Polycystic ovary syndrome is a hormonal disorder seen in many women. It occurs by the combination of many small and benign cysts in the ovaries. These cysts, called follicles, create a special pattern in the ovaries observed with ultrasound imaging. The number, structure, and size of these follicles provide important information for the diagnosis of ovarian diseases. In this study, two different methods of follicle detection are tested for Polycystic Ovary Syndrome. The first method consists of noise filtering, contrast adjustment, binarization, and morphological processes. For this method, Median Filter, Average Filter, Gaussian Filter, and Wiener Filter were used for noise reduction, and then histogram equalization and adaptive thresholding were tested. For the second method, Gaussian Filter and Wavelet Transform were selected for noise reduction, and k-means clustering and morphological operations were applied to the images. In the segmentation phase performed for both methods, follicles were detected with the Canny Edge Detection algorithm. False Acceptance Rate (FAR) and False Rejection Rate (FRR) were used to evaluate the accuracy of the results. Our results show that the most accurate follicle detection was obtained by using the Wiener Filter and Gaussian Filter.

The aim of this study was to determine the best method for follicle detection using ovarian ultrasound images and to classify the ultrasound images as pcos or normal ovaries using the proposed CNN architecture. Two different methods for follicle detection have been proposed to evaluate pcos. For this purpose, the Median, the Mean, the Wiener, and the Gaussian filters were tested using standard and adaptive thresholds. Second, Gaussian filtering, Discrete Wavelet Transform, and k-means clustering algorithms were tested. The Canny operator separates follicles from the background in the segmentation phase. In this study, a CNN architecture that classifies limited ultrasound ovary images was developed, and its success in the best follicle detection method was presented. The highest follicle detection accuracy of 97.63% was achieved with adaptive thresholding using a Wiener filter. Besides, the ultrasound images of the ovaries were classified as "normal" or "polycystic ovary syndrome" using CNN architecture with classification accuracy of 65.81% for unsegmented ovarian images and 77.81% for segmented images. In addition to the proposed method, classification was performed using SqueezeNet-based transfer learning, which was successful in limited datasets, and 74.18% classification accuracy was achieved for the unsegmented images and 75.54 % for segmented images . The results show that the combination of the Wiener filter with adaptive thresholding was quite successful in follicle detection and that the CNN can better classify ovaries using preprocessed ultrasound images.

Simultaneous Electroencephalography (EEG) and functional Magnetic Resonance Imaging (fMRI) provide a powerful framework for investigating brain dynamics; however, ballistocardiogram (BCG) artifacts in EEG compromise signal quality and limit the assessment of brain connectivity. This study evaluated three widely used artifact removal methods—Average Artifact Subtraction (AAS), Optimal Basis Set (OBS), and Independent Component Analysis (ICA)—together with two hybrid approaches (AAS + ICA and OBS + ICA). Unlike previous studies that focused solely on signal-level metrics, we adopted a holistic framework that combined signal quality indicators with graph-theoretical analysis of EEG-fMRI connectivity in static and dynamic contexts. The results show that AAS provides the best signal quality, whereas OBS better preserves structural similarity. ICA, although weaker in terms of signal metrics, demonstrates sensitivity to frequency-specific patterns in dynamic graphs. Hybrid methods yield benefits, with OBS + ICA producing the lowest p-values across frequency band pairs (e.g., theta–beta and delta–gamma), particularly in dynamic graphs. Topological analyses revealed that artifact removal significantly affected network structure, with dynamic analyses showing more pronounced frequency-specific effects than static analyses. High-frequency bands, such as beta and gamma, exhibit stronger differentiation under dynamic conditions. Overall, this study offers new insights into the relationship between artifact removal and brain network integrity, emphasizing the need for multimodal and frequency-sensitive evaluation strategies. The findings guide preprocessing decisions in EEG-fMRI studies and clarify how methodological choices shape the interpretation of brain connectivity.

Background: Simultaneous electroencephalography(EEG) and functional magnetic resonance imaging (fMRI) enable comprehensive investigationsof brain dynamics by combining high temporal and spatial resolutions. However, ballistocardiogram(BCG) artifacts in EEGsseverely affect signal quality and interpretation. This study aims to comparatively evaluate three widely used artifact removal techniques—average artifact subtraction (AAS), optimal basis set (OBS), and independent component analysis(ICA)—together with two hybrid approaches (AAS+ICA and OBS+ICA) within a multimodal and frequency-sensitive evaluation framework. Methods: Simultaneous EEG-fMRI data were preprocessed viafive artifact removal pipelines. Signal quality was assessed via time- and frequency-domain metrics, including power spectral density and correlation measures. Static and dynamic EEG-fMRI connectivity graphs were constructed on the basis of independent component networks (ICNs) and frequency-specific EEG features. Graph-theoretical measures (e.g., the clustering coefficient and global efficiency) were computed, and differences between methods were statistically evaluated via paired t tests with false discovery rate (FDR) correction. Results: AAS achieved the highest overall signal quality, whereasOBS preserved structural similarity more effectively. Although ICA yielded lower performance on traditional signal metrics, it demonstrated higher sensitivity to frequency-specific patterns, particularly in dynamic connectivity graphs. Among hybrid approaches, OBS+ICA provided the lowest p values across several frequency band pairs (e.g., theta–beta and delta–gamma), indicating improved detection of frequency-dependent interactions. Topological analyses revealed that artifact removal techniques substantially influence brain network organization. Dynamic connectivity analyses revealedstronger frequency-specific effects than static analyses did, with the beta and gamma bands showing the most pronounced differences. Conclusions: This study highlights the critical role of artifact removal strategies in shaping both EEG signal quality and EEG-fMRI connectivity outcomes. High-frequency bands, especially beta and gamma bands, exhibit distinctive network reconfigurations under dynamic conditions, underscoring their importance in cognitive and perceptual processes. By integrating signal-level, graph-theoretical, and multimodal evaluations, our findings provide practical guidelines for selecting preprocessing pipelines in simultaneous EEG-fMRI research and deepen our understanding of how methodological choices affect interpretations of brain connectivity.

The multimodal EEG-fMRI structure combines the advantages of EEG's high temporal resolution with the high spatial resolution of fMRI. However, BCG artifacts that arise due to the effect of the magnetic field during simultaneous recording distort EEG data, adversely affecting data analysis. In order to accurately interpret the data, these artifacts must be removed from the EEG recordings. Although classical artifact removal methods are widely used, they have many limitations. In recent years, deep learning-based methods have proven to be highly successful in eliminating BCG artifacts and offer an alternative to traditional approaches. The foundation of this success lies not only in the amount of data and computational power but also in the proper design of model architecture. In this study, an autoencoder-based model is proposed for the automatic removal of BCG artifacts. The effects of essential architectural components and attention mechanisms used in autoencoder-based models on model performance are examined. The advantages, disadvantages, and use-case scenarios of different components are discussed comparatively. This study highlights the critical role of architectural design in the success of the model.