Yazar "Karakis, Rukiye" seçeneğine göre listele
Listeleniyor 1 - 10 / 10
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe A Novel Color Image Watermarking Method with Adaptive Scaling Factor Using Similarity-Based Edge Region(Tech Science Press, 2023) Gurkahraman, Kali; Karakis, Rukiye; Takci, HidayetThis study aimed to deal with three challenges: robustness, imperceptibility, and capacity in the image watermarking field. To reach a high capacity, a novel similarity-based edge detection algorithm was developed that finds more edge points than traditional techniques. The colored watermark image was created by inserting a randomly generated message on the edge points detected by this algorithm. To ensure robustness and imperceptibility, watermark and cover images were combined in the high-frequency subbands using Discrete Wavelet Transform and Singular Value Decomposition. In the watermarking stage, the watermark image was weighted by the adaptive scaling factor calculated by the standard deviation of the similarity image. According to the results, the proposed edge-based color image watermarking technique has achieved high payload capacity, imperceptibility, and robustness to all attacks. In addition, the highest performance values were obtained against rotation attack, to which sufficient robustness has not been reached in the related studies. © 2023 CRL Publishing. All rights reserved.Öğe Brain tumors classification with deep learning using data augmentation(Gazi Univ, Fac Engineering Architecture, 2021) Gurkahraman, Kali; Karakis, RukiyeMedical image classification is the process of separating data into a specified number of classes. In recent years, Magnetic Resonance Imaging (MRI) has been widely used in the detection and diagnosis of brain tumors. In this study, it was aimed to classify three different brain tumors (glioma, meningioma and pituitary) using convolutional neural network (CNN) on T1-weighted MR images and to determine the efficiency of axial, coronal and sagittal MR planes in classification. The weights were initialized by transferring to CNN from DenseNet121 network, which was previously trained with ImageNet dataset. In addition, data augmentation was performed on MR images using affine and pixel-level transformations. The features obtained from the first fully connected layer of the trained CNN were also classified by support vector machine (SVM), k nearest neighbor (kNN), and Bayes methods. The performances of these classifiers were measured by the sensitivity, specificity, accuracy, area under curve, and the Pearson correlation coefficient on the test dataset. The accuracy values of the developed CNN and CNN-based SVM, kNN, and Bayes classifiers are 0.9860, 0.9979, 0.9907, and 0.8933, respectively. The CNN-based SVM model proposed for brain tumor classification obtained higher performance values than similar studies in the literature. In addition, coronal plane of the brain was found to give better results than other planes in determining the tumor type.Öğe Classification of Brain Tumors using Convolutional Neural Network from MR Images(IEEE, 2020) Gungen, Cahfer; Polat, Ozlem; Karakis, RukiyeThe classification of brain tumors has great importance in medical applications that benefit from computer-assisted diagnosis. Misdiagnosis of brain tumor types, both prevents the patient's response to treatment effectively and reduce the chance of survival. This study proposes a solution for the classification of brain tumors using MR images. The most common brain tumors, glioma, meningioma and pituitary, are detected using convolutional neural networks. The convolutional network is trained and tested on an accessible Figshare dataset containing 3064 MR images using four different optimizers. AUC, sensitivity, specificity and accuracy are used as performance measure. The proposed method is comparable to the literature and classifies brain tumors with an average accuracy of 96.84% and a maximum accuracy of 97.75%.Öğe Comparison of deep learning methods for the radiographic detection of patients with different periodontitis stages(Oxford Univ Press, 2024) Ayyildiz, Berceste Guler; Karakis, Rukiye; Terzioglu, Busra; Ozdemir, DurmusObjectives The objective of this study is to assess the accuracy of computer-assisted periodontal classification bone loss staging using deep learning (DL) methods on panoramic radiographs and to compare the performance of various models and layers.Methods Panoramic radiographs were diagnosed and classified into 3 groups, namely healthy, Stage1/2, and Stage3/4, and stored in separate folders. The feature extraction stage involved transferring and retraining the feature extraction layers and weights from 3 models, namely ResNet50, DenseNet121, and InceptionV3, which were proposed for classifying the ImageNet dataset, to 3 DL models designed for classifying periodontal bone loss. The features obtained from global average pooling (GAP), global max pooling (GMP), or flatten layers (FL) of convolutional neural network (CNN) models were used as input to the 8 different machine learning (ML) models. In addition, the features obtained from the GAP, GMP, or FL of the DL models were reduced using the minimum redundancy maximum relevance (mRMR) method and then classified again with 8 ML models.Results A total of 2533 panoramic radiographs, including 721 in the healthy group, 842 in the Stage1/2 group, and 970 in the Stage3/4 group, were included in the dataset. The average performance values of DenseNet121 + GAP-based and DenseNet121 + GAP + mRMR-based ML techniques on 10 subdatasets and ML models developed using 2 feature selection techniques outperformed CNN models.Conclusions The new DenseNet121 + GAP + mRMR-based support vector machine model developed in this study achieved higher performance in periodontal bone loss classification compared to other models in the literature by detecting effective features from raw images without the need for manual selection.Öğe Edge-Based Image Watermarking Method with Weighted Discrete Cosine Transform Coefficients(Institute of Electrical and Electronics Engineers Inc., 2022) Karakis, Rukiye; Gurkahraman, KaliIn image watermarking, hybrid approaches increase imperceptibility and robustness. Also, a scaling factor is used, which should be optimized when combining the cover image and watermark. In this study, discrete wavelet transform and discrete cosine transform (DCT) were used together. The watermark-edge image was obtained by randomly inserting the watermark on the horizontal, vertical and diagonal edge points of the cover image detected with Sobel. The DCT frequency components of the watermark-edge image were weighted with a generated matrix and combined with the DCT of the cover image. According to the obtained results, the proposed method is imperceptible and robust to various attacks, especially JPEG compression and noise attacks. © 2022 IEEE.Öğe Estimation of specific surface area and higher heating value of biochar and activated carbon produced by pyrolysis and physico-chemically assisted pyrolysis of biomass using an artificial neural network (ANN)(Springer Heidelberg, 2025) Balde, Mamadou Saliou; Karakis, Rukiye; Ates, AytenThe physical and chemical activation of biomass prior to pyrolysis significantly affects the properties of the activated carbon produced. In this study, raw tea waste (TW) and hazelnut shells (HS) were used to produce biochar and activated carbon samples by pyrolysis at different pyrolysis temperatures with and without chemical and physical activation. Subsequently, an artificial neural network (ANN) was developed based on the pyrolysis conditions, proximate and elemental analyses of the biomass feedstocks and the obtained biochar and activated carbon to predict the higher heating value (HHV) and specific surface area (SSA) of the biochar. For this purpose, machine learning algorithms such as ANN, Gaussian process regression (GPR), regression trees (RT), and support vector machines (SVM) were compared to find the best-performing algorithm for the prediction of HHV and SSA of biochar. Algorithms based on ANNs performed better than SVM, RT, and GPR models, with higher regressions and lower prediction errors. The resilient backpropagation (RProp) algorithm proved to be the most suitable training algorithm as it provided satisfactory results with a low percentage of mean squared error (MSE) and mean absolute error (MAE). The ANN models showed moderate to strong performance in the tests, with correlation coefficient (R) values of 0.82 and 0.95, coefficient of determination (R2) values of 0.67 and 0.90, and low MAE and MSE, indicating reasonable prediction accuracy for HHV and SSA of the biochar. The energy efficiency of biochar produced with conventional pyrolysis ranged from 9.84% to 21.13%, while the energy efficiency of activated carbon ranged from 45.26% to 67.21%, with the maximum reached at 300 degrees C. Based on the results of the thermodynamic analysis, it was found that the energy and exergy yields of the biochar and activated carbon produced depend on the activation conditions and temperature.Öğe Evaluation of segmented brain regions for medical image steganography(Gazi Univ, Fac Engineering Architecture, 2021) Karakis, Rukiye; Gurkahraman, Kali; Cigdem, Burhanettin; Oztoprak, Ibrahim; Topaktas, A. SuatIn medical image steganography, diagnosis and treatment of a disease can be affected as a result of the distortion caused by the embedding data in the images. For this reason, data is embedded in the region of non-interest determined by basic techniques such as manual or thresholding, and none of these methods involve the segmentation of brain tissues such as tumours. The present study aims to hide the data used in the diagnosis and treatment of a disease without affecting the medical information in the images with a segmentation-based steganography method by combining them into one file format. Magnetic Resonance (MR) images