Background and Objectives: Diffusion models, as a recent advancement in generative modeling, have become central to high-resolution image synthesis and reconstruction. Their rapid progress has notably shaped computer vision and health informatics, particularly by enhancing medical imaging and diagnostic workflows. However, despite these developments, researchers continue to face challenges due to the absence of a structured and comprehensive discussion on the use of diffusion models within clinical imaging. Methods: This systematic review investigates the application of diffusion models in medical imaging for diagnostic purposes. It provides an integrated overview of their underlying principles, major application areas, and existing research limitations. The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines and included peer-reviewed studies published between 2013 and 2024. Studies were eligible if they employed diffusion models for diagnostic tasks in medical imaging; non-medical studies and those not involving diffusion-based methods were excluded. Searches were conducted across major scientific databases prior to the review. Risk of bias was assessed based on methodological rigor and reporting quality. Given the heterogeneity of study designs, a narrative synthesis approach was used. Results: A total of 68 studies met the inclusion criteria, spanning multiple imaging modalities and falling into eight major application categories: anomaly detection, classification, denoising, generation, reconstruction, segmentation, super-resolution, and image-to-image translation. Explainable AI components were present in 22.06% of the studies, clinician engagement in 57.35%, and real-time implementation in 10.30%. Overall, the findings highlight the strong diagnostic potential of diffusion models but also emphasize the variability in reporting standards, methodological inconsistencies, and the limited validation in real-world clinical settings. Conclusions: Diffusion models offer significant promise for diagnostic imaging, yet their reliable clinical deployment requires advances in explainability, clinician integration, and real-time performance. This review identifies twelve key research directions that can guide future developments and support the translation of diffusion-based approaches into routine medical practice.

Bengali News Headline Categorization Using Machine Learning aims to categorize Bengali online news headlines into six distinct categories using Natural Language Processing. Researchers in different application fields have recently paid great attention to the fantastic accomplishments of Machine Learning Models in Natural Language Processing. Several machine learning algorithms categorize Bengali news headlines, including Logistic Regression, Random Forest Classifier, Multinomial Naive Bayes, and RBF Support Vector Machine. Also, deep learning models like LSTM, Bi-LSTM, GRU, Bi-GRU, and CNN, and the Bangla-BERT and XLM-RoBERTa transformer learning models are presented in this research. This paper’s primary purpose is to provide a comparative observation of several machine learning models, deep learning models, and transformer learning methods in Bengali news headline classification. We used 1,36,811 text data of Bengali news headlines for evaluation, and our dataset had an accuracy of 86.50% with XLM-RoBERTa.

Knee osteoarthritis (KOA) is a widespread degenerative joint disease that poses significant global challenges owing to delayed diagnosis and treatment, often resulting in severe disability. Despite extensive research on predicting KOA, many proposed methods lack reliability, because they fail to incorporate explainable AI (XAI) methodologies, robust preprocessing techniques, and appropriate hyperparameter tuning. This study introduces a deep learning framework for KOA classification, addressing both binary (diagnosis) and multi-class (severity prediction) classification tasks using the Osteoarthritis Initiative (OAI) dataset. Our approach is enhanced by a comprehensive image prepocessing pipeline that includes scaling, sharpening, denoising, histogram equalization, and contrast enhancement, which standardizes image quality and highlights crucial features for classification. The proposed stacked ensemble model, which integrates Xception, EfficientNetB5, and InceptionV3, surpasses individual models, achieving 86.29% accuracy in KOA diagnosis and 96.93% in KOA severity prediction. To ensure transparency and interpretability, we incorporated advanced explainability tools, including Gradient-weighted Class Activation Mapping (Grad-CAM), Faster Score-CAM, and Local Interpretable Model-agnostic Explanations (LIME), providing clear visual insights into the model’s decision-making process. Our findings present a balanced approach that combines performance with transparency, potentially leading to earlier and more accurate KOA diagnoses.

One of the most serious and prevalent diseases in the world is skin disease. Due to the intricacy of humans' tone and texture and the visible proximity impact of the disorders, it can be quite difficult to pinpoint the exact type of condition at times. Because of this, it's crucial to recognize skin illnesses at the earliest indication of them. Therefore, it is essential to have a system that can identify skin conditions regardless of these restrictions. With their major advancements, artificial intelligence, computer vision, and machine learning are now crucial to the area of medical research, particularly when it comes to the analysis of medical pictures. In this research, we introduced an automated system based on machine learning for classifying three distinct forms of skin illnesses, including melanoma, basal cell carcinoma, and eczema. For this, we have used different machine learning algorithms like Support Vector Machine, Naive Bayes, Logistic Regression and deep learning models like CNN, LSTM, Bi-LSTM, Inception V3, VGG-16, and Xception. We achieved impressive accuracy of above 90 % with every model on a bespoke dataset containing 9665 photos, and the maximum accuracy was 99.33% with the pre-trained deep learning model Xception.