MTMst. Sazia Tahosin

Mst. Sazia Tahosin

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mst. Sazia Tahosin has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Improving Medical X-Ray Imaging Diagnosis With Attention Mechanisms and Robust Transfer Learning Techniques

    by Isha Das, Alif Sheakh, Shahab Abdulla, Mst. Sazia Tahosin, Mehedi Hassan, Sadika Zaman, Abhay Shukla

    X-ray imaging remains a cornerstone in medical diagnostics for conditions such as bone fractures, knee osteoarthritis, and lung diseases. However, variability in image quality and dataset diversity presents significant challenges for automated analysis using deep learning models. This study addresses these issues by proposing an EfficientNet B0 architecture enhanced with a Convolutional Block Attention Module (CBAM) to improve classification accuracy and interpretability across multiple X-ray datasets: FracAtlas, Knee, and Lung X-ray. A robust preprocessing pipeline comprising LAB color space conversion, morphological filtering, gamma correction, Non-Local Means denoising, resizing, and normalization was applied to optimize image quality, with each step’s effectiveness verified through established image quality metrics. Additionally, geometric augmentation techniques were performed to increase dataset variability and improve model generalization. Comparative experiments with transfer learning, transformer-based, and attention-based models identified the attention-based EfficientNet B0 as the best performer. An extensive ablation study on the Knee X-ray dataset refined hyperparameters to maximize performance. The optimized model achieved high test accuracies of 98.09%, 97.12%, and 99.51% on FracAtlas, Knee, and Lung datasets, respectively. Further analyses, including noise robustness testing, k-fold cross-validation, and Grad-CAM visualization, demonstrated the model’s resilience, consistency, and explainability. The findings highlight the critical role of attention mechanisms in enhancing feature representation and generalization across heterogeneous medical imaging tasks. This work lays a foundation for reliable, interpretable AI systems that can support clinical decision-making, with future efforts focusing on expanding dataset diversity and real-world clinical validation to accelerate adoption.


  • Open Access

    FedVGM: Enhancing Federated Learning Performance on Multi-Dataset Medical Images With XAI

    by Mst. Sazia Tahosin, Alif Sheakh, Mohammad Jahangir Alam, Mehedi Hassan, Anupam Kumar Bairagi, Shahab Abdulla, Samah Alshathri, Walid El-Shafai

    Advances in deep learning have transformed medical imaging, yet progress is hindered by data privacy regulations and fragmented datasets across institutions. To address these challenges, we propose FedVGM, a privacy-preserving federated learning framework for multi-modal medical image analysis. FedVGM integrates four imaging modalities, including brain MRI, breast ultrasound, chest X-ray, and lung CT, across 14 diagnostic classes without centralizing patient data. Using transfer learning and an ensemble of VGG16 and MobileNetV2, FedVGM achieves 97.7% $\pm$ 0.01 accuracy on the combined dataset and 91.9–99.1% across individual modalities. We evaluated three aggregation strategies and demonstrated median aggregation to be the most effective. To ensure clinical interpretability, we apply explainable AI techniques and validate results through performance metrics, statistical analysis, and k-fold cross-validation. FedVGM offers a robust, scalable solution for collaborative medical diagnostics, supporting clinical deployment while preserving data privacy.


  • Open Access

    A Hybrid Transfer Learning Framework for Brain Tumor Diagnosis

    by Sadia Islam Tonni, Alif Sheakh, Mst. Sazia Tahosin, Zahid Hasan, Taslima Ferdaus Shuva, Touhid Bhuiyan, Muhammad Ali Abdullah Almoyad, Nabil Anan Orka, Tanvir Rahman, Risala Tasin Khan, et al.

    Brain tumors are among the most severe health challenges, necessitating early and precise diagnosis for effective treatment planning. This study introduces an optimized hybrid transfer learning (TL) framework for brain tumor classification using magnetic resonance imaging images. The proposed system integrates advanced preprocessing techniques, an ensemble of pretrained deep learning models, and explainable artificial intelligence (XAI) methods to achieve high accuracy and reliability. The methodology enhances image quality through noise reduction and contrast enhancement, facilitating robust feature extraction. The ensemble model combines VGG16 and ResNet152V2 architectures, achieving a classification accuracy of 99.47% on a challenging four‐class dataset. Additionally, gradient‐weighted class activation mapping and SHapley Additive exPlanations (SHAP)‐based XAI techniques provide visual and quantitative insights into model predictions, improving interpretability and clinical trust. This comprehensive framework demonstrates the potential of hybrid TL and XAI in advancing diagnostic accuracy and supporting clinical decision‐making for brain tumor detection. The results underscore its applicability in clinical settings, particularly in resource‐constrained environments.


