Mühendislik ve Doğa Bilimleri Fakültesi

Permanent URI for this collectionhttps://acikerisim.samsun.edu.tr/handle/123456789/24

Browse

Now showing 1 - 20 of 265

Siyam sinir ağlarını kullanarak Türk işaret dilindeki rakamların tanımlanması
(Dokuz Eylül Üniversitesi Mühendislik Fakültesi Fen ve Mühendislik Dergisi, 2020-04-27) Toğaçar, Mesut; Cömert, Zafer; Ergen, Burak
İletişim, insanların duygu, düşünce veya bilgiyi çeşitli yollar kullanarak karşı tarafa aktarma sürecidir. İletişimde en etkili yollardan birisi ise dildir. Dil, insanların günlük hayatını kolaylaştıran bir iletişim aracıdır ve bu iletişim aracını kullanamayan işitme engelli birçok insan vardır. İşitme engelli insanların, toplum içerisinde iletişimini kolaylaştırmak için işaret dilleri geliştirilmiştir. Her ülkenin kendi konuşma diline özgü işaret dili mevcuttur. Bu çalışma erişime açık Türk işaret dili rakamlarına odaklanmıştır. İşaret dili, toplumun her kesimi tarafından bilinmemektedir. Bu durum, işitme engelli insanların bulundukları sosyal ortamlarda iletişim aksaklıklarına neden olmaktadır. İşitme engelli olmayan ancak işaret dilini bilmeyen bir birey de aynı problemi yaşamaktadır. Bu çalışmanın amacı, işaret dilini kullanan insanların ne anlatmak istediğini derin öğrenme mimarisi üzerinde tespit etmektir. Bu amaçla, işaret dili rakamlarının, son zamanlarda popülerliği artan siyam sinir ağı ile tanımlanması bu çalışmada gerçekleştirilmiştir. Siyam sinir ağları, görüntü kümesinde aynı görüntüleri eşleştiren bir derin öğrenme modelidir. Bu ağları kullanarak Türk işaret dilinde kullanılan rakam görüntülerini tanımlamayı gerçekleştirdik. Elde edilen eşleştirme başarı oranı %98,16’dır. Sonuç olarak, bu çalışma ile Türk işaret dili rakamlarının tanımlanmasında siyam sinir ağlarının başarılı olduğu görülmüştür. Communication is the process of people transferring emotions, thoughts or information to the other party in various ways. One of the most effective ways of communication is language. Language is a communication tool that makes people's daily life easier and there are many hearing impaired people in our lives who cannot use this communication tool. Sign languages have been developed to facilitate the communication of hearing impaired people in society. There are specific sign languages varying according to the language of the countries. This study focuses on the Turkish sign language digits that are publicly available. Sign language is not known by all people of society. This situation causes communication disruptions in the social environments where hearing impaired people are present. A person who has not hearing impaired but cannot use sign language has the same problem. The aimof this study is to determine what people using sign language want to tell by using a deep learning architecture. For this purpose, the identification of digits in Turkish sign language has been realized by using the recently popular siamese neural network in this study. Siamese neural networks are a type of deep learning model that matches the same images in an image dataset. Using these networks, we have identified the digits used in Turkish sign language. The success rate of the matching was 98.16%. Consequently, siamese neural networks were found to be successful in identifying Turkish sign language digits with this study.
Intelligent skin cancer detection applying autoencoder, MobileNetV2 and spiking neural networks
(Chaos, Solitons and Fractals Nonlinear Science, and Nonequilibrium and Complex Phenomena, 2021-01-03) Toğaçar, Mesut; Ergen, Burhan; Cömert, Zafer
Melanocytes are skin cells that give color to the skin and form melanin color pigments. The unbalanced division and proliferation of these cells result in skin cancer. The early diagnosis and proper treatment of skin cancer are so important. In this scope, a novel model that relies upon the autoencoder, spiking, and convolutional neural networks is proposed to ensure a useful decision support tool in this study. The experiments were carried out on an open-access dataset called the ISIC skin cancer consisting of 1800 being and 1497 malignant tumor images. In the proposed approach, the dataset is reconstructed using the autoencoder model. The original dataset and structured dataset were trained and classified by the MobileNetV2 model that consists of residual blocks, and the spiking networks. The classification success rate of the study was 95.27%. As a result, it was seen that the autoencoder model and spiking networks contributed to enhancing the performance of the MobileNetV2 model. Thanks to the proposed model, a novel fully automated decision support tool with high sensitivity was ensured for skin cancer detection.
