Browsing by Author "Saiful, Md."

Now showing 1 - 3 of 3

Brain MRI Classification for Alzheimer’s Disease Based on Convolutional Neural Network
(Springer, 2023-12-20) Saiful, Md.; Saha, Arpita; Mim, Faria Tabassum; Tasnim, Nafisa; Reza, Ahmed Wasif; Arefin, Mohammad Shamsul
Alzheimer’s disease is a severe disorder of the brain that gradually increases and affects the function of the brain. It mainly affects middle-aged people or old aged person. Many researchers tried to train their model to classify or detect Alzheimer’s disease from MRI images automatically. In this paper, we also tried to classify four classes (Mild Demented, Moderate Demented, Non-Demented, Very Mild Demented) of Alzheimer’s diseases using ResNet (Residual neural network) on 6400 MRI images. In the paper, ResNet50v2 and ResNet101v2 used. By comparing their performance, ResNet101v2 gave a better result. The model’s precision is 74%, 27%, 75%, and 54%, recall percentage is 28%, 25%, 65%, and 77%, and f1 scores are 40%, 26%, 70%, and 63% for mild demented, moderate demented, non-demented, and very mild demented, respectively. By applying ResNet101v2, the percentage of accuracy is 98.35%.
Chained semantic retrieval for rare disease and gene identification using clinical phenotype
(BRAC University, 2025-10) Saiful, Md.; Alam, Md. Golam Rabiul
Accurate identification of rare diseases and associated genes from patient phenotypic data represents a critical challenge in precision medicine and genomic research. Current diagnostic approaches face substantial limitations in scalability, interpretability, and real-time processing when integrating heterogeneous phenotypic and genotypic databases. This study presents a novel artificial intelligence-driven system employing chained semantic retrieval and knowledge graph integration to identify rare diseases, genes, and Human Phenotype Ontology (HPO) terms from patient phenotype descriptions. The methodology leverages sentence transformers (based on Bidirectional and Auto-Regressive Transformers) for generating vector embeddings, Facebook AI Similarity Search (FAISS) for efficient similarity computation, and a fine-tuned Llama 3.2 model integrated with a Retrieval-Augmented Generation (RAG) pipeline. The system implements a tiered scoring mechanism that chains retrieval across Human Phenotype Ontology, disease, and gene databases to progressively refine predictions through contextual enhancement. Evaluation on 50 patient phenotypes with confirmed Duchenne Muscular Dystrophy diagnosis, consisting of 30 true positive and 20 true negative cases, demonstrated strong performance with 28 true positives, 17 true negatives, 2 false negatives, and 3 false positives in top-ten retrieval results. The system achieved 90 % overall accuracy, 93.3 % recall, 85 % specificity, 90.3 % precision, and an F1-score of 91.8 %, with an average computational efficiency of 3.2 seconds per response. The proposed framework effectively addresses critical gaps in cross-database integration while maintaining interpretability through tiered confidence scoring for clinical decision support applications.
MRI-Based Brain Tumor Classification Using Various Deep Learning Convolutional Networks and CNN
(Springer Nature, 2023-11-11) Saiful, Md.; Haider, Sakib; Rahman, S. M. Arafat; Reza, Nahid; Reza, Ahmed Wasif; Arefin, Mohammad Shamsul
"Yearly, brain tumors cause many fatalities and a significant portion of these victims come from rural regions. However, beginning brain tumor diagnosis technology is not as effective as anticipated. We thus set out to develop an accurate approach that would aid doctors in recognizing brain tumors. Even though there have been several types of research on this topic, we tried to develop a classification approach that is significantly more accurate and error-free and is trained using a sizable amount of authentic datasets rather than an enhanced data-modified version of the VGG-16 convolutional neural network architecture was used to analyze a dataset of 6328 MRI images that were categorized into three different types: Pituitary, Glioma, and Meningioma. The results were highly impressive, with the model achieving an overall accuracy of 99.5%. The precision rates for each type were also outstanding, with a precision rate of 99.4% for gliomas, 96.7% for meningiomas, and 100% for pituitaries. These results suggest that the modified VGG-16 architecture is highly effective in accurately classifying MRI images of the brain into these three distinct categories. Additionally, it outperformed various other current CNN designs and cutting-edge research in terms of outcomes."