Fatima Salman

Pharmacovigilance is a critical pillar of modern healthcare, yet the rapid expansion of drug inventories makes manual risk classification increasingly impractical. This paper proposes a validated unsupervised machine learning framework for automated drug segmentation based on the severity of reported side effects. Using a dataset of 11,498 unique pharmaceutical products, we apply a hybrid feature extraction pipeline combining TF-IDF text vectorization, Truncated Singular Value Decomposition (SVD), and a domain-informed severity scoring model. To ensure clinical accuracy, the scoring model was validated by a panel of clinical pharmacists, achieving high inter-rater reliability (Cohen’s Kappa = 0.82). K-Means clustering (k=4) partitions the drug portfolio into four distinct risk strata: Low Risk (42.4%), Moderate Risk (35.0%), High Risk (18.4%), and Critical Risk (4.2%). Cluster validity is confirmed through Silhouette, Davies-Bouldin, and Calinski-Harabasz indices, while sensitivity analysis demonstrates the model’s robustness against varying weight configurations. Results reveal an empirical "Review Paradox"—where high-risk medications receive superior patient ratings—which analysis of the "Uses" field attributes to selection bias in life-saving treatments, such as oncological agents. The framework’s external validity is further established through high concordance with U.S. FDA Black Box Warnings, particularly within the Critical Risk segment, and demonstrates clear alignment with global Anatomical Therapeutic Chemical (ATC) classification systems. By replacing generic token counts with a refined "Residual Side Effect Density" metric, the model maintains high interpretability and scalability. This research provides a robust, explainable methodology for drug safety stratification, offering a scalable adjunct to regulatory oversight and clinical decision support systems.

The proliferation of AI-powered tools in academic settings has prompted urgent questions about their cognitive and epistemic effects on researchers. Among emerging tools, Google's NotebookLM occupies a distinctive position: unlike general- purpose large language models (LLMs) such as ChatGPT or Gemini, it operates exclusively within a user-defined document corpus, grounding every response in cited, user-supplied sources. This architecture raises theoretically significant questions about cognitive load and research originality — two constructs central to the quality and integrity of academic knowledge production. This paper presents a critical narrative review of the literature at the intersection of source-grounded AI, Cognitive Load Theory (CLT), and academic originality, with particular attention to what current evidence suggests — and fails to address, about NotebookLM specifically. Drawing on 45 peer-reviewed studies and technical reports published between 2015 and 2024, we identify four major themes: (1) the cognitive affordances and constraints of source-grounded versus open-ended AI retrieval; (2) the applicability of CLT's load taxonomy to AI-mediated research tasks; (3) empirical evidence on AI's effects on creative and scholarly originality; and (4) the methodological and ethical challenges of studying AI-human research collaboration. We synthesise these themes into a conceptual framework — the Grounded Retrieval–Cognitive Redistribution (GRCR) model — that predicts when and for whom source-grounded AI tools are likely to reduce extraneous cognitive load without displacing germane elaboration. We conclude by identifying critical gaps in the literature and proposing a research agenda for empirical investigation of NotebookLM and similar grounded-retrieval systems in academic contexts.

The rapid advancement of generative artificial intelligence (AI) has introduced unprecedented challenges in distinguishing authentic human faces from synthetically generated counterparts. Existing research predominantly focuses on binary classification — real versus fake — without differentiating between distinct forgery mechanisms. This paper presents the first publicly available multi-class face forgery dataset comprising four categories: real photographs (1,081 images), Photoshop-manipulated faces (960 images), Generative Adversarial Network (GAN)-generated faces (2,000 images), and Diffusion model-generated faces (2,000 images), totaling 6,041 images. We conduct a comprehensive comparative evaluation of three state-of-the-art deep learning architectures — ResNet-50, EfficientNet-B4, and Swin Transformer (Swin-T) — under identical experimental conditions. The Swin Transformer achieves the highest test accuracy of 93.38%, with a macro-average F1-score of 90.32%, demonstrating superior capability in capturing fine-grained forgery artifacts. The results underscore that the Swin Transformer’s superiority stems from its ability to model long-range spatial dependencies, which proved instrumental in detecting global artifacts characteristic of diffusion- based synthesis. Notably, GAN and Diffusion-generated faces are detected with 100% accuracy, while Photoshop-manipulated faces remain the most challenging category (F1: 80%). Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations reveal distinct spatial attention patterns per forgery type, providing interpretable insights into model decision-making. The dataset is publicly released on Kaggle to facilitate future research in multi-class deepfake detection.

