PhD

FTSM

M.Sc. Computer Science

India

B.Sc. Computer Science

I got my B.Sc. from the departement of software Engineering, Faculty of Commuter Science and Mathematics from the University of Mosul, Iraq

Lecturer [Cihan University]

Dept. of Computer Science and IT.

Generative Federated Learning with Small and Large Models In Consumer Electronics for Privacy preserving Data Fusion in Healthcare Internet of Things

Journal IEEE Transactions on Consumer Electronics

publisher Scopus

DOI 10.1109/TCE.2025.3572629

Healthcare Internet of Things (HIoT) requires large-scale privacy features to ensure maximum security in sharing sensitive physiological data in consumer electronics. Recent approaches utilize the fusion concept to provide maximum privacy in health data sharing. Embedded signing data fusion with the health observed data ensures privacy preserved sharing across heterogeneous medical consumer devices for diagnosis. This article proposes a Dependency-correlated Data Fusion Scheme (DcDFS) to maximize the privacy of the health data-sharing process. The proposed scheme prepares separate key signing procedures using triple-DES (data encryption standard) to embed with the accumulated health data. The fusion process is carried out by defining key headers and integrity footers for authentication and verification. Therefore, the fusion generates a combined sequence of linear authentication and validation procedures enclosing the health data. In this scheme, the role of federated learning is to prevent permuted sequences for the same health data. This research integrates Small Language Model (SLM) and Large Language Model (LLM) into the federated learning module to support secure, scalable, and intelligent healthcare data sharing. Their collaboration enhances context-aware training while preserving privacy across decentralized, encrypted environments. A similar sequence mapped between the header and footer is responsible for discarding unauthorized data handling. The learning process verifies the permutation for many-to-one header to footer and vice versa. Therefore, the proposed fusion scheme generates a linear dependency between the actual and security-related data for maximum privacy. The proposed scheme achieves the following: the computation time is confined by 12.424%, the privacy leakage by 12.923%, and the computation efficiency is improved by 11.46%, as observed under the maximum sequences.

Read Publication

A Transfer Learning Based Framework for Diabetic Retinopathy Detection Using Data Fusion

Journal 2nd International Conference on Cyber Resilience, ICCR 2024

publisher IEEE

DOI 10.1109/ICCR61006.2024.10533112

Diabetic retinopathy is a condition associated with diabetes that damages the blood vessels within the retina resulting in vision impairment or even blindness. Early detection and classification of DR enables timely intervention, which is crucial for preventing vision loss and blindness in diabetic patients. This research presents a framework for binary classification, employing transfer learning to identify diabetic retinopathy in individuals with diabetes. APTOS19 and IDRiD, two datasets containing fundus images, are merged together for training the transfer learning models to predict the presence or absence of the disease. Many preprocessing techniques have been applied to these images like resizing, Gaussian filtering, and dataset splitting. After the split, training set is augmented using zooming, rotation, flipping etc. to increase diversity. The transfer learning models used are: ResNet50 and DenseNet121. These models are fine tuned for classification. The results highlight that the DenseNet121 model achieved a superior test accuracy of 97.22% as compared to ResNet50.

Read Publication

Synthesizing Retinal Images using End-To-End VAEs-GAN Pipeline-Based Sharpening and Varying Layer

Journal Multimedia Tools and Applications

publisher Springer

DOI 10.1007/s11042-023-17058-2

Issue 1

Volume 83

This study attempts to synthesize a realistic-looking fundus image from a morphologically changed vessel structure using the newly proposed sharpening and varying vessels technique (SVV). This technique sharpens the reconstructed vessels and introduces variation to their structure to generate multiple images from a single input mask. This helps to reduce the reliance on expensive and scarce annotated medical data. The study also aims to overcome the limitations of current methods, such as unrealistic optic disc boundaries, extreme vessel tortuosity, and missed optic discs. This is mainly due to the fact that existing models penalize their weights based on the difference between real and synthetic images using only a single mask. Therefore, their emphasis is on generating the input mask while disregarding other important fundoscopic features. Inspired by the recent progress in Generative Adversarial Nets (GANs) and Variational Autoencoder (VAE), the proposed approach was able to preserve the geometrical shape of critical fundus characteristics. Visual and quantitative results indicate that the produced images are considerably distinct from the ones used for training. However, they also exhibit anatomical coherence and a reasonable level of visual. The data utilized in this study and the programming code necessary to recreate the experiment can be accessed at https://github.com/AliSaeed86/SVV_GAN .