Identification of haploid and diploid maize seeds using convolutional neural networks and a transfer learning approach
(Computers and Electronics in Agriculture, 2019-06-24) Altuntaş, Yahya; Cömert, Zafer; Kocamaza, Adnan Fatih
Maize is one of the most significant grains cultivated all over the world. Doubled-haploid is an important technique in terms of advanced maize breeding, modern crop improvement and genetic programs, since this technique shortens the breeding period and increases breeding efficiency. However, the selection of the haploid seeds is a major problem of this breeding technique. This process is frequently conducted manually, and this unreliable situation leads to loss of time and labor. Inspired by the recent successes of deep transfer learning, in this study, we approached this problem as a computer vision task to provide a nondestructive, rapid and low-cost model. To achieve this objective, we adopted convolutional neural networks (CNNs) to recognize haploid and diploid maize seeds automatically through a transfer learning approach. More specifically, AlexNet, VVGNet, GoogLeNet, and ResNet were applied for this specific task. The experimental study was carried out using a new dataset consisting of 1230 haploid and 1770 diploid maize seed images. The samples in the dataset were classified considering a marker-assisted selection, known as the R1-nj anthocyanin marker. To measure the success of the CNN models, we utilized several performance metrics, such as accuracy, sensitivity, specificity, quality index, and F-score derived from the confusion matrix and receiver operating characteristic curves. According to the experimental results, the CNN models ensured promising results, and we achieved the most efficient results via VGG-19. The accuracy, sensitivity, specificity, quality index, and F-score of VGG-19 were 94.22%, 94.58%, 93.97%, 94.27%, and 93.07%, respectively. Consequently, the experimental results proved that CNN models can be a useful tool in recognizing haploid maize seeds. Furthermore, we conclude that this approach is significantly superior to machine learning-based methods and conventional manual selection.
Fusing fine-tuned deep features for recognizing different tympanic membranes
(ScienceDirect, 2019-11-08) Cömert, Zafer
Otitis media (OM) refers to a group of inflammatory diseases regarding the middle ear. Although there are a wide variety of disease types regarding OM, the most commonly seen disorders are acute otitis media (AOM), otitis media with effusion (OME) and chronic suppurative otitis media (CSOM). The examination of OM in the clinics is realized subjec tively. This subjective examination is error-prone and leads to a limited variability among specialist. For these reasons, computer-aided systems are in demand. In this study, we focus on recognizing normal, AOM, CSOM, and earwax tympanic membrane (TM) conditions using fused fine-tuned deep features provided by pre-trained deep convolutional neural networks (DCNNs). These features are applied as the input to several networks, such as an artificial neural network (ANN), k-nearest neighbor (k NN), decision tree (DT) and support vector machine (SVM). Moreover, we release a new publicly available TM data set consisting of totally 956 otoscope images. As a result, the DCNNs yielded promising results. Especially, the most efficient results were provided by VGG-16 with an accuracy of 93.05 %. The fused fine tuned deep features improved the overall classification success. Finally, the proposed model yielded promising results with an accuracy of 99.47 %, sensitivity of 99.35 %, and specificity of 99.77 % using the combination of the fused fine-tuned deep features and SVM model. Consequently, this study shows that fused fine-tuned deep features are rather useful in recognizing different TMs and these features can provide a fully automated model with high sensitivity.
Kulak içi hastalıklarının derin öğrenme mimarileriyle sınıflandırılması ve karşılaştırılması
(Avrupa Bilim ve Teknoloji Dergisi, 2023-04-12) Demircan, Furkancan; Ekinci, Murat; Cömert, Zafer
Otitis media (OM), kulak zarı içerisinde oluşan akıntılı, enfeksiyonel hastalıkları tanımlamaktadır. Kulak mumu (earwax), kulak zarı içerisinde bakteri oluşumunu önleyen savunma mekanizmasının aşırı birikimi sonucunda kulakta işitme kaybı oluşmasına neden olan hastalıktır. Kulak zarı içerisinde kalsiyum birikimi sonucunda saydamlığını ve esnekliğini kaybetmesine miringoskleroz denmektedir. Bu hastalıkların tanısı Kulak Burun Boğaz (KBB) uzmanları tarafından kulak zarının otoskopla incelenmesi sonucunda koyulmaktadır ve hataya açıktır. Bu çalışmada, bu problemin çözümüne katkı sağlamak ve bir karar destek sistemi sunmak amacıyla derin öğrenme modelleriyle kulak zarı hastalıklarına ait görüntüler sınıflandırılmıştır. Veri seti olarak 4 sınıf ve 880 görüntünün bulunduğu Ear Imagery veri seti seçilmiştir. Sınıflandırma işlemi için AlexNet, ResNet50, ResNet101, ResNet50V2, ResNet101V2, InceptionV3, Xception ve InceptionResNetV2 derin öğrenme modelleri seçilmiştir. En yüksek başarı