X-rays imaging is one of the possible method for detecting COVID19. Our research aimed to construct a model using deep learning for detecting COVID-19 pneumonia on high resolution X-rays, relieve working pressure of radiologists and contribute to the control of the epidemic. Methods: For the proposed model development, validation, and testing 260 images available from the repository of GitHub and Kaggle were used. The Images consists of 130 COVID-19 (ignoring SARS, MERS and ARDS) and 130 Normal X-ray images. Findings: The proposed model achieved: sensitivity of 100%, specificity of 100%, accuracy of 100%, PPV of 100%, and NPV of 100% in the dataset. For 260 images, the model achieved a comparable performance to that of expert radiologist. With the assistance of the model, the reading time of the radiologists will greatly be decreased. Conclusion: The deep learning model showed a comparable performance with expert radiologist, and greatly will improve the efficiency of radiologists in clinical practice. It holds great potential to relieve the pressure of frontline radiologists, improve early diagnosis, isolation and treatment, and thus contribute to the control of the epidemic.

Cardiovascular diseases remain the leading cause of global mortality, responsible for an estimated 17.9 million deaths annually. Early and interpretable diagnosis is critical to improving patient outcomes. This study bridges the gap between black-box machine learning models and clinical interpretability by developing a hybrid expert system that derives transparent, evidence-based diagnostic rules directly from patient data. We trained a constrained Decision Tree model (maximum depth = 3) on a publicly available dataset of 918 patients to ensure inherent explainability. The tree structure was systematically parsed and translated into a set of eight mutually exclusive IF-THEN clinicalrules, forming the knowledge base of a prototype rule-based expert system. Each rule is accompanied by its confidence score, support, and a clinically meaningful interpretation. The system achieved an accuracy of 82.1% (95% CI: 76.5 – 87.6%), a precision of 87.1%, and a recall of 79.4% on an independent test set. Five-fold cross-validation confirmed robustness (mean accuracy: 81.5% ± 6.1%), and the area under the ROC curve (AUC) was 0.886, indicating excellent discriminative ability. Crucially, for each case, the system outputs the specific diagnostic rule applied, a risk-stratified recommendation (e.g., urgent cardiology consultation, routine screening), and a confidence score. and a clinical interpretation, moving beyond a mere prediction to an auditable diagnostic aid. This transforms the model from a predictive endpoint into a fully auditable and interpretable clinical decision support tool. This work provides a reproducible and scalable framework for building accurate, transparent, and clinically deployable diagnostic systems, facilitating greater trust and adoption in real-world healthcare settings.

Bone fracture detection is a critical task in orthopedic radiology, where timely and accurate diagnosis significantly impacts patient recovery. Manual interpretation of X-ray images can be challenging due to the subtle nature of certain fractures and the high workload of clinicians. This study explores the efficacy of Deep Learning models in automating the detection of bone fractures. We implemented and compared two prominent architectures: ResNet50 and a proposed optimized MobileNetV2. To address the challenges of class imbalance and limited clinical data, we utilized advanced Data Augmentation techniques and Dropout regularization. The models were evaluated on a dataset of 425 unseen test images. The results demonstrate that the proposed MobileNetV2 model significantly outperformed the baseline, achieving a remarkable 99% accuracy, a 1.00 Recall for fracture detection, and a Macro F1-score of 0.95. In contrast, while ResNet50 achieved 94% accuracy, it exhibited severe bias and failed to generalize to normal cases. Furthermore, we employed Grad-CAM (Gradient-weighted Class Activation Mapping) to provide visual interpretability, confirming that the model’s "attention" aligns with actual anatomical fracture sites. These findings suggest that the lightweight MobileNetV2 architecture is not only highly accurate but also suitable for integration into portable medical devices and real-time clinical decision support systems, offering a reliable tool for enhancing diagnostic precision in emergency settings.

Background: Coronavirus is a family of viruses that can cause disease such as severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS). In 2019, Coronavirus was identified as the cause of the outbreak in China. Therefore, in this paper we will review what the Coronavirus is and explore the symptoms associated with it. Objectives: The main objective of this expert system is to obtain appropriate diagnosis of the disease. Methods: In this paper, the expert system is designed for the ability of doctors to detect and diagnose symptoms of coronavirus, a common symptom of this disease are fever, cough, pneumonia in both lungs and shortness of breath. This system presents the disease symptoms, which day will be recognizing symptoms, survival and spread, favorable conditions and image of this symptoms. Clips and Delphi expert system languages are used for designing and implementing the proposed expert system. Result: The expert system in the diagnosis of Coronavirus disease was assessed by doctors and they were satisfied and accepted with its quality of performance. Conclusions: The expert system is easy for doctors and people have interested in the coronavirus to detect and diagnosis the symptoms that may face this disease.

Background: Safflower is a highly branched, herbaceous, thistle-like annual plant. Any section of the safflower plant sections can suffering from a disease that weakens the ability to grow and eliminates its production. Therefore, in this paper will identify the pests and diseases present in safflower culture and detect the symptoms in each disease. Also images are showing the symptom form in this disease. Objectives: The main objective of this expert system is to obtain appropriate diagnosis of the disease. Methods: In this paper, the expert system is designed for the ability of agricultural engineers to detect and diagnose disease of safflower like as: alternaria blight, cercospora leaf spot, powdery mildew, head rot and wilt, mosaic, ramularia leaf spot, rust, wilt, and root rot. This system displays the disease symptoms, survival and spread, favorable conditions and image for each disease. Clips and Delphi expert system languages are used for designing and implementing the proposed expert system. Results: The expert system in the diagnosis of safflower diseases was evaluated by farmers and agricultural engineers and they were satisfied and accepted with its quality of performance. Conclusions: The expert system is easy for farmers and people have experience in the plant of safflower to detect and diagnosis the symptoms that may face this plant from several disease.