Read Publication

Enhancing Email Spam Detection Using Advanced AI Techniques

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836525

This review paper investigates the advances of artificial intelligence (AI) in the field of email spam detection. The study addresses AI-based techniques used for email spam filtering by classifying threats such as keywords, content-based, sender IDs, and phrases. Email spam remains one of the most pressing issues today, with spammers sending malicious links in junk folders to compromise confidential information. As spam filtering becomes increasingly complex, AI and machine learning techniques like Naïve Bayes, Support Vector Machine (SVM), and Random Forest are commonly used to mitigate these threats. This study proposes a novel framework that combines traditional machine learning models with deep neural networks to develop a robust and flexible spam detection system. The framework is designed to enhance the detection capabilities of the classifier by applying a series of key sequential steps, each intended to improve the system's ability to distinguish between legitimate emails and spam. The paper details the methodology, providing a step-by-step explanation of each phase to demonstrate how the framework advances email spam detection.

Read Publication

Integrating Three Machine Learning Algorithms in Ensemble Learning Model for Improving Content-based Spam Email Recognition

Journal Journal of Soft Computing and Data Mining

publisher Scopus / WOS

DOI 10.30880/jscdm.2024.05.02.014

Issue 2

Volume 5

Email spam refers to junk files, images, or data sent through email that might contain links leading to phishing websites. This email is often sent repeatedly to random users, and sometimes it may be dangerous. The objective of this study is to predict and recognize whether the emails sent to users are spam or not by using machine learning classification algorithms. Email Spam Classification (ESC) datasets are used in this study for spam detection tests. The ESC datasets contain 5172 rows and 3002 columns of spam and non-spam features. The methodology used in this study is the CRISP-DM to guide the process of evaluating the performance of three machine learning algorithms: Naive Bayes (NB), Logistic Regression (LR), and Random Forest (RF). Subsequently, an ensemble model that integrates the three machine learning algorithms is proposed to improve the performance of spam email recognition. The selected evaluation metrics are F1-Score, accuracy, precision, and recall. Based on the results, the RF algorithm has the highest accuracy of 97.3% in classifying spam emails, with an F1 score of 96.8%, precision of 96.2%, and recall of 96.0%. The NB achieves the best second results, which are slightly different from the RF, and the LR achieves considerably lower results than the other two algorithms. The ensemble model that integrates the three algorithms performs best in classifying spam emails with 98.9% accuracy, 97.6% precision, 97.4% recall, and 96.7% F1-score.

Artificial Intelligence in Account Management: Innovation, Challenges, and Strategic Outlook

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836501

This article presents the growing role of artificial intelligence (AI) in account management, through highlighting its increasing significance in computing and financial administration. AI is rapidly becoming a transformative tool across various sectors, and its application in tax administration presents both opportunities and challenges. The article explores AI-driven business opportunities, growth trends, and strategic planning within the field. in addition to address the critical technical, ethical, and legal considerations, including intellectual property rights, that must be navigated to ensure responsible and effective implementation.

