Mohamed I. Marie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mohamed I. Marie has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
Ahmad Essayed Yakoub
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt
1 shared publication- Amira M. Idrees
College of Business, King Khalid University, Abha 61421, Saudi Arabia
1 shared publication - Ayman E. Khedr
Faculty of Computers and Information Technology, Future University in Egypt, Cairo, Egypt
1 shared publication Hoda A. Elnaggar
Management Information System Department Higher Institute of Management Sciences and Foreign Trade Cairo 379 Egypt
1 shared publicationMarwa Elsherif
Economics Department Helwan University Cairo 11795 Egypt
1 shared publication- Mohamed S. Elredeny
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt
1 shared publication
Latest Publications
- Open Access
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 120327-120344
A Hybrid Model for Early Melanoma Detection: Integrating YOLOv9 and Faster R-CNN for Enhanced Diagnostic Accuracy
Melanoma accounts for only 1% of skin cancer diagnoses yet causes the majority of skin cancer-related deaths due to its rapid progression and high metastatic potential. Early and accurate detection is crucial for improving patient outcomes; however, existing deep learning models often struggle to balance diagnostic precision with real-time efficiency. This study presents the Melano Hybrid Model, a novel architecture that integrates the rapid detection capabilities of YOLOv9 with the boundary localization accuracy of Faster R-CNN through an adaptive feature fusion mechanism. The model was rigorously evaluated on three benchmark datasets—ISIC 2019, HAM10000, and ISIC 2020—using 5-fold cross-validation. On ISIC 2020, the hybrid model achieved a 96.2% classification accuracy (95% CI: 95.8–96.6%) and a 95.1% F1-score (95% CI: 94.7–95.5%), significantly outperforming standalone models (p < 0.001). The architecture delivers an average inference speed of 31.3 frames per second (FPS), surpassing clinical real-time thresholds. Additionally, computational profiling confirms its practical feasibility with 78.3 million parameters, 134.8 GFLOPs, and a 324 MB memory footprint. These results support the hybrid framework as a robust AI-assisted tool for real-world melanoma screening, offering an optimal trade-off between speed and diagnostic performance.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 120327-120344
Open AccessA Hybrid Model for Early Melanoma Detection: Integrating YOLOv9 and Faster R-CNN for Enhanced Diagnostic Accuracy
Melanoma accounts for only 1% of skin cancer diagnoses yet causes the majority of skin cancer-related deaths due to its rapid progression and high metastatic potential. Early and accurate detection is crucial for improving patient outcomes; however, existing deep learning models often struggle to balance diagnostic precision with real-time efficiency. This study presents the Melano Hybrid Model, a novel architecture that integrates the rapid detection capabilities of YOLOv9 with the boundary localization accuracy of Faster R-CNN through an adaptive feature fusion mechanism. The model was rigorously evaluated on three benchmark datasets—ISIC 2019, HAM10000, and ISIC 2020—using 5-fold cross-validation. On ISIC 2020, the hybrid model achieved a 96.2% classification accuracy (95% CI: 95.8–96.6%) and a 95.1% F1-score (95% CI: 94.7–95.5%), significantly outperforming standalone models (p < 0.001). The architecture delivers an average inference speed of 31.3 frames per second (FPS), surpassing clinical real-time thresholds. Additionally, computational profiling confirms its practical feasibility with 78.3 million parameters, 134.8 GFLOPs, and a 324 MB memory footprint. These results support the hybrid framework as a robust AI-assisted tool for real-world melanoma screening, offering an optimal trade-off between speed and diagnostic performance.
- Open Access
- 1 January 2024
- Published by Southwest Jiaotong University in Journal of Southwest Jiaotong University
- Vol. 59 (2)
MODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES
MODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES MODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES
- 1 January 2024
- Published by Southwest Jiaotong University in Journal of Southwest Jiaotong University
- Vol. 59 (2)
Open AccessMODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES
MODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES MODELS FOR FINANCIAL CRISIS PREDICTION USING DEEP LEARNING TECHNIQUES
- Open Access
- 31 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 124597-124608
Enhancing the Efficiency of Gaussian Naïve Bayes Machine Learning Classifier in the Detection of DDOS in Cloud Computing
Distributed Denial of services is one of the most dangerously planned attacks in cloud computing, resulting in huge losses of data and money for both the cloud services providers and the users of these services. Many efforts have been performed to help protect the cloud from these attacks using machine learning techniques. This study focuses on enhancing the efficiency of the Gaussian Naïve Bayes classifier, considered one of the cheapest and fastest classifiers. Still, it has some problems resulting from its equation’s statistical nature. The nature of this classifier is based on multiplication, resulting in inaccurate classification due to the zero-frequency issue and the fact that it assumes that features are independent. This research proposed a framework handling the selection of a set of highly independent features following an iterative feature selection approach using the Pearson Correlation Coefficient, Mutual Information, and Chi-squared and then selecting other subsets of features from these sets to reach a set of highly independent features. After that, we used a specific algorithm handling the data pre-processing to handle the zero-frequency problem where we used the Mode to replace the missing values, and if the mode was zero, it used the mean instead. Still, if the record’s label is zero, we get the value of the previous record with zero labels. After that, we handled the data imbalances using SMOTE. These enhancements increased both accuracy for the mutual information model by 2% and the average overall accuracy and precision by 1.5%.
- 31 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 124597-124608
Open AccessEnhancing the Efficiency of Gaussian Naïve Bayes Machine Learning Classifier in the Detection of DDOS in Cloud Computing
Distributed Denial of services is one of the most dangerously planned attacks in cloud computing, resulting in huge losses of data and money for both the cloud services providers and the users of these services. Many efforts have been performed to help protect the cloud from these attacks using machine learning techniques. This study focuses on enhancing the efficiency of the Gaussian Naïve Bayes classifier, considered one of the cheapest and fastest classifiers. Still, it has some problems resulting from its equation’s statistical nature. The nature of this classifier is based on multiplication, resulting in inaccurate classification due to the zero-frequency issue and the fact that it assumes that features are independent. This research proposed a framework handling the selection of a set of highly independent features following an iterative feature selection approach using the Pearson Correlation Coefficient, Mutual Information, and Chi-squared and then selecting other subsets of features from these sets to reach a set of highly independent features. After that, we used a specific algorithm handling the data pre-processing to handle the zero-frequency problem where we used the Mode to replace the missing values, and if the mode was zero, it used the mean instead. Still, if the record’s label is zero, we get the value of the previous record with zero labels. After that, we handled the data imbalances using SMOTE. These enhancements increased both accuracy for the mutual information model by 2% and the average overall accuracy and precision by 1.5%.