arXiv subjects

Mujdat Cetin

Publications and source records attributed to Mujdat Cetin.

At least 19 recordsLinked to original sources

arXiv2025

PG-ControlNet: A Physics-Guided ControlNet for Generative Spatially Varying Image Deblurring

Spatially varying image deblurring remains a fundamentally ill-posed problem, especially when degradations arise from complex mixtures of motion and other forms of blur under significant noise. State-of-the-art learning-based approaches generally fall into two paradigms: model-based deep unrolling methods that enforce physical constraints by modeling the degradations, but often produce over-smoothed, artifact-laden textures, and generative models that achieve superior perceptual quality yet hallucinate details due to weak physical constraints. In this paper, we propose a novel framework that uniquely reconciles these paradigms by taming a powerful generative prior with explicit, dense physical constraints. Rather than oversimplifying the degradation field, we model it as a dense continuum of high-dimensional compressed kernels, ensuring that minute variations in motion and other degradation patterns are captured. We leverage this rich descriptor field to condition a ControlNet architecture, strongly guiding the diffusion sampling process. Extensive experiments demonstrate that our method effectively bridges the gap between physical accuracy and perceptual realism, outperforming state-of-the-art model-based methods as well as generative baselines in challenging, severely blurred scenarios.

cs.CV↗

arXiv2025

Conformalized Generative Bayesian Imaging: An Uncertainty Quantification Framework for Computational Imaging

Uncertainty quantification plays an important role in achieving trustworthy and reliable learning-based computational imaging. Recent advances in generative modeling and Bayesian neural networks have enabled the development of uncertainty-aware image reconstruction methods. Current generative model-based methods seek to quantify the inherent (aleatoric) uncertainty on the underlying image for given measurements by learning to sample from the posterior distribution of the underlying image. On the other hand, Bayesian neural network-based approaches aim to quantify the model (epistemic) uncertainty on the parameters of a deep neural network-based reconstruction method by approximating the posterior distribution of those parameters. Unfortunately, an ongoing need for an inversion method that can jointly quantify complex aleatoric uncertainty and epistemic uncertainty patterns still persists. In this paper, we present a scalable framework that can quantify both aleatoric and epistemic uncertainties. The proposed framework accepts an existing generative model-based posterior sampling method as an input and introduces an epistemic uncertainty quantification capability through Bayesian neural networks with latent variables and deep ensembling. Furthermore, by leveraging the conformal prediction methodology, the proposed framework can be easily calibrated to ensure rigorous uncertainty quantification. We evaluated the proposed framework on magnetic resonance imaging, computed tomography, and image inpainting problems and showed that the epistemic and aleatoric uncertainty estimates produced by the proposed framework display the characteristic features of true epistemic and aleatoric uncertainties. Furthermore, our results demonstrated that the use of conformal prediction on top of the proposed framework enables marginal coverage guarantees consistent with frequentist principles.

eess.IV↗

arXiv2024

LG-Sleep: Local and Global Temporal Dependencies for Mice Sleep Scoring

Efficiently identifying sleep stages is crucial for unraveling the intricacies of sleep in both preclinical and clinical research. The labor-intensive nature of manual sleep scoring, demanding substantial expertise, has prompted a surge of interest in automated alternatives. Sleep studies in mice play a significant role in understanding sleep patterns and disorders and underscore the need for robust scoring methodologies. In response, this study introduces LG-Sleep, a novel subject-independent deep neural network architecture designed for mice sleep scoring through electroencephalogram (EEG) signals. LG-Sleep extracts local and global temporal transitions within EEG signals to categorize sleep data into three stages: wake, rapid eye movement (REM) sleep, and non-rapid eye movement (NREM) sleep. The model leverages local and global temporal information by employing time-distributed convolutional neural networks to discern local temporal transitions in EEG data. Subsequently, features derived from the convolutional filters traverse long short-term memory blocks, capturing global transitions over extended periods. Crucially, the model is optimized in an autoencoder-decoder fashion, facilitating generalization across distinct subjects and adapting to limited training samples. Experimental findings demonstrate superior performance of LG-Sleep compared to conventional deep neural networks. Moreover, the model exhibits good performance across different sleep stages even when tasked with scoring based on limited training samples.

