This survey explores the application of multimodal machine learning techniques in predictive healthcare analytics. By integrating various data modalities, such as medical imaging, clinical text, time-series signals, and structured tabular data, these approaches aim to emulate clinical reasoning and enhance diagnostic and prognostic accuracy. Across studies, multimodal ML consistently outperformed unimodal baselines, with intermediate fusion employed in 60% of cases and achieving average AUC improvements of 5–12% over single-modality models. Oncology and neurology emerged as the leading domains, where combining imaging with genomic and cognitive data significantly improved cancer survival and Alzheimer’s detection. Despite progress, key challenges persist, including modality misalignment (23%), missing data (18%), and limited external validation (12%). Recent trends highlight transformer-based cross-modal attention, self-supervised learning for data-scarce settings, and hybrid fusion architectures in critical care. While multimodal ML offers clear clinical advantages, regulatory constraints and interoperability gaps continue to hinder deployment. This survey contributes to the existing literature by providing a comprehensive synthesis of multimodal ML applications across healthcare domains. It documents comparative fusion strategies, modelling approaches, and empirical performance outcomes. The paper’s primary contribution lies in identifying intermediate fusion as the most effective integration strategy and revealing systematic gaps in external validation and model transparency that must be addressed for clinically trustworthy multimodal systems.

Sindhi, a low-resource language spoken by millions, faces significant challenges in Natural Language Processing (NLP) due to the scarcity of annotated datasets. This paper presents DAugSindhi, a study focused on enhancing Sindhi text classification through data augmentation techniques. These methods aim to address data scarcity by artificially expanding the dataset to improve model performance. The study explores various augmentation techniques, including Easy Data Augmentation (EDA), Back Translation, Paraphrasing, and Text Generation using Large Language Models (LLMs). EDA methods like synonym replacement, random insertion, random deletion, and random swapping introduce semantic diversity, while Back Translation creates contextual variations. Paraphrasing and Text Generation leverage LLMs to produce enriched and diverse text samples. Experiments were conducted on a Sindhi dataset of 3364 news articles categorized into sports, entertainment, and technology. A multilingual BERT model (bert-base-multilingual-cased) was fine-tuned for binary and multi-class classification tasks. Results revealed that EDA, particularly Random Deletion, outperformed all other techniques, achieving a 99% F1 score for binary classification. Back Translation and Paraphrasing also delivered substantial improvements, highlighting their utility in low-resource settings. This work establishes a robust baseline for Sindhi text classification, demonstrating that simple augmentation techniques can significantly enhance NLP applications for underrepresented languages. Future work will explore hybrid strategies and larger datasets to further advance Sindhi NLP.

Obtaining information about the electrical activity of the heart in the form of electrocardiograms (ECG) has become a standard way of monitoring patients’ heart rhythm and function. It is used for diagnosing a variety of cardiac anomalies such as arrhythmia and other heart diseases. However, the interpretation of ECGs requires the expertise of trained physicians, thus bearing the need for tools that automatically classify such signals. In this study we train deep convolutional neural networks (CNNs) to perform binary classification of ECG beats to normal and abnormal. We use transfer learning in order to build models that are fine-tuned on specific patients’ data, after pre-training a generic network on a set of different ECGs selected from the MIT-BIH arrhythmia database. We then compare the performance of the fine-tuned networks against that of individual networks, which are trained only on the ECG data of a single patient, in order to evaluate the overall efficacy of transfer learning on the given problem. We managed to achieve adequate results on both scenarios as the individual classifiers yielded an average of 94.6% balanced accuracy on the test set, whereas the fine-tuned models a marginally worse 93.5%.

The rapid advancement of large language models (LLMs), such as ChatGPT, has revolutionized natural language processing, enabling these models to generate coherent and contextually relevant text. These capabilities hold significant potential for various applications, particularly in the medical field. LLMs can assist healthcare professionals by providing accurate information, supporting diagnostic processes, and enhancing medical education. However, the implementation of LLMs in clinical settings faces several barriers, including technical limitations, ethical concerns, regulatory constraints, and practical challenges. Ensuring accuracy, reliability, and interpretability is crucial, along with addressing biases, ensuring data privacy, and establishing ethical guidelines. Future research should focus on improving training techniques, developing explainable AI methods, and creating efficient, resource-saving models. Expanding applications through domain-specific models and multi-modal integration can further unlock the potential of LLMs. By addressing these challenges and prioritizing responsible development, LLM technology can be safely and effectively integrated into various domains, driving innovation and improving outcomes. This paper explores the current state of LLM technology, its medical applications, barriers to implementation, and directions for future research and development

Skin cancer, particularly melanoma, poses a significant global health challenge due to its prevalence and mortality rate. Early detection is critical to improving outcomes, as advanced cases become increasingly difficult to treat. The advent of Artificial Intelligence (AI) and Explainable AI (XAI) techniques has revolutionized dermatological diagnostics by offering accurate and interpretable solutions. This systematic review investigates the integration of XAI in skin lesion classification, analyzing 22 recent studies published between 2019 and 2023. The studies encompass diverse approaches, including deep learning models like CNNs, ResNet, DenseNet, and MobileNet, as well as explainability techniques such as Grad-CAM, SHAP, and saliency maps. Results highlight significant advancements in accuracy and interpretability, with some models achieving over 99% accuracy on datasets like ISIC 2018 and HAM10000. However, challenges persist, including dataset imbalances, limited diversity in patient metadata, and generalizability across different skin types and imaging conditions. XAI methods, by visualizing model decision pathways, enhance transparency, fostering trust among clinicians and enabling seamless AI integration into clinical practice. This review underscores the potential of combining state-of-the-art AI models with explainable frameworks to address the complexities of skin lesion diagnostics. It advocates for future research to prioritize diverse, metadata-rich datasets, innovative optimization techniques, and robust architectures to develop reliable, interpretable diagnostic tools. By bridging the gap between advanced AI and user understanding, this work contributes to the creation of clinically applicable, trustable AI-driven healthcare solutions.