  • Optimizing medical image analysis through MViTX on multiple datasets with explainable AI

    by Alif Sheakh, Mst. Sazia Tahosin, Mohammad Jahangir Alam, Mahbuba Begum


  • Open Access

    Enhancing Bangla Fake News Detection Using Bidirectional Gated Recurrent Units and Deep Learning Techniques

    by Utsha Roy, Mst Sazia Tahosin, Mahedi Hasan, Taminul Islam, Fahim Imtiaz, Rezwane Sadik, Yassine Maleh, Rejwan Bin Sulaiman, Simul Hassan Talukder

    The rise of fake news has made the need for effective detection methods, including in languages other than English, increasingly important. The study aims to address the challenges of Bangla which is considered a less important language. To this end, a complete dataset containing about 50,000 news items is proposed. Several deep learning models have been tested on this dataset, including the bidirectional gated recurrent unit (GRU), the long short-term memory (LSTM), the 1D convolutional neural network (CNN), and hybrid architectures. For this research, we assessed the efficacy of the model utilizing a range of useful measures, including recall, precision, F1 score, and accuracy. This was done by employing a big application. We carry out comprehensive trials to show the effectiveness of these models in identifying bogus news in Bangla, with the Bidirectional GRU model having a stunning accuracy of 99.16%. Our analysis highlights the importance of dataset balance and the need for continual improvement efforts to a substantial degree. This study makes a major contribution to the creation of Bangla fake news detecting systems with limited resources, thereby setting the stage for future improvements in the detection process.


  • Open Access

    Enhancing Bangla Fake News Detection Using Bidirectional Gated Recurrent Units and Deep Learning Techniques

    by Utsha Roy, Mst. Sazia Tahosin, Mahedi Hassan, Taminul Islam, Fahim Imtiaz, Rezwane Sadik, Yassine Maleh, Rejwan Bin Sulaiman, Simul Hasan Talukder

    The rise of fake news has made the need for effective detection methods, including in languages other than English, increasingly important. The study aims to address the challenges of Bangla which is considered a less important language. To this end, a complete dataset containing about 50,000 news items is proposed. Several deep learning models have been tested on this dataset, including the bidirectional gated recurrent unit (GRU), the long short-term memory (LSTM), the 1D convolutional neural network (CNN), and hybrid architectures. For this research, we assessed the efficacy of the model utilizing a range of useful measures, including recall, precision, F1 score, and accuracy. This was done by employing a big application. We carry out comprehensive trials to show the effectiveness of these models in identifying bogus news in Bangla, with the Bidirectional GRU model having a stunning accuracy of 99.16%. Our analysis highlights the importance of dataset balance and the need for continual improvement efforts to a substantial degree. This study makes a major contribution to the creation of Bangla fake news detecting systems with limited resources, thereby setting the stage for future improvements in the detection process.

  • Child and Maternal Mortality Risk Factor Analysis Using Machine Learning Approaches

    by Alif Sheakh, Mst. Sazia Tahosin, Maruf Hasan, Taminul Islam, Onamika Islam, Masum Rana

    Global attention is now being paid to maternal and child mortality. The incidence of maternal mortality is high in low and middle-income countries, particularly among adolescents and young adults. Healthcare professionals can monitor the mother's heartbeat during pregnancy to determine fetal viability using CTGs to prevent these deaths. To reduce child and maternal mortality, this work presented a risk factor analysis using machine learning approaches. As part of this study, this work evaluated seven machine learning algorithms. To assess the performance of different categorization algorithms, accuracy, precision, and recall were used. The random forest has achieved the highest 99.98% accuracy among the other algorithms. Initially, the dataset was imbalanced, after applying undersampling and oversampling methods, all algorithms performed excellently. A major focus of the present study was to predict the risk factor of child and maternal mortality using clinical data. Sending an ultrasound pulse and reading the response is how ultrasound devices work. To prevent child and maternal mortality, this analysis is an effective and cost-effective option for healthcare professionals.