Read Publication

Performance Analysis of Training Optimizers on Diabetic Retinopathy Detection Using SP-Net CNN

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836543

Diabetic retinopathy is an ocular disease linked to long-term diabetes mellitus and it can lead to permanent loss of vision if not addressed promptly. Timely detection of this disease is crucial as it can significantly reduce the risk of severe complications. Our aim is to identify the most effective optimization algorithm for training our deep learning (CNN) model. Our goal is to achieve high accuracy and efficiency in grading diabetic retinopathy into 5 different classes. In this research, we conducted a comparative evaluation of three optimizers namely Adam, SGD and RMSProp. For classification, we developed a convolutional neural network called SP-Net. An image based dataset named APTOS 2019 is taken from Kaggle and various pre-processing techniques are applied on it such as cropping, denoising, histogram equalization, and resizing. We used data augmentation techniques to tackle the class imbalance issue. Then the dataset was divided into three subsets (training, validation, and testing) using a split of 70:10:20. Our model SP-Net incorporates several layers including convolutional, dropout, max pooling, and fully connected layers. The results demonstrate that SP-Net achieved a highest test accuracy of 97.76% when it was trained using the SGD optimizer. Moreover, this model trained with the best optimizer outperformed the existing state-of-the-art techniques on comparison.

Read Publication

Classification of Lung Diseases Using Machine Learning Technique

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836302

The advancement of medicine depends on the use of modern technologies. To provide precise and tailored treatment options for a variety of illnesses, comprehensive research conducted in collaboration with medical experts, patients, and researchers is crucial. This study uses deep learning to determine the acceptable level of accuracy in the medical field based on data that is accessible to the public. To begin, we took the annotated lung sound recordings and extracted the spectrogram features and labels to feed into our 2D Convolutional Neural Network (CNN) model. In this work, we address the issue of scarce medical data by employing small-volume datasets with less than a thousand samples to identify pulmonary disorders from chest X-ray images. Deep learning algorithms are utilized to treat pneumonia, tuberculosis, lung cancer, and other lung diseases. A review of various typical deep-learning network topologies used in medical image processing is also provided. Our ensemble model achieved a classification accuracy of 96.2% and an F1-score of 96.1 %, outperforming individual models such as VDSNet, ResNet18, DenseNet201, and SqueezeN et. This demonstrates the effectiveness of our approach in enhancing diagnostic accuracy for lung disease classification

Read Publication

Leveraging AI for Enhancing Robustness and Optimization in Constraint-Based Job-Shop Scheduling with Operators

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836349

in complex scheduling scenarios, such as the Job Scheduling Problem (JSP) faced by human operators, the challenge is to optimize the total time mandatory to complete all jobs and the reliability of the schedule. This paper presents a unified approach that combines constraint satisfaction (CSP) techniques with heuristic algorithms to resolve JSP using operators called JSO(n, p), where n is the number of tasks and p is the number of data. The main aims are to minimize the total time required to complete all tasks and to maximize system consistency under uncertainties such as equipment failure or operator unavailability. We analysis branch-and-bound (BB) methods, neural networks, and hybrid algorithms to achieve optimal software solutions. This paper also assesses the impact of bus constraints on traditional scheduling models and suggests solutions for reliability. The evaluation of various algorithmic strategies and extensive experiments on many sampling problems demonstrate the balance between robustness and optimality.

AI-Driven Strategic Foresight: Anticipating Future Trends and Modelling Business Strategies

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836619

In a generation characterized with the aid of speedy technological advancements, socio-monetary shifts, and environmental challenges, groups need to undertake proactive strategies to stay aggressive and resilient. Strategic foresight, which entails expecting destiny tendencies and modeling enterprise techniques, accordingly, emerges as a vital device in navigating this uncertainty. The trouble this have a look at addresses is the growing trouble groups face in predicting and making ready for destiny disruptions the use of conventional forecasting techniques. The number one goals of this have a look at are to discover the important thing methodologies and gear of strategic foresight, display how those techniques may be implemented to assume destiny tendencies, and offer insights on integrating foresight into enterprise approach formulation. This has a look at contributes to the present frame of information with the aid of using providing a complete evaluation of strategic foresight techniques, which includes environmental scanning, situation planning, the Delphi method, fashion evaluation, SWOT and PESTLE evaluation, and era street mapping. Research findings show that forward-looking strategies provide an organization with the ability to identify potential risks and opportunities, develop adaptive strategies, and achieve long-term growth Specifically, strategic organizations using anticipation reports a 30% increase in the ability to anticipate market fluctuations, a 25% improvement in strategic haste and a 20% reduction in unexpected disruptions, of consumer goods etc. Case studies of various industries demonstrate the practical application and usefulness of forecasting in real-world situations The study concludes that incorporating forward-looking strategies into business planning does not that it not only prepares organizations for future challenges but empowers them to shape their future and ensure sustainable competitive advantage.

