Sürdürülebilir Kalkınma Amaçları

• 4

Yazarlar

• Acharya, U. Rajendra(2)
• Dogan, Sengul(2)
• Muezzinoglu, Taha(2)
• Talo, Muhammed(2)
• Tuncer, Turker(2)
• Yildirim, Ozal(2)
• Acharya, U Rajendra(1)
• Ay, B.(1)
• Aydin, G.(1)
• Baloglu, Ulas Baran(1)
• Barua, Prabal Datta(1)
• Baygin, Mehmet(1)
• Baygin, Nursena(1)
• Celik, Yusuf(1)
• Cheong, Kang Hao(1)
• Gul, Yeliz(1)
• Karabatak, Murat(1)
• Kilicarslan, Gulhan(1)
• Palmer, Elizabeth Emma(1)
• Tuncer, Ilknur(1)

• MakaleAçık Erişim2023Scopus SJRQ1/r/munzur/1681
  PatchResNet: Multiple Patch Division–Based Deep Feature Fusion Framework for Brain Tumor Classification Using MRI Images

  Muezzinoglu, Taha; Baygin, Nursena; Tuncer, Ilknur; Barua, Prabal Datta; Baygin, Mehmet; Dogan, Sengul; Tuncer, Turker; Palmer, Elizabeth Emma; Cheong, Kang Hao; Acharya, U. Rajendra

  Modern computer vision algorithms are based on convolutional neural networks (CNNs), and both end-to-end learning and transfer learning modes have been used with CNN for image classification. Thus, automated brain tumor classification models have been proposed by deploying CNNs to help medical professionals. Our primary objective is to increase the classification performance using CNN. Therefore, a patch-based deep feature engineering model has been proposed in this work. Nowadays, patch division techniques have been used to attain high classification performance, and variable-sized patches have achieved good results. In this work, we have used three types of patches of different sizes (32 × 32, 56 × 56, 112 × 112). Six feature vectors have been obtained using these patches and two layers of the pretrained ResNet50 (global average pooling and fully connected layers). In the feature selection phase, three selectors—neighborhood component analysis (NCA), Chi2, and ReliefF—have been used, and 18 final feature vectors have been obtained. By deploying k nearest neighbors (kNN), 18 results have been calculated. Iterative hard majority voting (IHMV) has been applied to compute the general classification accuracy of this framework. This model uses different patches, feature extractors (two layers of the ResNet50 have been utilized as feature extractors), and selectors, making this a framework that we have named PatchResNet. A public brain image dataset containing four classes (glioblastoma multiforme (GBM), meningioma, pituitary tumor, healthy) has been used to develop the proposed PatchResNet model. Our proposed PatchResNet attained 98.10% classification accuracy using the public brain tumor image dataset. The developed PatchResNet model obtained high classification accuracy and has the advantage of being a self-organized framework. Therefore, the proposed method can choose the best result validation prediction vectors and achieve high image classification performance.

  • 
  AtıfScopus103·OpenAlex101

  İncele
• MakaleAçık Erişim2023Scopus SJRQ2/r/munzur/1587
  Application of the deep transfer learning framework for hydatid cyst classification using CT images

  Gul, Yeliz; Muezzinoglu, Taha; Kilicarslan, Gulhan; Dogan, Sengul; Tuncer, Turker