Background: The castor bean is a large grassy or semi-wooden shrub or small tree. Any part of the castor plant parts can suffering from a disease that weakens the ability to grow and eliminates its production. Therefore, in this paper will identify the pests and diseases present in castor culture and detect the symptoms in each disease. Also images is showing the symptom form in this disease. Objectives: The main objective of this expert system is to obtain appropriate diagnosis of the disease. Methods: In this paper, the expert system is designed for the ability of agricultural engineers to detect and diagnose disease of castor like as: seeding blight, alternaria blight, cercospora leaf spot, powdery mildew and wilt. This system presents the disease symptoms, survival and spread, favorable conditions and image for each disease. Clips and Delphi expert system languages are used for designing and implementing the proposed expert system. Results: The expert system in the diagnosis of castor diseases was assessed by farmers and agricultural engineers and they were satisfied and accepted with its quality of performance. Conclusions: The expert system is easy for farmers and people have experience in the plant of castor to detect and diagnosis the symptoms that may face this plant from several disease.

Background: Thyroid disease is a group of disorders that affects the thyroid gland. The thyroid is a small, butterflyshaped gland in the front of your neck that makes thyroid hormones. Thyroid hormones control how your body uses energy, so they affect the way nearly every organ in your body works—even the way your heart beats. Therefore, in this paper will identify what is the thyroid and diseases present in thyroid and detect the symptoms in each disease. Objectives: The main objective of this expert system is to obtain appropriate diagnosis of this disease. Methods: In this paper, the expert system is designed for the ability of doctors to detect and diagnose disease of thyroid. The proposed expert system presents an overview about thyroid diseases are given, the cause of diseases are outlined and the treatment of disease whenever possible is given out. SL5 Object Expert System language was used for designing and implementing the proposed expert system. Results: The expert system in the diagnosis of thyroid diseases was assessed by Medical student and doctors they were satisfied and accepted with its quality of performance. Conclusions: The expert system is easy for podiatric physician, patients and people have experience to detect and diagnosis the symptoms that may face this disease.

Abstract: The integration of artificial intelligence (AI) into mental health care has the potential to revolutionize the field by enhancing diagnostic accuracy, personalizing treatment, and improving access to care. This paper explores the advancements in AI technologies applied to mental health, including machine learning algorithms for diagnosis, natural language processing for therapeutic applications, and predictive analytics for personalized care. It also examines the ethical and practical challenges associated with these technologies, such as privacy concerns, algorithmic bias, and the need for regulatory frameworks. By analyzing current applications and emerging trends, this paper aims to provide a comprehensive overview of how AI can transform mental health care while addressing the associated challenges and opportunities.

Early detection of breast cancer plays a pivotal role in improving patient prognosis and reducing mortality rates. Recent technological advancements have significantly enhanced the accuracy and effectiveness of breast cancer screening methods. This paper explores the latest innovations in early detection, including the evolution of digital mammography, the impact of 3D mammography (tomosynthesis), and the use of advanced imaging techniques such as molecular imaging and MRI. Furthermore, the integration of artificial intelligence (AI) in diagnostic tools is discussed, highlighting how machine learning algorithms are refining imaging analyses and reducing diagnostic errors. Advances in genetic screening, liquid biopsies, and biomarkers are also examined, showcasing their potential in identifying high-risk individuals and enabling personalized treatment plans. This paper provides insights into the future of breast cancer detection, outlining both the opportunities and challenges that lie ahead in adopting these innovative technologies to improve patient outcomes.

Artificial intelligence (AI), deep learning, machine learning and neural networks represent extremely exciting and powerful machine learning-based techniques used to solve many real-world problems. Artificial intelligence is the branch of computer sciences that emphasizes the development of intelligent machines, thinking and working like humans. For example, recognition, problem-solving, learning, visual perception, decision-making and planning. Deep learning is a subset of machine learning in artificial intelligence that has networks capable of learning unsupervised from data that is unstructured or unlabeled. Deep learning is a technique used to generate face detection and recognize it for real or fake by using profile images and determine the differences between them. In this study, we used deep learning techniques to generate models for Real and Fake face detection. The goal is determining a suitable way to detect real and fake faces. The model was designed and implemented, including both Dataset of images: Real and Fake faces detection through the use of Deep learning algorithms based on neural networks. We have trained dataset which consists of 9,000 images for total in 150 epochs, and got the ResNet50 model to be the best model of network architectures used with 100% training accuracy, 99.18% validation accuracy, training loss 0.0003, validation loss 0.0265, and testing accuracy 99%.