cs.LG↗

arXiv2024

Integrating Generative and Physics-Based Models for Ptychographic Imaging with Uncertainty Quantification

Ptychography is a scanning coherent diffractive imaging technique that enables imaging nanometer-scale features in extended samples. One main challenge is that widely used iterative image reconstruction methods often require significant amount of overlap between adjacent scan locations, leading to large data volumes and prolonged acquisition times. To address this key limitation, this paper proposes a Bayesian inversion method for ptychography that performs effectively even with less overlap between neighboring scan locations. Furthermore, the proposed method can quantify the inherent uncertainty on the ptychographic object, which is created by the ill-posed nature of the ptychographic inverse problem. At a high level, the proposed method first utilizes a deep generative model to learn the prior distribution of the object and then generates samples from the posterior distribution of the object by using a Markov Chain Monte Carlo algorithm. Our results from simulated ptychography experiments show that the proposed framework can consistently outperform a widely used iterative reconstruction algorithm in cases of reduced overlap. Moreover, the proposed framework can provide uncertainty estimates that closely correlate with the true error, which is not available in practice. The project website is available here.

eess.IV↗

arXiv2024

Robust EEG-based Emotion Recognition Using an Inception and Two-sided Perturbation Model

Automated emotion recognition using electroencephalogram (EEG) signals has gained substantial attention. Although deep learning approaches exhibit strong performance, they often suffer from vulnerabilities to various perturbations, like environmental noise and adversarial attacks. In this paper, we propose an Inception feature generator and two-sided perturbation (INC-TSP) approach to enhance emotion recognition in brain-computer interfaces. INC-TSP integrates the Inception module for EEG data analysis and employs two-sided perturbation (TSP) as a defensive mechanism against input perturbations. TSP introduces worst-case perturbations to the model's weights and inputs, reinforcing the model's elasticity against adversarial attacks. The proposed approach addresses the challenge of maintaining accurate emotion recognition in the presence of input uncertainties. We validate INC-TSP in a subject-independent three-class emotion recognition scenario, demonstrating robust performance.

eess.SP↗

arXiv2024

Subject-Independent Deep Architecture for EEG-based Motor Imagery Classification

Motor imagery (MI) classification based on electroencephalogram (EEG) is a widely-used technique in non-invasive brain-computer interface (BCI) systems. Since EEG recordings suffer from heterogeneity across subjects and labeled data insufficiency, designing a classifier that performs the MI independently from the subject with limited labeled samples would be desirable. To overcome these limitations, we propose a novel subject-independent semi-supervised deep architecture (SSDA). The proposed SSDA consists of two parts: an unsupervised and a supervised element. The training set contains both labeled and unlabeled data samples from multiple subjects. First, the unsupervised part, known as the columnar spatiotemporal auto-encoder (CST-AE), extracts latent features from all the training samples by maximizing the similarity between the original and reconstructed data. A dimensional scaling approach is employed to reduce the dimensionality of the representations while preserving their discriminability. Second, a supervised part learns a classifier based on the labeled training samples using the latent features acquired in the unsupervised part. Moreover, we employ center loss in the supervised part to minimize the embedding space distance of each point in a class to its center. The model optimizes both parts of the network in an end-to-end fashion. The performance of the proposed SSDA is evaluated on test subjects who were not seen by the model during the training phase. To assess the performance, we use two benchmark EEG-based MI task datasets. The results demonstrate that SSDA outperforms state-of-the-art methods and that a small number of labeled training samples can be sufficient for strong classification performance.

eess.SP↗

arXiv2024

Multi-Source Domain Adaptation with Transformer-based Feature Generation for Subject-Independent EEG-based Emotion Recognition