değeri %94 doğruluk ile InceptionResNetV2 mimarisinden ve en hızlı sonuç 438 saniye ile AlexNet mimarisinden elde edilmiştir. Bu yaklaşımla kulak zarına ait hastalıkların potansiyel uzman hatalarından arındırılarak otonom bir sistem ile gerçekleştirilebileceği gösterilmiştir. Gelecekte klinik alanda böyle bir sistemin kullanılması; uzmanların karar verme sürecini destekleyebilir ve hataya açık olan değerlendirme sürecinin daha objektif ve tekrar edilebilir bir şekilde yönetilmesini sağlayabilir. Otitis Media (OM) is infectious disease with discharge in the eardrum. Earwax is a disease that causes hearing loss in the ear as a result of excessive accumulation of the defense mechanism that prevents the formation of bacteria in the eardrum. The loss of transparency and flexibility as a result of calcium accumulation in the eardrum is called myringosclerosis. The diagnosis of these diseases is made by otolaryngologists using an otoscopy examination of the eardrum and this process is prone to error. In this study, otoscopy images were classified with deep learning models to solve this problem. The Ear Imagery dataset with 4 classes and 880 images was chosen as the dataset. AlexNet, ResNet50, ResNet101, ResNet50V2, ResNet101V2, InceptionV3, Xception and InceptionResNetV2 deep learning models were selected for classification. The highest success value was obtained fromInceptionResNetV2 architecture with 94% and the fastest result was obtained from AlexNet architecture with 438 seconds. With this approach, it has been shown that diseases of the eardrum can be treated with an autonomous system, freeing from expert error. In the future, such a system in the clinical field will be able to reduce labor and error.
Sinir ağı dil modelleri ve evrensel cümle kodlayıcı kullanarak havayolu müşteri yorumlarının duygu analizi
(Fırat Üniversitesi Müh. Bil. Dergisi, 2024-04-10) Saka, Semih Osman; Cömert, Zafer
Duygu analizi, metin tabanlı verilerin duygusal tonlarını belirlemede kullanılan önemli bir doğal dil işleme (NLP) tekniğidir. İşletmeler, müşteri memnuniyetini artırmak ve hizmet kalitesini iyileştirmek için müşteri yorumlarından elde edilen duygusal içgörülerle stratejik kararlar alabilir. Bu çalışmada ise havayolu müşteri yorumlarından oluşan bir veri seti kullanılmıştır. Veri seti, her bir yorumun doğrulama durumu, içerik, değerlendirme puanı, öneri durumu, duygu analizini içermektedir ve toplamda 1100 örnekten oluşmaktadır. Çalışmada dört farklı model incelenmiştir. Bu modellerden CNNLSTM, sıfırdan öğrenme stratejisiyle eğitilmiştir. Ayrıca, iki farklı sinir ağı dil modeli (Neural Network Language Model, NNLM) ve evrensel cümle kodlayıcı (Universal Sentence Encoder,USE) transfer öğrenme yaklaşımıyla eğitilmiştir. CNNLSTM modeli %92,06 doğruluk oranı ile yüksek performans göstermiştir. nnlm-en-dim50 modeli %90,87 doğruluk oranı elde ederken, nnlm-en-dim128 modeli %92,46 doğruluk oranı ileöne çıkmıştır. En yüksek performansı ise %95,63 doğruluk oranı ile USE modeli göstermiştir. Bu sonuçlar, derin öğrenme ve transfer öğrenme tekniklerinin duygu analizinde etkili araçlar olduğunu göstermektedir. Çalışma, işletmelerin müşteri memnuniyetini artırmak ve hizmet kalitesini iyileştirmek için duygu analizi teknolojilerini nasıl etkin bir şekilde kullanabileceklerine dair önemli içgörüler sunmaktadır. Gelecek çalışmalarda, farklı veri setleri ve daha geniş örneklem büyüklükleri ile modellerin performanslarının daha detaylı incelenmesi önerilmektedir. Sentiment analysis represents a fundamental technique within the domain of natural language processing (NLP), employed for the purpose of discerning the emotional tenor of text-based data. Businesses may make strategic decisions by employing emotional insights derived from customer reviews to enhance customer satisfaction and improve service quality. In this study, a dataset comprising reviews from airline customers was employed. The dataset comprises the verification status, content, rating score, recommendation status, and sentiment analysis of each review, with a total of 1100 examples. The study examined four distinct models. The CNN-LSTM model was trained using a training-from-scratch strategy. Furthermore, two distinct neural network language models (Neural Network Language Model, NNLM) and the Universal Sentence Encoder (USE) were trained using a transfer learning approach. The CNN-LSTM model exhibited robust performance, achieving an accuracy rate of 92.06%. The nnlm-en-dim50 model achieved an accuracy rate of 90,87%, while the nnlm-en-dim128 model demonstrated a notably higher level of accuracy at 92.46%. The USE model exhibited the highest performance, with an accuracy rate of 95.63%. These findings suggest that deep learning and transfer learning techniques are effective tools for sentiment analysis. The study offers valuable insights into how businesses can utilize sentiment analysis technologies to enhance customer satisfaction and service quality. It is recommended that future studies investigate the performance of these models with different datasets and larger sample sizes.
