Afnan M. Alhassan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Afnan M. Alhassan has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
Nouf I. Altmami
Department of Computer Science, College of Computing and Information Technology, Shaqra University, Shaqra, Saudi Arabia
5 shared publications- Wan Mohd Nazmee Wan Zainon
School of Computer Sciences, Universiti Sains Malaysia, USM, Penang, Malaysia
3 shared publications - Nouf I. Altmami
Department of Computer Science, College of Computing and Information Technology, Shaqra University, Shaqra 11961, Saudi Arabia
2 shared publications - Nouf Ibrahim Altmami
Department of Computer Science, College of Computing and Information Technology, Shaqra University, Shaqra, Saudi Arabia
2 shared publications Wan Mohd Nazmee Wan Zainon
School of Computer Science, Universiti Sains Malaysia, George Town, Malaysia
2 shared publicationsNouf Altmami
Department of Computer Science, College of Computing and Information Technology, Shaqra University, Shaqra, Saudi Arabia
1 shared publication
Latest Publications
- Open Access
- 20 February 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (2) , e0342592
Enhanced medical image segmentation using optimized bidirectional LSTM and dolphin partner optimizer
Medical imaging has become a standard in diagnosing and treating how organs and tissues operate. In earlier systems, machine learning approaches were used for segmentation over a long period. Thus, processing and analyzing these medical images are highly important for clinical diagnosis. Deep learning-based image segmentation has attracted a lot of interest in recent years. This paper introduces the Optimized Bidirectional Long-Short Term Memory (OBi-LSTM) segmented technique for medical image classification. The OBi-LSTM classifier contains the sequencing data in both directions, backward and forward. It is utilized to simultaneously be aware of the most significant spatial placements, channels, and scales. Using Dolphin Partner Optimizer (DPO), the weight and bias parameters of the Bi-LSTM classifier are tuned for each cell. Since these variables are shared across the entire sequence, the network can maintain effective representation while requiring fewer unique weight parameters and hidden neurons, reducing overall computational complexity. OBi-LSTM classification outperforms state-of-the-art techniques in the MRI segmented test, demonstrating the proposed approach’s usefulness and strong explanatory power. The proposed OBi-LSTM model achieves a high dice similarity coefficient of 94.05%, a Jaccard resemblance index of 88.77%, and an accuracy ratio of 93.05% compared to other existing models. The proposed OBi-LSTM + DPO model significantly enhances medical image segmentation performance, particularly in boundary delineation and feature extraction.
- 20 February 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (2) , e0342592
Open AccessEnhanced medical image segmentation using optimized bidirectional LSTM and dolphin partner optimizer
Medical imaging has become a standard in diagnosing and treating how organs and tissues operate. In earlier systems, machine learning approaches were used for segmentation over a long period. Thus, processing and analyzing these medical images are highly important for clinical diagnosis. Deep learning-based image segmentation has attracted a lot of interest in recent years. This paper introduces the Optimized Bidirectional Long-Short Term Memory (OBi-LSTM) segmented technique for medical image classification. The OBi-LSTM classifier contains the sequencing data in both directions, backward and forward. It is utilized to simultaneously be aware of the most significant spatial placements, channels, and scales. Using Dolphin Partner Optimizer (DPO), the weight and bias parameters of the Bi-LSTM classifier are tuned for each cell. Since these variables are shared across the entire sequence, the network can maintain effective representation while requiring fewer unique weight parameters and hidden neurons, reducing overall computational complexity. OBi-LSTM classification outperforms state-of-the-art techniques in the MRI segmented test, demonstrating the proposed approach’s usefulness and strong explanatory power. The proposed OBi-LSTM model achieves a high dice similarity coefficient of 94.05%, a Jaccard resemblance index of 88.77%, and an accuracy ratio of 93.05% compared to other existing models. The proposed OBi-LSTM + DPO model significantly enhances medical image segmentation performance, particularly in boundary delineation and feature extraction.