Although deep learning-based algorithms have demonstrated excellent performance in automated emotion recognition via electroencephalogram (EEG) signals, variations across brain signal patterns of individuals can diminish the model's effectiveness when applied across different subjects. While transfer learning techniques have exhibited promising outcomes, they still encounter challenges related to inadequate feature representations and may overlook the fact that source subjects themselves can possess distinct characteristics. In this work, we propose a multi-source domain adaptation approach with a transformer-based feature generator (MSDA-TF) designed to leverage information from multiple sources. The proposed feature generator retains convolutional layers to capture shallow spatial, temporal, and spectral EEG data representations, while self-attention mechanisms extract global dependencies within these features. During the adaptation process, we group the source subjects based on correlation values and aim to align the moments of the target subject with each source as well as within the sources. MSDA-TF is validated on the SEED dataset and is shown to yield promising results.

cs.LG↗

arXiv2023

A Hybrid End-to-End Spatio-Temporal Attention Neural Network with Graph-Smooth Signals for EEG Emotion Recognition

Recently, physiological data such as electroencephalography (EEG) signals have attracted significant attention in affective computing. In this context, the main goal is to design an automated model that can assess emotional states. Lately, deep neural networks have shown promising performance in emotion recognition tasks. However, designing a deep architecture that can extract practical information from raw data is still a challenge. Here, we introduce a deep neural network that acquires interpretable physiological representations by a hybrid structure of spatio-temporal encoding and recurrent attention network blocks. Furthermore, a preprocessing step is applied to the raw data using graph signal processing tools to perform graph smoothing in the spatial domain. We demonstrate that our proposed architecture exceeds state-of-the-art results for emotion classification on the publicly available DEAP dataset. To explore the generality of the learned model, we also evaluate the performance of our architecture towards transfer learning (TL) by transferring the model parameters from a specific source to other target domains. Using DEAP as the source dataset, we demonstrate the effectiveness of our model in performing cross-modality TL and improving emotion classification accuracy on DREAMER and the Emotional English Word (EEWD) datasets, which involve EEG-based emotion classification tasks with different stimuli.

cs.LG↗

arXiv2022

Uncertainty Quantification for Deep Unrolling-Based Computational Imaging

Deep unrolling is an emerging deep learning-based image reconstruction methodology that bridges the gap between model-based and purely deep learning-based image reconstruction methods. Although deep unrolling methods achieve state-of-the-art performance for imaging problems and allow the incorporation of the observation model into the reconstruction process, they do not provide any uncertainty information about the reconstructed image, which severely limits their use in practice, especially for safety-critical imaging applications. In this paper, we propose a learning-based image reconstruction framework that incorporates the observation model into the reconstruction task and that is capable of quantifying epistemic and aleatoric uncertainties, based on deep unrolling and Bayesian neural networks. We demonstrate the uncertainty characterization capability of the proposed framework on magnetic resonance imaging and computed tomography reconstruction problems. We investigate the characteristics of the epistemic and aleatoric uncertainty information provided by the proposed framework to motivate future research on utilizing uncertainty information to develop more accurate, robust, trustworthy, uncertainty-aware, learning-based image reconstruction and analysis methods for imaging problems. We show that the proposed framework can provide uncertainty information while achieving comparable reconstruction performance to state-of-the-art deep unrolling methods.

eess.IV↗

arXiv2021

Combining physics-based modeling and deep learning for ultrasound elastography

Ultrasound elasticity images which enable the visualization of quantitative maps of tissue stiffness can be reconstructed by solving an inverse problem. Classical model-based approaches for ultrasound elastography use deterministic finite element methods (FEMs) to incorporate the governing physical laws resulting in poor performance in noisy conditions. Moreover, these approaches utilize fixed regularizers for various tissue patterns while appropriate data-adaptive priors might be required for capturing the complex spatial elasticity distribution. In this regard, we propose a joint model-based and learning-based framework for estimating the elasticity distribution by solving a regularized optimization problem. We present an integrated objective function composed of a statistical physics-based forward model and a data-driven regularizer to leverage deep neural networks for learning the underlying elasticity prior. This constrained optimization problem is solved using the gradient descent (GD) method and the gradient of regularizer is simply replaced by the residual of the trained denoiser network for having an explicit objective function with reduced computation time.