Otitis media için evrişimsel sinir ağlarına dayalı bütünleşik bir tanı sistemi
(BEÜ Fen Bilimleri Dergisi, 2019-11-18) Cömert, Zafer
Otitis media (OM) bir dizi iltihaplı orta kulak rahatsızlıklarını temsil eden tıbbi bir kavramdır. OM dünya genelinde, özellikle çocukluk çağında, görülen en yaygın hastalıklardan biridir. Klinik pratikte OM tanısı, otoskop cihazıyla elde edilen orta kulak görüntüsünün kulak buran boğaz uzmanları tarafından incelenmesiyle gerçekleştirilir. İncelemenin sübjektif olarak yapılması, gözlemciler arasında değişkenliklerin ortaya çıkmasına neden olmaktadır. Aynı zamanda, bu alanda bilgisayar destekli sistemlerinin kullanımının da yeteri kadar yaygın olmadığı görülmektedir. OM rahatsızlıklarının zamanında teşhis edilememesi, hastalıkların ilerlemesine ve buna bağlı olarak da işitme, konuşma ve bilişsel rahatsızlıkların ortaya çıkmasına nedenolmaktadır. Tüm bu dezavantajların üstesinden gelmek üzere, bu çalışmada OM teşhisi için önceden eğitilmiş evrişimsel sinir ağlarına dayalı bütünleşik bir tanı sistemi önerilmiştir. Deneysel çalışmalar, Özel Van Akdamar Hastanesinde gönüllü hastalardan toplanan ve toplamda beş farklı sınıfı temsil eden 898 adet otoskop imgeleri üzerinde gerçekleştirilmiştir. Sonuç olarak, önerilen model %82.16 sınıflandırma başarısı sağlanmıştır. Evrişimsel sinir ağlarına dayalı önerilen modelin sağladığı uçtan uca öğrenme ve yüksek hassasiyetle, OM teşhisinin objektif bir şekilde yapılabilmesi ve tanı sürecinde hekimlerin karar verme sürecinin desteklenmesi sağlanabilir. Önerilen yöntem bu açılardan umut verici sonuçlar üretmiştir. Otitis media (OM) is a medical concept representing a range of inflammatory middle ear disorders. OM is one of the most common diseases worldwide, especially in childhood. In clinical practice, the diagnosis of OM is carried out by examining the images of the middle ear obtained via the otoscope device by specialists. The subjective examination leads to arise the variabilities among observers. At the same time, the use of computer-aided systems in this area is not common enough. Failure to diagnose OM disorders in a timely manner leads to the progression of the diseases, the emergence of hearing, speech, and cognitive disorders. To overcome all these disadvantages, an integrated diagnostic system based on the pretrained deep convolutional neural networks is proposed for the diagnosis of OM in this study. Experimental studies were carried out on 898 otoscope images, representing five different classes, collected from volunteer patients admitted to Özel Van Akdamar Hospital. As a result, the proposed model achieved 82.16% classification success. With the end-to-end learning and high sensitivity provided by the proposed model based on convolutional neural networks, OM diagnosis can be realized objectively and physicians' decision-making process can be supported using this system. The proposed method has produced promising results in these respects.
Detection of weather images by using spiking neural networks of deep learning models
(Neural Computing and Applications, 2020-10-24) Toğaçar, Mesut; Ergen, Burak; Cömert, Zafer
The transmission of weather information of a location at certain time intervals affects the living conditions of the people there directly or indirectly. According to weather information, people shape their behavior in daily life. Besides, agricultural activities are carried out according to the weather conditions. Considering the importance of this subject, it is possible to make weather predictions based on the weather images in today’s technology exploiting the computer systems. However, the recent mention of the name of artificial intelligence technology in every field has made it compulsory for computer systems to benefit from this technology. The dataset used in the study has four classes: cloudy, rain, shine, and sunrise. In the study, GoogLeNet and VGG-16 models and the spiking neural network (SNN) were used together. The features extracted from GoogLeNet and VGG-16 models were combined and given to the SNNs as the input. As a result, the SNNs contributed to the success of classification with the proposed approach. The classification accuracy rates of cloudy, rain, shine, and sunrise classes were 98.48%, 97.58%, 97%, and 98.48%, respectively, together with SNN. Also, the use of SNNs in combination with deep learning models to obtain a successful result is proved in this study.