  Hydatid cyst, which causes important public health problems, is a disease frequently reported by radiologists. Classification of hydatid cyst types is very important to give the most suitable cure to patients. In order to decrease the misclassification rates of these cysts, machine learning is a good solution to create intelligent medical assistants. The primary objective of this work is to develop a new image classification framework to classify hydatid cyst types. Firstly, we collected a computed tomography (CT) image dataset. This CT image dataset consists of 2416 CT images with five classes (hydatid cyst types). We proposed a new deep feature engineering dataset framework for this dataset. This framework consists of (i) deep feature extraction by deploying transfer learning, (ii) iterative feature selection using INCA (iterative neighborhood component analysis), and (iii) classification. In the first phase—the feature extraction phase—we extracted deep features by deploying ten pretrained convolutional neural networks (CNNs). The used CNNs are commonly used CNNs, and the deep features have been extracted using the last pooling and fully connected layers of these networks. INCA has been applied to the generated deep features by each feature vector. k-nearest neighbors (kNN) classifier has been utilized to generate predicted values of these 10 CNNs using the selected features by INCA. Our proposed INCA-based deep feature extractor attained over 92% classification accuracy for each pretrained model. Our proposed framework reached over 92% classification accuracies, clearly demonstrating that our proposed framework is good image classification.

  • 
  AtıfScopus10·WoS5·OpenAlex8

  İncele
• MakaleAçık Erişim2020Scopus SJRQ1/r/munzur/3456
  Automated invasive ductal carcinoma detection based using deep transfer learning with whole-slide images

  Celik, Yusuf; Talo, Muhammed; Yildirim, Ozal; Karabatak, Murat; Acharya, U Rajendra

  Advances in artificial intelligence technologies have made it possible to obtain more accurate and reliable results using digital images. Due to the advances in digital histopathological images obtained using whole slide image (WSI) scanners, automated analysis of digital images by computer support systems has become interesting. In particular, deep learning architectures, are one of the preferred approaches in the analysis of digital histopathology images. The deeper networks trained on large amounts of image data are adapted for different tasks using transfer learning technique. In this study, automated detection of invasive ductal carcinoma (IDC), which is the most common subtype of breast cancers, is proposed using deep transfer learning technique. We have used deep learning pre-trained models, ResNet-50 and DenseNet-161 for the IDC detection task. The public histopathology dataset containing 277,524 image patches were used in our experimental studies. As a result of training on the last layers of pre-trained deep networks, DenseNet-161 model has yielded F-sore of 92.38% and balanced accuracy value of 91.57%. Similarly, we have obtained F-score of 94.11% and balanced accuracy value of 90.96% using ResNet-50 architecture. In addition, our developed model is validated using the publicly available BreakHis breast cancer dataset and obtained promising results in classifying magnification independent histopathology images into benign and malignant classes. Our developed system obtained the highest classification performance as compared to the state-of-art techniques and is ready to be tested with more diverse huge databases.

  • 
  AtıfScopus297·OpenAlex318

  İncele
• MakaleAçık Erişim2019Scopus SJRQ2/r/munzur/3255
  Automated detection of diabetic subject using pre-trained 2D-CNN models with frequency spectrum images extracted from heart rate signals

  Yildirim, Ozal; Talo, Muhammed; Ay, B.; Baloglu, Ulas Baran; Aydin, G.; Acharya, U. Rajendra

  In this study, a deep-transfer learning approach is proposed for the automated diagnosis of diabetes mellitus (DM), using heart rate (HR) signals obtained from electrocardiogram (ECG) data. Recent progress in deep learning has contributed significantly to improvement in the quality of healthcare. In order for deep learning models to perform well, large datasets are required for training. However, a difficulty in the biomedical field is the lack of clinical data with expert annotation. A recent, commonly implemented technique to train deep learning models using small datasets is to transfer the weighting, developed from a large dataset, to the current model. This deep learning transfer strategy is generally employed for two-dimensional signals. Herein, the weighting of models pre-trained using two-dimensional large image data was applied to one-dimensional HR signals. The one-dimensional HR signals were then converted into frequency spectrum images, which were utilized for application to well-known pre-trained models, specifically: AlexNet, VggNet, ResNet, and DenseNet. The DenseNet pre-trained model yielded the highest classification average accuracy of 97.62%, and sensitivity of 100%, to detect DM subjects via HR signal recordings. In the future, we intend to further test this developed model by utilizing additional data along with cloud-based storage to diagnose DM via heart signal analysis.

  • 
  AtıfScopus145·OpenAlex152

  İncele