eess.IV↗

arXiv2021

Regularization by Adversarial Learning for Ultrasound Elasticity Imaging

Classical model-based imaging methods for ultrasound elasticity inverse problem require prior constraints about the underlying elasticity patterns, while finding the appropriate hand-crafted prior for each tissue type is a challenge. In contrast, standard data-driven methods count solely on supervised learning on the training data pairs leading to massive network parameters for unnecessary physical model relearning which might not be consistent with the governing physical models of the imaging system. Fusing the physical forward model and noise statistics with data-adaptive priors leads to a united reconstruction framework that guarantees the learned reconstruction agrees with the physical models while coping with the limited training data. In this paper, we propose a new methodology for estimating the elasticity image by solving a regularized optimization problem which benefits from the physics-based modeling via a data-fidelity term and adversarially learned priors via a regularization term. In this method, the regularizer is trained based on the Wasserstein Generative Adversarial Network (WGAN) objective function which tries to distinguish the distribution of clean and noisy images. Leveraging such an adversarial regularizer for parameterizing the distribution of latent images and using gradient descent (GD) for solving the corresponding regularized optimization task leads to stability and convergence of the reconstruction compared to pixel-wise supervised learning schemes. Our simulation results verify the effectiveness and robustness of the proposed methodology with limited training datasets.

eess.IV↗

arXiv2021

MR elasticity reconstruction using statistical physical modeling and explicit data-driven denoising regularizer

Elasticity image, visualizing the quantitative map of tissue stiffness, can be reconstructed by solving an inverse problem. Classical methods for magnetic resonance elastography (MRE) try to solve a regularized optimization problem comprising a deterministic physical model and a prior constraint as data-fidelity term and regularization term, respectively. For improving the elasticity reconstructions, appropriate prior about the underlying elasticity distribution is required which is not unique. This article proposes an infused approach for MRE reconstruction by integrating the statistical representation of the physical laws of harmonic motions and learning-based prior. For data-fidelity term, we use a statistical linear-algebraic model of equilibrium equations and for the regularizer, data-driven regularization by denoising (RED) is utilized. In the proposed optimization paradigm, the regularizer gradient is simply replaced by the residual of learned denoiser leading to time-efficient computation and convex explicit objective function. Simulation results of elasticity reconstruction verify the effectiveness of the proposed approach.

eess.IV↗

arXiv2021

Finite Element Reconstruction Of Stiffness Images In MR Elastography Using Statistical Physical Forward Modeling And Proximal Optimization Methods

Quantitative characterization of tissue properties, known as elasticity imaging, can be cast as solving an ill-posed inverse problem. The finite element methods (FEMs) in magnetic resonance elastography (MRE) imaging are based on solving a constrained optimization problem consisting of a physical forward model and a regularizer as the data-fidelity term and the prior term, respectively. In existing formulation for the elasticity forward model, physical laws that arise from equilibrium equation of harmonic motion, indicate a deterministic relationship between MRE-measured data and unknown elasticity distribution which leads to the poor and unstable elasticity distribution estimation in the presence of noise. Toward this end, we propose an efficient statistical methodology for physical forward model refinement by formulating it as linear algebraic representation with respect to the unknown elasticity distribution and incorporating an analytical noise model. To solve the subsequent total variation regularized optimization task, we benefit from a fixed-point scheme involving proximal gradient methods. Simulation results of elasticity reconstruction in various SNR conditions verify the effectiveness of the proposed approach.

eess.IV↗

arXiv2021

Ultrasound Elasticity Imaging Using Physics-based Models And Learning-based Plug-And-Play Priors

Existing physical model-based imaging methods for ultrasound elasticity reconstruction utilize fixed variational regularizers that may not be appropriate for the application of interest or may not capture complex spatial prior information about the underlying tissues. On the other hand, end-to-end learning-based methods count solely on the training data, not taking advantage of the governing physical laws of the imaging system. Integrating learning-based priors with physical forward models for ultrasound elasticity imaging, we present a joint reconstruction framework which guarantees that learning driven reconstructions are consistent with the underlying physics. For solving the elasticity inverse problem as a regularized optimization problem, we propose a plug-and-play (PnP) reconstruction approach in which each iteration of the elasticity image estimation process involves separate updates incorporating data fidelity and learning-based regularization. In this methodology, the data fidelity term is developed using a statistical linear algebraic model of quasi-static equilibrium equation revealing the relationship of the observed displacement fields \cmmnt{measured deformation data} to the unobserved elastic modulus. The regularizer comprises a convolutional neural network (CNN) based denoiser that captures the learned prior structure of the underlying tissues. Preliminary simulation results demonstrate the robustness and effectiveness of the proposed approach with limited training datasets and noisy displacement measurements.