Enhancing of dataset using DeepDream, fuzzy color image enhancement and hypercolumn techniques to detection of the Alzheimer’s disease stages by deep learning model
(Neural Computing and Applications, 2021-01-20) Toğaçar, Mesut; Cömert, Zafer; Ergen, Burhan
Alzheimer’s disease (AD), which occurs as a result of the loss of cognitive functions in the brain, causes near-forgetfulness in the case and dementia in subsequent processes. Dataset consists of MR images containing four phases of AD. The dataset was re-enhanced separately with DeepDream, fuzzy color image enhancement, hypercolumn techniques. Visual Geometry Group-16 (VGG-16) deep learning model is used in the enhancing process and deep features are combined. Linear Regression is used for the selection of efficient features. The Support Vector Machine is preferred as a classifier With the proposed approach, the classification achievement was obtained as 100% in Mild Dementia, 99.94% in Moderate Dementia, 100% in non-Dementia, 99.94% in Very Mild Dementia. The overall accuracy was 99.94%. The proposed approach increased the prediction success in detecting Alzheimer’s stages by re-enhancing MR images. Thus, an efficient early diagnosis model was realized at an affordable cost for individuals likely to progress with dementia.
Efficient approach for digitization of the cardiotocography signals
(Physica A, 2019-04-23) Cömert, Zafer; Şengür, Abdulkadir; Akbulut, Yaman; Budak, Ümit; Kocamaz, Adnan Fatih; Bajaj, Varun
Cardiotocography (CTG) is generally provided on printed traces, and digitization of CTG signal is important for forthcoming assessments. In this paper, a new algorithm relies on the box-counting method is offered for the digitization of the CTG signals from CTG printed traces. The introduced algorithm inputs the CTG printed traces and outputs the digital fetal heart rate (FHR) and uterine contraction (UC) signals. The proposed method initially extracts the CTG signal image and gridded background image. Retrieving of the FHR and UC signals on the gridded background disrupts the background grids. So, we employ an algorithm to fix the degraded lines in the gridded background. After the line fixing operation, the boxes in the horizontal and vertical axes are counted for determining the calibration parameters. A set of specific equations are used to determine the calibration parameters. The signal extraction is performed on by redchannel thresholding of input CTG printing images. An open-access CTG intrapartum database comprises 552 samples is used in the experiment. As a result, the average correlation coefficients of FHR and UC signals are 0.9811 ± 0.0251 and 0.9905 ± 0.0126, respectively.
Neighbourhood component analysis and deep feature-based diagnosis model for middle ear otoscope images
(Neural Computing and Applications, 2021-11-26) Başaran, Erdal; Cömert, Zafer; Çelik, Yüksel
Otitis media (OM), known as inflammation of the middle ear, is a condition especially seen in children. To carry out a definitive diagnosis of the discomfort that manifests itself with various symptoms such as pain in the ear, fever, and discharge, the eardrum in the middle ear should be examined by a specialist. In this study, a convolution neural network was used for feature extraction from middle ear otoscope images to diagnose different types of OM. These features were extracted using AlexNet, VGG-16, GoogLeNet, ResNet-50 models. The deep features extracted from these models were combined into a new deep feature vector. This feature vector consisting of 4000 deep features was examined, and the most relevant 222 deep features were selected from this large feature set by using the neighbourhood component analysis. In this case, the number of features was decreased and a more effective feature set was obtained. In the next stage of this experimental study, this new feature set was applied as the input to the support vector machine. As a result of the experimental study, an accuracy rate of 79.02% was achieved. The results point out that the use of deep features in detecting OM provides efficient results, and the proposed approach is beneficial in reducing the number of deep features as well as achieving better classification results.
Waste classification using AutoEncoder network with integrated feature selection method in convolutional neural network models
(Measurement, 2019-12-27) Toğaçar, Mesut; Ergen, Burak; Cömert, Zafer
Unless adequate measures are taken for waste litter, the ecological balance may deteriorate over time. The wastes disposed of the trash can be divided into two classes that are organic and recycling types. In recent years, artificial intelligence is frequently mentioned in all areas of technology. In this study, the dataset used for the classification of waste is reconstructed with the AutoEncoder network. The feature sets are then extracted using two datasets by Convolutional Neural Network (CNN) architectures and these feature sets are combined. The Ridge Regression (RR) method performed on the combined feature set reduced the number of features and also revealed the efficient features. Support Vector Machines (SVMs) were used as classifiers in all experiments. The most successful classification accuracy in the experiments was 99.95%. In this study, it is seen that the proposed approach is successful in the classification of waste types.