of epilepsy patients were segmented as background, gray matter, white matter, and tumour by discrete wavelet transform (DWT) and k-means clustering-based segmentation method. The hidden data includes confidential patient information, doctor's comment, selected Electroencephalogram (EEG) signals, and EEG health reports. The high-capacity message, which encoded by DNA encryption using chaotic and hash functions, and then compressed, is hidden in the least significant bits of non-tumour pixels of images. In the study, the difference between the cover and the stego images was measured by the peak signal-to-noise ratio, the structural similarity measure, the universal quality index, and the correlation coefficient. These values were obtained as 64.0334 decibels (dB), 0.9979, 0.9971, 0.9993, respectively. A comparison of the results indicates that the proposed method combines the high capacity data of the patients in a single file format and increases both the security and recording space of medical data.Öğe NOISE REMOVAL IN MAGNETIC RESONANCE IMAGING USING 3D DEEP LEARNING MODEL(Mugla Sitki Kocman University, 2024) Karakis, Rukiye; Topdag, TugbaMagnetic Resonance Imaging (MRI) is a widely used imaging technique for examining brain tissues and diagnosing various conditions. However, MRI images often contain noise caused by factors such as equipment limitations, environmental conditions, patient movement, and magnetic field interference. This noise can obscure critical details, making accurate diagnosis and treatment planning challenging. In this study, the focus is on the removal of Rician noise from MRI images. To address this challenge, two 3D autoencoder models, named M-UNet+ResNet and M-UNet+DenseNet, were developed. These models are based on an enhanced UNet architecture that integrates dense and residual connections, aimed at improving noise reduction capabilities. The models were trained using T1 and T2-weighted MRI images from the IXI dataset, incorporating noise levels varying from 3% to 15%. Their performance was evaluated using metrics such as peak signal-to-noise ratio, structural similarity index measure, and mean absolute error. The results demonstrated that both models effectively reduced noise across various levels, with M-UNet+ResNet generally outperforming M-UNet+DenseNet. Notably, M-UNet+ResNet achieved PSNR values of 38.72 dB and 37.04 dB, and SSIM values of 0.82 and 0.81 in the IXI-HH-T2 and IXI-Guys-T2 datasets, respectively, indicating its strong capability in preserving image quality. This study concludes that incorporating residual connections in DL models enhances their ability to remove noise from MRI images, offering a solution for maintaining the integrity of medical images in clinical settings.Öğe Steganalysis on Medical Images with Support Vector Machine(IEEE, 2020) Maroof Ozcan, Fatmanur Betul; Karakis, Rukiye; Guler, IvanDICOM is a file format standard for medical images. DICOM file format includes a header with the patient's personal information as well as information related to the image. In steganography applications on medical data, patient's personal information, reports, and other data are hidden in the medical images. Steganalysis detects whether any secret data is hidden in a file or not. In this study, a steganalysis method is proposed to detect the presence of hidden data in a medical image. For this reason, the secret message that contains personal data extracted from the medical image header is embedded into the LSBs of the MR image pixels by using six different methods. The support vector machine (SVM) is used as a classifier for steganalysis between cover and stego images. As a result of the analysis, the presence of hidden data in medical images is found with 99.28% accuracy and 0.9856 correlation coefficient value.Öğe Steganography and Medical Data Security(CRC Press, 2018) Karakis, Rukiye; Guler, InanThis chapter describes medical information systems, the details of picture archiving and communication system (PACS) and Digital Imaging and Communications in Medicine (DICOM). It presents medical steganography technique formulations and demonstrates their performance on medical images. The chapter provides the comparison metrics of medical image steganography. The development of computer systems has affected the healthcare industry. The electronic health record of a patient is a systematic collection of personal health information from birth to the time after death. PACS manages the software and platforms to acquire, store, distribute, and retrieve the medical images. In PACS, medical images are stored in DICOM file format. DICOM includes definitions for different devices, software, and platforms that are compatible with each other in PACS. In medical steganography, a cover object can be medical images or biological signals. The capacity of embedding message and robustness are essential problem for medical steganography application. © 2019 by Taylor & Francis Group, LLC.