Analytical Study of the Features and Parameters of Off-Angle Iris Image vs. Deep Learning Model

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836647

This paper presents an analytical study and a technique for extracting the features of a common case of images of the iris called off-angle iris which was taken for persons identification systems. The main problem when using biological iris measurements to identify the persons is the difficulty of identifying and extracting features of the iris. This problem is increasing when dealing with off-angle iris and it leads to decreased system accuracy and increased system rate error. In return, all the transfer learning techniques face difficulties in the case of heavily degraded data and partially occluded, and off-angle. A new method has a whole new methodology to deal with the image as it is without transformation processes. A new algorithm has been included for extracting features of the iris through the pupil switching points. The most discriminating points of the iris depend on the biological human eye statistics and analytical study. It has been trained and tested on the common images of the off-angle iris database so-called: 'CASIA Iris 1.0'. It has been implemented in the MATLAB environment. The results showed the efficiency of this technique, high precision and more importantly low failed acceptance rate. It emphasized that it is adaptive as well as efficiency improvement of the system.

Enhancing Scalability in Reinforcement Learning for Open Spaces

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836237

Reinforcement Learning (RL) has been successful when the environment has specific objectives and boundaries. But with the emerging focus on open-world application which makes all or some of the rules or purpose go to naught, it makes traditional methods of RL a bit more difficult. This paper goes over various advancements and changes in Reinforcement Learning which can be employed for open-ended environments. Among the other strategies, hierarchical reinforcement learning, intrinsic motivation-based exploration, meta-learning and unsupervised skill acquisition are also among the ones that are examined. As a result, such a position based on the technology argues the promising future of open-ended methods for the management of complex problems and high level of uncertainty associated with the preset target or purpose. Also, we study cases in video games, robotics and autonomous systems, where RL is implemented in an open-ended and dynamic environment. We also outline existing limitations and perspectives, highlighting the need for more flexible methods and inter-scientific collaboration to fully realize RL's potential in open-ended contexts.

Deep Learning-Based Approaches for Accurate Brain Tumor Detection in MRI Images

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836556

Brain tumor classification is considered one of the major tasks in medical image analysis, where correct and timely diagnosis could be achieved to serve as the key to effective treatment. This research paper proposes a deep learning-based approach for automatic classification of brain tumors from MRI images by fine tuning the ResNet50V2 CNN model. The dataset is made up of four classes, namely: Glioma Tumor, Meningioma Tumor, No Tumor, and Pituitary Tumor, totaling 2,844 images. We have done some strategy data augmentation and class weighting in order to prevent class imbalance, which ensures good performance in all tumor classes. Pre-processing and augmentation of the input images are done via rotation, shifting, zooming, and flipping. It yielded 95.29% on the validation set, with the major highlights of performance metrics for the minority class being class precision equal to 0.97 and class recall equal to 0.92, hence proving the efficiency of performed class balancing. The range of F1-scores, from 0.93 to 0.97, means fantastic predictive capability across all tumor types. Other techniques were also used to make this model optimal, such as early stopping, learning rate reduction, and checkpointing. The high performance of this model shows great promise in helping clinicians toward the right diagnosis of brain tumors in an effective way. Future efforts will involve incorporating larger datasets, exploring advanced augmentation techniques to enhance further model generalizability. Model explainability tools such as Grad-CAM will be used to extract insight into the model's decision-making process. This enhances the clinical interpretability of the results.

