity of adjacent organs, partial volume effects and injection of contrast media. In addition to different image characteristics with... more
Robust kidney segmentation from MR images is a very difficult task due to the especially gray level similarity of adjacent organs, partial volume effects and injection of contrast media. In addition to different image characteristics with different MR scanners, the variations of the kidney shapes, gray levels and positions make the identification and segmentation task even harder. In this paper, we propose an automatic kidney segmentation approach using Gaussian mixture model (GMM) that adapts all parameters according to each MR image dataset to handle all these challenging problems. The efficiency in terms of the segmentation performance is achieved by the estimation of the GMM parameters using the Expectation Maximization (EM) method. The segmentation approach is compared to k-means method. The results show that the model based probabilistic segmentation technique gives better performance for both low contrast images and atypical kidney shapes where several algorithms fail on abdominal MR images.