- Open Access
- 19 February 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (4)
D2EBTN: Deep stacked dual learning ensembled attentional-based deep learning model for malware detection in the cloud
Cloud computing is quite popular because of its scalability and flexibility merits. However, faces significant security risks from malware, which exploits infrastructure weaknesses to cause unauthorized access, identity theft, and data breaches. Most current deep learning approaches struggle to detect malware effectively due to significant challenges like high computational costs, issues with overfitting, lack of interpretability, and the potential for inaccurate results. As a consequence, the research proposes a Deep Stacked Dual Learning Ensembled Attentional Bidirectional Long Short-Term Memory and Deep Belief Network (D2EBTN) for malware detection in the cloud. The D2EBTN utilizes a stacked dual learning strategy with the combination of incremental and federated learning concepts, offering a robust, efficient, and privacy-preserving malware detection across distributed systems. In addition to that, the combination of the ensemble attention module and the deep belief network enhances performance accuracy. Leverages the distinct strengths of both components to address the complex and evolving nature of malware threats, thus addressing existing challenges. Owing to these advancements, the model outperforms state-of-the-art methods and attains a precision of 98.63%, sensitivity of 97.75%, F1-score of 98.19%, specificity of 98.41%, and accuracy of 98.04% under the CIC-IDS2017 dataset.
- 19 February 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (4)
Open AccessD2EBTN: Deep stacked dual learning ensembled attentional-based deep learning model for malware detection in the cloud
Cloud computing is quite popular because of its scalability and flexibility merits. However, faces significant security risks from malware, which exploits infrastructure weaknesses to cause unauthorized access, identity theft, and data breaches. Most current deep learning approaches struggle to detect malware effectively due to significant challenges like high computational costs, issues with overfitting, lack of interpretability, and the potential for inaccurate results. As a consequence, the research proposes a Deep Stacked Dual Learning Ensembled Attentional Bidirectional Long Short-Term Memory and Deep Belief Network (D2EBTN) for malware detection in the cloud. The D2EBTN utilizes a stacked dual learning strategy with the combination of incremental and federated learning concepts, offering a robust, efficient, and privacy-preserving malware detection across distributed systems. In addition to that, the combination of the ensemble attention module and the deep belief network enhances performance accuracy. Leverages the distinct strengths of both components to address the complex and evolving nature of malware threats, thus addressing existing challenges. Owing to these advancements, the model outperforms state-of-the-art methods and attains a precision of 98.63%, sensitivity of 97.75%, F1-score of 98.19%, specificity of 98.41%, and accuracy of 98.04% under the CIC-IDS2017 dataset.
- Open Access
- 26 January 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (3) , 393
Quadra Sense: A Fusion of Deep Learning Classifiers for Mitosis Detection in Breast Cancer Histopathology
Background/Objectives: The difficulties caused by breast cancer have been addressed in a number of ways. Since it is said to be the second most common cause of death from cancer among women, early intervention is crucial. Early detection is difficult because of the existing detection tools’ shortcomings in objectivity and accuracy. Quadra Sense, a fusion of deep learning (DL) classifiers for mitosis detection in breast cancer histopathology, is proposed to address the shortcomings of current approaches. It demonstrates a greater capacity to produce more accurate results. Methods: Initially, the raw dataset is preprocessed by using a normalization by means of color channel normalization (zero-mean normalization) and stain normalization (Macenko Stain Normalization), and the artifact can be removed via median filtering and contrast enhancement using histogram equalization; ROI identification is performed using modified Fully Convolutional Networks (FCNs) followed by the feature extraction (FE) with Modified InceptionV4 (M-IV4), by which the deep features are retrieved and the feature are selected by means of a Self-Improved Seagull Optimization Algorithm (SA-SOA), and finally, classification is performed using Mito-Quartet. Results: Ultimately, using a performance evaluation, the suggested approach achieved a higher accuracy of 99.2% in comparison with the current methods. Conclusions: From the outcomes, the recommended technique performs well.
- 26 January 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (3) , 393
Open AccessQuadra Sense: A Fusion of Deep Learning Classifiers for Mitosis Detection in Breast Cancer Histopathology
Background/Objectives: The difficulties caused by breast cancer have been addressed in a number of ways. Since it is said to be the second most common cause of death from cancer among women, early intervention is crucial. Early detection is difficult because of the existing detection tools’ shortcomings in objectivity and accuracy. Quadra Sense, a fusion of deep learning (DL) classifiers for mitosis detection in breast cancer histopathology, is proposed to address the shortcomings of current approaches. It demonstrates a greater capacity to produce more accurate results. Methods: Initially, the raw dataset is preprocessed by using a normalization by means of color channel normalization (zero-mean normalization) and stain normalization (Macenko Stain Normalization), and the artifact can be removed via median filtering and contrast enhancement using histogram equalization; ROI identification is performed using modified Fully Convolutional Networks (FCNs) followed by the feature extraction (FE) with Modified InceptionV4 (M-IV4), by which the deep features are retrieved and the feature are selected by means of a Self-Improved Seagull Optimization Algorithm (SA-SOA), and finally, classification is performed using Mito-Quartet. Results: Ultimately, using a performance evaluation, the suggested approach achieved a higher accuracy of 99.2% in comparison with the current methods. Conclusions: From the outcomes, the recommended technique performs well.