Enhancing Food Engineering Through Machine Learning in Analyzing Biomolecular Structures

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE

DOI 10.1109/DASA63652.2024.10836446

In this paper, a detailed investigation has been outlined of protein sequence data with the help of a machine learning model for the identification of macromolecule types. The dataset was preprocessed from the Protein Data Bank and included missing values and duplicate structures, and the nominal variables have been encoded. The approach that was used in our study entailed the employment of Random Forest Classifier to forecast macromolecule types using the sequence data. Researchers divided the data into the training and testing data sets, thereafter, training the classifier, they tested the classifier's accuracy, precision, recall, and F1-score metrics. It can be stated that the relevance of this study is based on the necessity to discover the structural and functional properties of proteins in the field of bioinformatics. Hence, through an accurate classification of the macromolecule types, this approach assists in defining and describing the proteins, thus demystifying the roles that they play within biological cells. Our contribution also comprises the derivation of a highly reliable classification model that would be useful in identifying the type of macromolecules present from protein sequences. The model thus made high accurate determination of the different macromolecules and provided a clear framework which showed the model's efficacy in the differentiation of different macromolecules. Also, we describe the results of the analysis of macromolecule types of distribution and reveal the most often used categories within the set. This paper, therefore, demonstrates the potential of ML in bioinformatics and improves knowledge in predicting protein structure and function which may be of value in numerous biomedicine sectors.

Nanostructured Photonics Probes: A Transformative Approach in Neurotherapeutics and Brain Circuitry

Journal Neuroscience

publisher Elsevier Ltd

DOI 10.1016/j.neuroscience.2024.10.046

Volume 562

Neuroprobes that use nanostructured photonic interfaces are capable of multimodal sensing, stimulation, and imaging with unprecedented spatio-temporal resolution. In addition to electrical recording, optogenetic modulation, high-resolution optical imaging, and molecular sensing, these advanced probes combine nanophotonic waveguides, optical transducers, nanostructured electrodes, and biochemical sensors. The potential of this technology lies in unraveling the mysteries of neural coding principles, mapping functional connectivity in complex brain circuits, and developing new therapeutic interventions for neurological disorders. Nevertheless, achieving the full potential of nanostructured photonic neural probes requires overcoming challenges such as ensuring long-term biocompatibility, integrating nanoscale components at high density, and developing robust data-analysis pipelines. In this review, we summarize and discuss the role of photonics in neural probes, trends in electrode diameter for neural interface technologies, nanophotonic technologies using nanostructured materials, advances in nanofabrication photonics interface engineering, and challenges and opportunities. Finally, interdisciplinary efforts are required to unlock the transformative potential of next-generation neuroscience therapies.

Read Publication

Fine Vessel Segmentation With Refinement Gate in Deep Learning Architectures

Journal Malaysian Journal of Computer Science

publisher Faculty of Computer Science and Information Technology

DOI 10.22452/mjcs.vol37no3.2

Issue 3

Volume 37

Automated vessel segmentation is essential in diagnosing eye-related disorders and monitoring progressive retinal diseases. State-of-the-art methods have achieved excellent results in this field, but very few have considered the post-processing of feature maps. As a result, there is often a lack of small and fine vessels or discontinuities in segmented vessels. To address this issue, this study introduces a novel post-processing technique called the refinement gate, which works with a deep learning model during training. The refinement gate enhances contextual information to extract important features from feature maps better. The proposed technique is applied with U-net architecture and placed after every convolution block in the encoder path. Visual and statistical comparisons demonstrate the robustness of the proposed method using three publicly available datasets, namely: the DRIVE DB, the STARE DB, and CHASE_DB1 datasets, showing significant improvements to segment weak and tiny vessels. The reported results confirm the potential of the model to be used as a segmentation tool in the medical field. This study is the first to propose such a gating mechanism without additional trainable parameters or standalone networks as in other literature.

