Bilgisayar ve Bilişim Fakültesi Koleksiyonu

Permanent URI for this collectionhttps://hdl.handle.net/20.500.13091/10834

Browse

Browsing Bilgisayar ve Bilişim Fakültesi Koleksiyonu by Publisher "Elsevier Sci Ltd"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Article
    Citation - WoS: 4
    Citation - Scopus: 5
    Identification of Full-Night Sleep Parameters Using Morphological Features of Ecg Signals: a Practical Alternative To Eeg and Eog Signals
    (Elsevier Sci Ltd, 2024) Yücelbaş, Şule; Yücelbaş, Cüneyt; Tezel, Gülay; Özsen, Seral; Yosunkaya, Şebnem
    Electroencephalogram (EEG) signals, which are among the most important recordings used in Polysomnography for sleep staging, are more challenging and demanding than electrocardiography (ECG) signals, both in terms of acquisition and interpretation. When examining the studies of other researchers on sleep parameters in the literature, it is evident that EEG signals are predominantly used for determining arousal (AR), K-complex (Kc), and sleep spindle (Ss) parameters. Furthermore, it is understood that electrooculography (EOG) signals are employed for detecting slow eye movements (SEM) and rapid eye movements (REM) parameters.This study is a continuation of our previous research, where we used only EEG signals for Kc and Ss detection. In this study, an approach that includes ECG signals in the determination of sleep parameters to bring practicality to sleep staging studies was adopted. For this purpose, firstly, 16 morphological features were extracted from ECG recordings taken from a total of 24 subjects after various preprocessing steps. Subsequently, these data were used to work on the detection of five different sleep parameters: AR, Kc, Ss, SEM, and REM, using the Random Subspace (RaSE) ensemble learning algorithm. The results were calculated according to various statistical criteria and a classification accuracy of over 78 % was obtained in all parameters. As a result, the sleep parameters that could be determined most successfully using the ECG signal were SEM and arousal, respectively. In addition, feature elimination was performed for these datasets using Symmetric Uncertainty (SU) ranking. As a result of the reclassification process using 9 and 12 features, the effectiveness of which was determined for both datasets, respectively, significant increases were observed in the performance outputs. Experimental results have shown that ECG signals can be used as an alternative to EEG and EOG signals in the determination of full-night sleep parameters.
  • Thumbnail Image
    Article
    Citation - WoS: 6
    Citation - Scopus: 7
    Improving Efficiency in Convolutional Neural Networks With 3d Image Filters
    (Elsevier Sci Ltd, 2022) Uyar, Kübra; Taşdemir, Şakir; Ülker, Erkan; Ünlükal, Nejat; Solmaz, Merve
    Background and objective: The effective performance of deep networks has provided the solution to various stateof-the-art problems. Convolutional Neural Network (CNN) is accepted as an accurate, effective, and reliable practice in image-based applications. However, there is a need to use pre-trained models in case of insufficient data in CNN. This study aims to present an alternative solution to this problem with the proposed 3D image based filter generation approach with simpler CNNs for the classification of small datasets. Methods: In this study, a novel 3D image filters-based CNN (Hist3DCNN) is proposed. The proposed filter generation approach is based on 3D object images taken from different perspectives. The efficiency of Hist3DCNN is shown on a novel histological dataset that contains blood, connective, epithelium, muscle, and nerve tissue images. Various case studies are carried out with generated filters assigned as the initial value to AlexNet and the designed Hist3DCNN model that is simpler than AlexNet. Results: Based on results, the classification accuracy of AlexNet with proposed filters used in convolution layers were 84.65% and 85.34%. The accuracy was increased to 85.47% by Hist3DCNN on the histological image classification. Moreover, four different benchmark datasets were tested to demonstrate the robustness of Hist3DCNN on various datasets. Conclusions: This study provides a new aspect to literature due to 3D image-based filter generation approach to initialize convolution filters. Experimental results validate that Hist3DCNN can be used as a filter value initialization method with simple CNN models that contain less learnable parameters for the classification task of small datasets.