- Open Access
- 22 January 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (2) , 359
DL-PCMNet: Distributed Learning Enabled Parallel Convolutional Memory Network for Skin Cancer Classification with Dermatoscopic Images
Background/Objectives: Globally, one of the most dreadful and rapidly spreading illnesses is skin cancer, and it is acknowledged as a lethal form of cancer due to the abnormal growth of skin cells. Mostly, classifying and diagnosing the types of skin lesions is complex, and recognizing tumors from dermoscopic images remains challenging. The existing methods have limitations like insufficient datasets, computational complexity, class imbalance issues, and poor classification performance. Methods: This research presents a method named the Distributed Learning enabled Parallel Convolutional Memory Network (DL-PCMNet) model to effectively classify skin cancer by overcoming the existing limitations. Hence, the proposed DL-PCMNet model utilizes a distributed learning framework to provide greater flexibility during the learning process, and it increases the reliability of the model. Moreover, the model integrates the Convolutional Neural Network (CNN) and Long Short-Term Memory model (LSTM) in a parallel distribution, which enhances robustness and accuracy by capturing the information of long-term dependencies. Furthermore, the utilization of advanced preprocessing and feature extraction techniques increases the accuracy of classification. Results: The evaluation results exhibit an accuracy of 97.28%, precision of 97.30%, sensitivity of 97.17%, and specificity of 97.72% at 90% of training by using the ISIC 2019 skin lesion dataset, respectively. Conclusions: Specifically, the proposed DL-PCMNet model achieved efficient and accurate skin cancer classification compared with other existing models.
- 22 January 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (2) , 359
Open AccessDL-PCMNet: Distributed Learning Enabled Parallel Convolutional Memory Network for Skin Cancer Classification with Dermatoscopic Images
Background/Objectives: Globally, one of the most dreadful and rapidly spreading illnesses is skin cancer, and it is acknowledged as a lethal form of cancer due to the abnormal growth of skin cells. Mostly, classifying and diagnosing the types of skin lesions is complex, and recognizing tumors from dermoscopic images remains challenging. The existing methods have limitations like insufficient datasets, computational complexity, class imbalance issues, and poor classification performance. Methods: This research presents a method named the Distributed Learning enabled Parallel Convolutional Memory Network (DL-PCMNet) model to effectively classify skin cancer by overcoming the existing limitations. Hence, the proposed DL-PCMNet model utilizes a distributed learning framework to provide greater flexibility during the learning process, and it increases the reliability of the model. Moreover, the model integrates the Convolutional Neural Network (CNN) and Long Short-Term Memory model (LSTM) in a parallel distribution, which enhances robustness and accuracy by capturing the information of long-term dependencies. Furthermore, the utilization of advanced preprocessing and feature extraction techniques increases the accuracy of classification. Results: The evaluation results exhibit an accuracy of 97.28%, precision of 97.30%, sensitivity of 97.17%, and specificity of 97.72% at 90% of training by using the ISIC 2019 skin lesion dataset, respectively. Conclusions: Specifically, the proposed DL-PCMNet model achieved efficient and accurate skin cancer classification compared with other existing models.
- Open Access
- 1 January 2026
- Published by American Institute of Mathematical Sciences (AIMS) in AIMS Mathematics
- Vol. 11 (3) , 7945-7979
S2A-DL-EGBNet: Land Cover classification using an Attention-enabled Distributed Learning-based Encoder Generative Bidirectional Network
Land Cover (LC) classification is an effective technique that categorizes the Earth's surface into urban, forest, and agricultural land classes by utilizing remote sensing data. Nevertheless, the hyperspectral remote sensing images are afflicted with a lack of labeled data, spectral variability, and the curse of dimensionality, which limit their competence in remote sensing applications. Hence, this research presents a Spatial Split Attention-enabled Distributed Learning-based Encoder Generative Bidirectional Network (S2A-DL-EGBNet) model for accurate LC classification using hyperspectral images (HSIs). The model integrates a distributed learning module to process large datasets, and training is done in parallel by reducing complexity while improving scalability. Also, a Generative Adversarial Network (GAN)-based data balancing is employed to address the class imbalance problem and enables the model's effectiveness. Thereafter, the parallel Bidirectional Long Short-Term Memory (BiLSTM) is integrated to speed up the training process and minimize computation time. The research applies multiple feature extraction techniques, capturing complex features, scaling variations in photographic distortions, and illumination changes from the aspects of input data. Notably, the S2A-DL-EGBNet model is validated using the Hyperspectral Remote Sensing Scenes dataset, and the S2A-DL-EGBNet model shows remarkable performance by achieving 98.44% sensitivity, 98.84% accuracy, and 99.24% sensitivity on the Indian Pines dataset for training data of 90%.