Read Publication

Classification of Lung Diseases Using Machine Learning Technique

Journal 2024 International Conference on Decision Aid Sciences and Applications, DASA 2024

publisher IEEE / Scopus

DOI 10.1109/DASA63652.2024.10836302

The advancement of medicine depends on the use of modern technologies. To provide precise and tailored treatment options for a variety of illnesses, comprehensive research conducted in collaboration with medical experts, patients, and researchers is crucial. This study uses deep learning to determine the acceptable level of accuracy in the medical field based on data that is accessible to the public. To begin, we took the annotated lung sound recordings and extracted the spectrogram features and labels to feed into our 2D Convolutional Neural Network (CNN) model. In this work, we address the issue of scarce medical data by employing small-volume datasets with less than a thousand samples to identify pulmonary disorders from chest X-ray images. Deep learning algorithms are utilized to treat pneumonia, tuberculosis, lung cancer, and other lung diseases. A review of various typical deep-learning network topologies used in medical image processing is also provided. Our ensemble model achieved a classification accuracy of 96.2% and an F1-score of 96.1 %, outperforming individual models such as VDSNet, ResNet18, DenseNet201, and SqueezeN et. This demonstrates the effectiveness of our approach in enhancing diagnostic accuracy for lung disease classification

Read Publication

Accuracy of using generative adversarial networks for glaucoma detection: Systematic review and bibliometric analysis

Journal Journal of Medical Internet Research

publisher Scopus / WOS

DOI 10.2196/27414

Issue 9

Volume 23

Background: Glaucoma leads to irreversible blindness. Globally, it is the second most common retinal disease that leads to blindness, slightly less common than cataracts. Therefore, there is a great need to avoid the silent growth of this disease using recently developed generative adversarial networks (GANs). Objective: This paper aims to introduce a GAN technology for the diagnosis of eye disorders, particularly glaucoma. This paper illustrates deep adversarial learning as a potential diagnostic tool and the challenges involved in its implementation. This study describes and analyzes many of the pitfalls and problems that researchers will need to overcome to implement this kind of technology. Methods: To organize this review comprehensively, articles and reviews were collected using the following keywords: ("Glaucoma,""optic disc,""blood vessels") and ("receptive field,""loss function,""GAN,""Generative Adversarial Network,""Deep learning,""CNN,""convolutional neural network"OR encoder). The records were identified from 5 highly reputed databases: IEEE Xplore, Web of Science, Scopus, ScienceDirect, and PubMed. These libraries broadly cover the technical and medical literature. Publications within the last 5 years, specifically 2015-2020, were included because the target GAN technique was invented only in 2014 and the publishing date of the collected papers was not earlier than 2016. Duplicate records were removed, and irrelevant titles and abstracts were excluded. In addition, we excluded papers that used optical coherence tomography and visual field images, except for those with 2D images. A large-scale systematic analysis was performed, and then a summarized taxonomy was generated. Furthermore, the results of the collected articles were summarized and a visual representation of the results was presented on a T-shaped matrix diagram. This study was conducted between March 2020 and November 2020. Results: We found 59 articles after conducting a comprehensive survey of the literature. Among the 59 articles, 30 present actual attempts to synthesize images and provide accurate segmentation/classification using single/multiple landmarks or share certain experiences. The other 29 articles discuss the recent advances in GANs, do practical experiments, and contain analytical studies of retinal disease. Conclusions: Recent deep learning techniques, namely GANs, have shown encouraging performance in retinal disease detection. Although this methodology involves an extensive computing budget and optimization process, it saturates the greedy nature of deep learning techniques by synthesizing images and solves major medical issues. This paper contributes to this research field by offering a thorough analysis of existing works, highlighting current limitations, and suggesting alternatives to support other researchers and participants in further improving and strengthening future work. Finally, new directions for this research have been identified.

Read Publication

The 3rd International Conference on Cyber Resilience (ICCR 2025)

Publisher IEEE / Scopus

Country UAE

Location British University in Dubai

Visit Conference

The 3rd International Conference on Business Analytics for Technology and Security (ICBATS-2025)

Publisher IEEE

Country UAE

Location The British University in Dubai, Dubai,

Visit Conference

2nd International Conference on Sustainable Development Techniques

Publisher Springer

Country Türkiye

Location Alanya Alaaddin Keykubat University, Alanya, Türkiye

Visit Conference