Permütasyonlu otoregresif dizi modeli ve YOLOv8 kullanarak sahne metni tanıma
(2024) Arı, Berna Gürler; Cömert, Zafer
Metin, yüksek düzeyli bilgi çıkarımı için en güçlü kaynaklardan biri olarak görülür. Bilgisayarlı görüde sıcak bir çalışma alanı olan sahne metni tanıma, önce metin bölgelerini tespit etme ardından bu bölgelerdeki metnin analizine dayanmaktadır. Ülkemizde her kademede ölçme-değerlendirme için en sık kullanılan yöntem olan çoktan seçmeli soruları ve soru kağıdının analizini sahne metni tanıma amacıyla araştırmamızın hedefine almaktayız. Farklı kişi ve işaretleme stilleriyle oluşturulan veriseti üzerinde deneyler yapılırken üç ana prosedür gerçekleştirilmiştir. Soru kağıdı üzerinde gerçekleştirilen bu hedefler sırasıyla; YOLOv8 ile soru tespitinin yapılması, YOLOv8 ile seçenek tespitinin yapılması ve Permütasyonlu Otoregresif Dizi Modeli ile soru numarası tespitinin yapılmasıdır. Gerçek veriler üzerinden elde edilen model doğruluğu, ölçümler sonucunda çalışmanın uygulanabilirliğini göstermektedir. Text is considered one of the most powerful sources for high-level information extraction. Scene text recognition, a hot field of study in computer vision, is based on first detecting text regions and then analyzing the text in these regions. We target multiple choice questions and question paper analysis, which are the most frequently used methods for measurement and evaluation at all levels in our country, as the target of our research. Three main goals were achieved while conducting experiments on the dataset created with different people and marking styles. These objectives realized on the question paper are respectively; Determining the question with YOLOv8, determining the option with YOLOv8, and determining the question number with the Permutated Autoregressive Sequence Model. The model accuracy obtained from real data shows the feasibility of the study as a result of the measurements.
Timpanik membran görüntü özellikleri kullanılarak sınıflandırılması
(Fırat Üniversitesi Müh. Bil. Dergisi, 2021-06-03) Başaran, Erdal; Cömert, Zafer; Çelik, Yüksel
Orta kulak inflamasyonu olarak bilinen otitis media rahatsızlığının teşhis edilmesi için otoskop cihazı ile zar bölgesine bakılarak karar verilmektedir. Dokusal özellik çıkarma algoritmaları, görüntüler üzerinde bölge tespiti ve görüntüye ait özelliklerin elde edilmesinde yaygın olarak kullanılmaktadır. Bu çalışmada gerekli yasal izinler alındıktan sonra elde edilen orta kulak görüntülerinde normal ve otitis media görüntülerinin ayırt edilmesi için literatürde yaygın olarak kullanılan gri seviyeli eş-oluşum matrisi, yerel ikili örüntüler, yönlü gradyanların histogram algoritmaları kullanılmıştır. Bu dokusal özellik çıkarma algoritmalarının görüntüleri sınıflandırma üzerinde başarıları incelendikten sonra her bir özellik setine görüntülere ait renk kanallarının ortalamaları da eklenerek bu özelliğin sınıflandırma başarısına etkisi incelenmiştir. Sonuç olarak tek başına bir dokusal özellik çıkarma algoritması kullanıldığında en iyi sonuçlar yerel ikili örüntü algoritması ile elde edilmiştir. Bu algoritmaya renk kanallarının ortalaması da eklendiği zaman sınıflandırma başarısını olumlu yönde etkilediği sonucuna varılmıştır. Sınıflandırma sonucunda % 78.67 doğruluk oranı elde edilmiştir. In order to diagnose otitis media, known as middle ear inflammation, is made by looking at the membrane area with an otoscope device. Textural feature extraction algorithms are widely used to detect regions on images and to obtain image properties. In this study, gray-level co-occurance matrix, local binary patterns, histogram of oriented gradients, which are widely used in the literature, were used to distinguish normal and otitis media images in middle ear images obtained after obtaining the necessary legal permissions. After examining the success of these textural feature extraction algorithms on image classification, the averages of the color channels of the images were added to each feature set, and theeffect of this feature on the classification success was examined. As a result, when using a textural feature extraction algorithm alone, the best results were obtained with the local binary pattern algorithm. When the average of color channels is added to this algorithm, it is concluded that it affects the classification success positively. As a result of the classification, an accuracy rate of 78.67% was obtained.