- 1 January 2026
- Published by American Institute of Mathematical Sciences (AIMS) in AIMS Mathematics
- Vol. 11 (3) , 7945-7979
Open AccessS2A-DL-EGBNet: Land Cover classification using an Attention-enabled Distributed Learning-based Encoder Generative Bidirectional Network
Land Cover (LC) classification is an effective technique that categorizes the Earth's surface into urban, forest, and agricultural land classes by utilizing remote sensing data. Nevertheless, the hyperspectral remote sensing images are afflicted with a lack of labeled data, spectral variability, and the curse of dimensionality, which limit their competence in remote sensing applications. Hence, this research presents a Spatial Split Attention-enabled Distributed Learning-based Encoder Generative Bidirectional Network (S2A-DL-EGBNet) model for accurate LC classification using hyperspectral images (HSIs). The model integrates a distributed learning module to process large datasets, and training is done in parallel by reducing complexity while improving scalability. Also, a Generative Adversarial Network (GAN)-based data balancing is employed to address the class imbalance problem and enables the model's effectiveness. Thereafter, the parallel Bidirectional Long Short-Term Memory (BiLSTM) is integrated to speed up the training process and minimize computation time. The research applies multiple feature extraction techniques, capturing complex features, scaling variations in photographic distortions, and illumination changes from the aspects of input data. Notably, the S2A-DL-EGBNet model is validated using the Hyperspectral Remote Sensing Scenes dataset, and the S2A-DL-EGBNet model shows remarkable performance by achieving 98.44% sensitivity, 98.84% accuracy, and 99.24% sensitivity on the Indian Pines dataset for training data of 90%.
- Open Access
- 1 January 2026
- Published by American Institute of Mathematical Sciences (AIMS) in AIMS Mathematics
- Vol. 11 (2) , 5062-5091
Echo whale optimization-based incentive learning: A bidirectional long short-term memory model for intrusion detection in wireless networks
Intrusion detection distinguishes unauthorized system access by monitoring network activities, where identifiable patterns and behaviors are difficult to recognize due to factors such as false positives and negatives. The inability to work on large and complex networks, poor generalizability, high computation time, and the adoption of new threats that emerge over time are major concerns. Therefore, we established a model for handling the aforementioned challenges using the Echo whale optimization-incentive learning-based Bidirectional Long Short-Term Memory (EWO-IL-BiLSTM) model, which was proposed to improve the accuracy of intrusion detection. The EWO algorithm was incorporated to effectively tune the best parameters of the BiLSTM model to improve detection accuracy while increasing convergence speed and reducing false errors. Through incentive learning integration, the system demonstrated an award mechanism during training, which effectively captures the temporal dependencies in the network data and prompts the BiLSTM classifier to reach the best results. This research demonstrated that the benefits of the EWO and the Incentive Learning mechanism-based BiLSTM are a major advancement that improves the accurate intrusion detection, thereby promoting the network security and robustness in the ever-changing cyber-threat environments. The evaluation findings indicated that the proposed EWO-IL-BiLSTM model demonstrated superior performance with a high accuracy of 97.06%, F1-Score of 96.78%, precision of 97.59%, and recall of 95.98% at 80% training using the WSN-BFSF dataset.