eess.IV↗

arXiv2021

Online Graph Learning under Smoothness Priors

The growing success of graph signal processing (GSP) approaches relies heavily on prior identification of a graph over which network data admit certain regularity. However, adaptation to increasingly dynamic environments as well as demands for real-time processing of streaming data pose major challenges to this end. In this context, we develop novel algorithms for online network topology inference given streaming observations assumed to be smooth on the sought graph. Unlike existing batch algorithms, our goal is to track the (possibly) time-varying network topology while maintaining the memory and computational costs in check by processing graph signals sequentially-in-time. To recover the graph in an online fashion, we leverage proximal gradient (PG) methods to solve a judicious smoothness-regularized, time-varying optimization problem. Under mild technical conditions, we establish that the online graph learning algorithm converges to within a neighborhood of (i.e., it tracks) the optimal time-varying batch solution. Computer simulations using both synthetic and real financial market data illustrate the effectiveness of the proposed algorithm in adapting to streaming signals to track slowly-varying network connectivity.

cs.LG↗

arXiv2021

Depth extraction from a single compressive hologram

We propose a novel method that records a single compressive hologram in a short time and extracts the depth of a scene from that hologram using a stereo disparity technique. The method is verified with numerical simulations, but there is no restriction on adapting this into an optical experiment. In the simulations, a computer-generated hologram is first sampled with random binary patterns, and measurements are utilized in a recovery algorithm to form a compressive hologram. The compressive hologram is then divided into two parts (two apertures), and these parts are separately reconstructed to form a stereo image pair. The pair is eventually utilized in stereo disparity method for depth map extraction. The depth maps of the compressive holograms with the sampling rates of 2, 25, and 50 percent are compared with the depth map extracted from the original hologram, on which compressed sensing is not applied. It is demonstrated that the depth profiles obtained from the compressive holograms are in very good agreement with the depth profile obtained from the original hologram despite the data reduction.

eess.IV↗

arXiv2021

Online Discriminative Graph Learning from Multi-Class Smooth Signals

Graph signal processing (GSP) is a key tool for satisfying the growing demand for information processing over networks. However, the success of GSP in downstream learning and inference tasks is heavily dependent on the prior identification of the relational structures. Graphs are natural descriptors of the relationships between entities of complex environments. The underlying graph is not readily detectable in many cases and one has to infer the topology from the observed signals. Firstly, we address the problem of graph signal classification by proposing a novel framework for discriminative graph learning. To learn discriminative graphs, we invoke the assumption that signals belonging to each class are smooth with respect to the corresponding graph while maintaining non-smoothness with respect to the graphs corresponding to other classes. Secondly, we extend our work to tackle increasingly dynamic environments and real-time topology inference. We develop a proximal gradient (PG) method which can be adapted to situations where the data are acquired on-the-fly. Beyond discrimination, this is the first work that addresses the problem of dynamic graph learning from smooth signals where the sought network alters slowly. The validation of the proposed frameworks is comprehensively investigated using both synthetic and real data.

eess.SP↗

arXiv2020

A statistical framework for model-based inverse problems in ultrasound elastography

Model-based computational elasticity imaging of tissues can be posed as solving an inverse problem over finite elements spanning the displacement image. As most existing quasi-static elastography methods count on deterministic formulations of the forward model resulting in a constrained optimization problem, the impact of displacement observation errors has not been well addressed. To this end, we propose a new statistical technique that leads to a unified optimization problem for elasticity imaging. Our statistical model takes the imperfect nature of the displacement measurements into account, and leads to an observation model for the Young's modulus that involves signal dependent colored noise. To solve the resulting regularized optimization problem, we propose a fixed-point algorithm that leverages proximal splitting methods. Preliminary qualitative and quantitative results demonstrate the effectiveness and robustness of the proposed methodology.

eess.IV↗