Predicting fetal hypoxia using common spatial pattern and machine learning from cardiotocography signals
(Applied Acoustics, 2020-05-13) Alsaggaf, Wafaa; Cömert, Zafer; Nour, Majid; Polat, Kemal; Brdesee, Hani; Toğaçar, Mesut
Cardiotocography (CTG) is a screening tool used in daily obstetric practice to determine fetal wellbeing. Its interpretation is generally performed visually by the field experts, and this visual inspection is an error-prone and subjective process. In addition, it leads to several drawbacks, such as variability among the observers and low reproducibility rates. To tackle these drawbacks, a novel computer-aided diagnostic (CAD) model is proposed. As novel diagnostic indices, the features provided by the common spatial patterns (CSP) were considered in this study. The experiments were carried out on a publicly available CTU-UHB Intrapartum CTG database. Four different data division criteria were evaluated individually. The proposed model relied upon a combination of the conventional as well as the CSP features and machine learning models such as an artificial neural network (ANN), support vector machine (SVM), and k-nearest neighbor (kNN). To validate the successes of the models, the five-fold cross-validation method was employed. The results validated that the CSP features ensured an increase in the performances of the machine learning models in the fetal hypoxia detection task. Also, the most effective results were provided by the SVM classifier with an accuracy of 94.75%, a sensitivity of 74.29% and a specificity of 99.55%. Consequently, thanks to the proposed model, a novel, consistent, and robust diagnostic model ensured for predicting fetal hypoxia.
Machine learning approach equipped with neighbourhood component analysis for DDoS attack detection in software-defined networking
(Electronics, 2021-05-19) Tonkal, Özgür; Polat, Hüseyin; Başaran, Erdal; Cömert, Zafer; Kocaoğlu, Ramazan
The Software-Defined Network (SDN) is a new network paradigm that promises more dynamic and efficiently manageable network architecture for new-generation networks. With its programmable central controller approach, network operators can easily manage and control the whole network. However, at the same time, due to its centralized structure, it is the target of many attack vectors. Distributed Denial of Service (DDoS) attacks are the most effective attack vector to the SDN. The purpose of this study is to classify the SDN traffic as normal or attack traffic using machine learning algorithms equipped with Neighbourhood Component Analysis (NCA). We handle a public “DDoS attack SDN Dataset” including a total of 23 features. The dataset consists of Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and Internet Control Message Protocol (ICMP) normal and attack traffics. The dataset, including more than 100 thousand recordings, has statistical features such as byte_count, duration_sec, packet rate, and packet per flow, except for features that define source and target machines. We use the NCA algorithm to reveal the most relevant features by feature selection and perform an effective classification. After preprocessing and feature selection stages, the obtained dataset was classified by k-Nearest Neighbor (kNN), Decision Tree (DT), Artificial Neural Network (ANN), and Support Vector Machine (SVM) algorithms. The experimental results show that DT has a better accuracy rate than the other algorithms with 100% classification achievement.
Normal ve kronik hastalıklı orta kulak imgelerinin evrişimsel sinir ağları yöntemiyle tespit edilmesi
(Türkiye Bilişim Vakfı Bilgisayar Bilimleri ve Mühendisliği Dergisi, 2020-03-30) Başaran, Erdal; Cömert, Zafer; Şengür, Abdulkadir; Budak, Ümit; Çelik, Yüksel; Toğaçar, Mesut
Orta kulak iltihabı kulak zarının arkasında sıvı birikmesi olarak bilinmektedir. Orta kulak iltihabının uzun süreli tedaviye yanıt vermemesi ve kulak zarının delinmesi ile karakterize olan kronik orta kulak iltihabı işitme kaybına bile sebep olabilen ciddi bir rahatsızlıktır. Bu çalışmada gerekli etik kurulu izni alındıktan sonra Özel Van Akdamar Hastanesinde gönüllü hastalardan otoskop cihazı ile elde edilen 598 adet normal orta kulak görüntüsü ve kronik hastalıklı orta kulak görüntüleri ile sınıflandırma işlemi gerçekleştirilmiştir. Son yıllarda yapay zekâ kapsamında değerlendirilen algoritmalar hemen her alanda kullanılmaktadır. Sağlık alanında da tanı ve karar destek sistemleri geliştirilerek başarılı çalışmalar yapılmaktadır. Bu çalışmada yapay zekâ algoritmalarından olan ve özellikle biyomedikal görüntü sınıflandırma çalışmalarında da iyi sonuçlar elde edilen evrişimsel sinir ağı mimarilerinden olan AlexNet, VGG16, VGG19, GoogleNet, ResNet18, ResNet50, ResNet101 modelleri kullanılmıştır. Deneysel çalışmalar sonucu VGG19 mimarisi ile %97.2067 başarı oranı elde edilmiştir. Evrişimsel sinir ağları yöntemi normal ve kronik orta kulak görüntülerini ayırt etmede başarılı bir yöntemdir. Middle ear inflammation is known as the accumulation of fluid behind the eardrum. Chronic middle ear inflammation, which is characterized by the failure to respond to long-term treatment and perforation of the eardrum, is a serious condition that can even cause hearing loss. In this study, 598 normal middle ear images obtained from the volunteer patients with otoscope device and middle ear images with chronic disease were performed after obtaining the necessary ethics committee permission. In recent years, algorithms evaluated within the scope of artificial intelligence have been used in almost every field. In the field of health, diagnostic and decision support systems are developed and successful studies are carried out. In this study, AlexNet, VGG16, VGG19,GoogleNet, ResNet18, ResNet50, ResNet101modelswhich are one of the artificial intelligence algorithms and the convolutional neural network architectures which have good results especially in biomedical image classification studies are used. As a result of the experimental studies, 97.2067% success rate was achieved with VGG19 architecture. The convolutional neural network method is a successful method to distinguish between normal and chronic middle ear images.