- 1 January 2026
- Published by American Institute of Mathematical Sciences (AIMS) in AIMS Mathematics
- Vol. 11 (2) , 5062-5091
Open AccessEcho whale optimization-based incentive learning: A bidirectional long short-term memory model for intrusion detection in wireless networks
Intrusion detection distinguishes unauthorized system access by monitoring network activities, where identifiable patterns and behaviors are difficult to recognize due to factors such as false positives and negatives. The inability to work on large and complex networks, poor generalizability, high computation time, and the adoption of new threats that emerge over time are major concerns. Therefore, we established a model for handling the aforementioned challenges using the Echo whale optimization-incentive learning-based Bidirectional Long Short-Term Memory (EWO-IL-BiLSTM) model, which was proposed to improve the accuracy of intrusion detection. The EWO algorithm was incorporated to effectively tune the best parameters of the BiLSTM model to improve detection accuracy while increasing convergence speed and reducing false errors. Through incentive learning integration, the system demonstrated an award mechanism during training, which effectively captures the temporal dependencies in the network data and prompts the BiLSTM classifier to reach the best results. This research demonstrated that the benefits of the EWO and the Incentive Learning mechanism-based BiLSTM are a major advancement that improves the accurate intrusion detection, thereby promoting the network security and robustness in the ever-changing cyber-threat environments. The evaluation findings indicated that the proposed EWO-IL-BiLSTM model demonstrated superior performance with a high accuracy of 97.06%, F1-Score of 96.78%, precision of 97.59%, and recall of 95.98% at 80% training using the WSN-BFSF dataset.
- Open Access
- 29 December 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (12) , e0339822
Intrusion detection using search-based learning optimized ensemble tree classifier model
An Intrusion Detection System (IDS) is an important component of cybersecurity, meant to monitor malicious behaviour, detect, and respond to unauthorized activities in computer systems or networks. Generally, Intrusion detection (IDS) is classified into host-based IDS (HIDS) and network-based IDS (NIDS), which monitor individual devices and network traffic, respectively. Existing models faced certain limitations, including the dilemma of balancing false positives against false negatives, the challenge of adjusting to evolving threats, handling issues with high-dimensional information and encrypted traffic, and limited resource competence when dealing with privacy concerns. The proposed research work currently aims at developing an intrusion detection system that is more adaptive and effective to hinder these existing challenges and improve the security of digital environments. The study is related to applying an elaborate Search-based learning-optimized ensemble tree classifier (SBO-based ensemble tree classifier) for improving ID in Vehicular Ad Hoc Networks (VANETs). The ensemble classifier incorporates decision tree, random forest, extra tree, and eXtreme Gradient Boosting (XG Boost) classifiers, which are fused to provide a comprehensive interpretation of potential attacks within the VANET environment. Moreover, the research is enriched by incorporating Search-based learning optimization that takes advantage of their collective and adaptive nature. This innovative amalgamation attempts to perfect the aggregated response generated by the ensemble classifier, which fine-tunes the proposed model for effective intrusion detection. To facilitate the multi-dimensional orientation, four separate outputs, such as alpha, beta, gamma, and delta, were introduced, which allow the categorization of intrusion attacks based on specific types. More specifically, the experimental results illustrate that the proposed SBO-based ensemble tree classifier achieved superior performance with an accuracy of 96.56%, F1-score of 96.63%, FPR of 0.97, MCC of 0.97, Precision of 96.59%, Sensitivity of 96.68%, and Specificity of 96.52% for intrusion detection and outperforms the other existing methods using the BOT-IOT Dataset.
- 29 December 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (12) , e0339822
Open AccessIntrusion detection using search-based learning optimized ensemble tree classifier model
An Intrusion Detection System (IDS) is an important component of cybersecurity, meant to monitor malicious behaviour, detect, and respond to unauthorized activities in computer systems or networks. Generally, Intrusion detection (IDS) is classified into host-based IDS (HIDS) and network-based IDS (NIDS), which monitor individual devices and network traffic, respectively. Existing models faced certain limitations, including the dilemma of balancing false positives against false negatives, the challenge of adjusting to evolving threats, handling issues with high-dimensional information and encrypted traffic, and limited resource competence when dealing with privacy concerns. The proposed research work currently aims at developing an intrusion detection system that is more adaptive and effective to hinder these existing challenges and improve the security of digital environments. The study is related to applying an elaborate Search-based learning-optimized ensemble tree classifier (SBO-based ensemble tree classifier) for improving ID in Vehicular Ad Hoc Networks (VANETs). The ensemble classifier incorporates decision tree, random forest, extra tree, and eXtreme Gradient Boosting (XG Boost) classifiers, which are fused to provide a comprehensive interpretation of potential attacks within the VANET environment. Moreover, the research is enriched by incorporating Search-based learning optimization that takes advantage of their collective and adaptive nature. This innovative amalgamation attempts to perfect the aggregated response generated by the ensemble classifier, which fine-tunes the proposed model for effective intrusion detection. To facilitate the multi-dimensional orientation, four separate outputs, such as alpha, beta, gamma, and delta, were introduced, which allow the categorization of intrusion attacks based on specific types. More specifically, the experimental results illustrate that the proposed SBO-based ensemble tree classifier achieved superior performance with an accuracy of 96.56%, F1-score of 96.63%, FPR of 0.97, MCC of 0.97, Precision of 96.59%, Sensitivity of 96.68%, and Specificity of 96.52% for intrusion detection and outperforms the other existing methods using the BOT-IOT Dataset.
- Open Access
- 24 October 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (10) , e0335397
IV3TM: Inception V3 enabled bidirectional long short-term memory network for brain tumor classification
A brain tumor is one of the life-threatening neurological conditions affecting millions of people worldwide. Early diagnosis and classification of brain tumor types facilitate prompt treatment, thereby increasing the patient’s chances of survival. The advent of Deep Learning methods has significantly improved the field of medical image classification and aids neurologists in brain tumor diagnosis. However, the existing methods using Magnetic Resonance Imaging (MRI) face significant difficulties due to the complexities of brain tumors and the variability in tumor characteristics. Consequently, this research proposes the Inception V3 enabled Bidirectional Long Short Term Memory Network (IV3TM) for Brain Tumor Classification. In the proposed approach, the preprocessing and data augmentation techniques are presented to enhance classification performance. At the pre-processing stage, an iterative weighted-mean Filter approach is utilized to cope with bias field-effect fluctuations, noise, and blurring in input images to enhance the edges. Further, the data augmentation strategy increases the size of the available training data. SqueezeNet is used to segment images for further classification operations. Further, the proposed model combines the strengths of Inception V3 and BiLSTM to learn the sequential dependencies significant for understanding the intricate structural relationships in brain MRI data. The effectiveness of the proposed method is evaluated using several metrics, including specificity, accuracy, precision, F1-score, and sensitivity. Furthermore, the proposed method’s error is evaluated using root mean square error (RMSE). Experiments using the Brain Magnetic Resonance Imaging (MRI) images dataset and Figshare brain tumor datasets have shown encouraging results.
- 24 October 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (10) , e0335397
Open AccessIV3TM: Inception V3 enabled bidirectional long short-term memory network for brain tumor classification
A brain tumor is one of the life-threatening neurological conditions affecting millions of people worldwide. Early diagnosis and classification of brain tumor types facilitate prompt treatment, thereby increasing the patient’s chances of survival. The advent of Deep Learning methods has significantly improved the field of medical image classification and aids neurologists in brain tumor diagnosis. However, the existing methods using Magnetic Resonance Imaging (MRI) face significant difficulties due to the complexities of brain tumors and the variability in tumor characteristics. Consequently, this research proposes the Inception V3 enabled Bidirectional Long Short Term Memory Network (IV3TM) for Brain Tumor Classification. In the proposed approach, the preprocessing and data augmentation techniques are presented to enhance classification performance. At the pre-processing stage, an iterative weighted-mean Filter approach is utilized to cope with bias field-effect fluctuations, noise, and blurring in input images to enhance the edges. Further, the data augmentation strategy increases the size of the available training data. SqueezeNet is used to segment images for further classification operations. Further, the proposed model combines the strengths of Inception V3 and BiLSTM to learn the sequential dependencies significant for understanding the intricate structural relationships in brain MRI data. The effectiveness of the proposed method is evaluated using several metrics, including specificity, accuracy, precision, F1-score, and sensitivity. Furthermore, the proposed method’s error is evaluated using root mean square error (RMSE). Experiments using the Brain Magnetic Resonance Imaging (MRI) images dataset and Figshare brain tumor datasets have shown encouraging results.
- Open Access
- 16 October 2025
- Published by Elsevier in Ain Shams Engineering Journal
- Vol. 16 (12)
An efficient Alzheimer’s disease detection using optimized hybrid light GBM-based CNN
- 16 October 2025
- Published by Elsevier in Ain Shams Engineering Journal
- Vol. 16 (12)
Open AccessAn efficient Alzheimer’s disease detection using optimized hybrid light GBM-based CNN
- 24 September 2025
- Published by Elsevier in Measurement
- Vol. 258
IoT-based anomaly detection with the hybrid activation and the attention mechanism enabled LSTM-GRU
- 24 September 2025
- Published by Elsevier in Measurement
- Vol. 258
IoT-based anomaly detection with the hybrid activation and the attention mechanism enabled LSTM-GRU