Convolutional neural network approach for automatic tympanic membrane detection and classification
(Biomedical Signal Processing and Control, 2019-10-13) Başaran, Erdal; Cömert, Zafer; Çelik, Yüksel
Otitis media (OM) is a term used to describe the inflammation of the middle ear. The clinical inspection of the tympanic membrane is conducted visually by experts. Visual inspection leads to limited variability among the observers and includes human-induced errors. In this study, we sought to solve these problems using a novel diagnostic model based on a faster regional convolutional neural network(Faster RCNN) fortympanic membrane detection,and pretrainedCNNs for tympanic membrane classification. The experimental study was conducted on a neweardrum dataset. The Faster R-CNN was initially applied to the original images. The number of images in the dataset was subsequently increased using basic image augmentation techniques such as flip and rotation. We also evaluated the success of the model in the presence of various noise effects. The original and automatically extracted tympanic membrane patches were finally input separately to the CNNs. The AlexNet, VGGNets, GoogLeNet, and ResNets models were employed. This resulted in an average precision of 75.85% in the tympanic membrane detection. All CNNs in the classification produced satisfactory results, with the proposed approach achieving an accuracy of 90.48% with the VGG-16 model. This approach can potentially be used in future otological clinical decision support systems to increase the diagnostic accuracy of the physicians and reduce the overall rate of misdiagnosis. Future studies will focus on increasing the number of samples in the eardrum dataset to cover a full range of ontological conditions. This would enable us to realize a multi-class classification in OM diagnosis.
DeepMaizeNet: A novel hybrid approach based on CBAM for implementing the doubled haploid technique
(Special Section: Machine Learning in Agriculture, 2023-05-18) Ayaz, İbrahim; Kutlu, Fatih; Cömert, Zafer
Maize (Zea mays L.) is an important cereal plant in the family of wheatgrass cultivated all over the world. With the increase in human population and environmental factors, the need for maize plants is increasing day by day. One of the efficient methods of increasing production of the maize is maize breeding. The most effective and rapid method for maize breeding is the doubled haploid (DH) technique. This technique reduces maize breeding time and increases productivity. There are different selection methods to select haploid maize seeds in the maize breeding process. Among these selection methods, the most common and most successful selection method is the visual checking of the R1-Navajo marker. Maize seed separation by hand is a time-consuming and error-prone operation. It is labor-intensive and very tiring; therefore, it is essential to develop a fast and highly accurate intelligent system that separates diploid and haploid maize seeds from each other. This study presents a pioneering approach, introducing the DeepMaizeNet, a hybrid deep learning model that showcasesitsprowessinaccuratelyclassifyinghaploidmaizeseeds.Theclassification of haploidseedsholdssignificant valuefortheDHtechnique,andtheproposed model’s success is a promising step towardenhanced efficiency. The proposed hybrid model exploits some new techniques such as convolution block attention module, hypercolumn, 2Dupsampling,andresidualblock.Fortheassessmentoftheproposed model, the five-fold cross-validation technique is employed. The result shows that DeepMaizeNet provides a promising performance by achieving 94.13% accuracy, 94.91% F1-score, and 97.27% sensitivity.
COVID-19 detection using deep learning models to exploit Social Mimic Optimization and structured chest X-ray images using fuzzy color and stacking approaches
(Computers in Biology and Medicine, 2020-05-02) Toğaçar, Mesut; Ergen, Burhan; Cömert, Zafer
Coronavirus causes a wide variety of respiratory infections and it is an RNA-type virus that can infect both humans and animal species. It often causes pneumonia in humans. Artificial intelligence models have been helpful for successful analyses in the biomedical field. In this study, Coronavirus was detected using a deep learning model, which is a sub-branch of artificial intelligence. Our dataset consists of three classes namely: coronavirus, pneumonia, and normal X-ray imagery. In this study, the data classes were restructured using the Fuzzy Color technique as a preprocessing step and the images that were structured with the original images were stacked. In the next step, the stacked dataset was trained with deep learning models (MobileNetV2, SqueezeNet) and the feature sets obtained by the models were processed using the Social Mimic optimization method. Thereafter, efficient features were combined and classified using Support Vector Machines (SVM). The overall classification rate obtained with the proposed approach was 99.27%. With the proposed approach in this study, it is evident that the model can efficiently contribute to the detection of COVID-19 disease.

Browse

Recent Submissions