International Journal of Advances in Applied Computational Intelligence

Rapid growth in the adoption of Electronic Health Records (EHRs), Internet of Medical Things (IoMT) devices, wearable sensor technology, and digital healthcare systems offers immense scope for intelligent healthcare decision support. However, most AI-enabled healthcare systems in use today still lack explainability, contextual reasoning capabilities, and effective decision-making. For these reasons, this research develops an Agentic Generative AI Framework for Intelligent Disease Prediction and Decision-Making in smart healthcare. The framework incorporates predictive analytics, Generative AI-based clinical reasoning, and autonomous intelligent agents into a coherent healthcare framework. Six specific agents are used for data gathering, data analysis, disease prediction, clinical reasoning, treatment recommendations, and patient monitoring. The combined functionality of these agents supports disease prediction, clinical reasoning, and personalized treatment plans. Evaluation was performed on healthcare datasets related to heart disease, diabetes, chronic kidney disease, and breast cancer. Experimental results show high efficiency, stable accuracy across diseases, reliable recommendation generation, and enhanced healthcare intelligence compared with traditional ML, DL, and LLM methods. Results show that combining Agentic AI with Generative AI increases explainability, adaptability, and efficiency in medical decision support. The proposed model represents an encouraging path toward intelligent, patient-centered, and explainable smart healthcare systems.

Due to the increasing adoption of IoT applications, there is a growing necessity for energy-efficient and sustainable WSN. Yet, traditional routing protocols tend to face problems like energy wastage, congestion, unreliable communication, and shorter network life spans under dynamic network conditions. This study presents the development of a DRL-powered Green IoT framework to enhance efficient communication through WSN while optimizing QoS performance. Specifically, the proposed framework employs the Deep Q-Network, Double Deep Q-Learning, adaptive clustering, and multi-objective optimization in order to enhance both routing and QoS performance. The model makes use of residual energy, congestion levels, throughput, delivery rate, and communication delays during its decision-making processes. Experimentation with the model was performed by making use of Python and NS-3. The simulation results showed that the presented model performed better than traditional routing methods like LEACH, PEGASIS, and HEED when evaluated on factors like energy preservation, enhanced throughput, minimized congestion, reduced delays, and increased network life spans. It can be concluded that DRL-powered communication optimization is a viable solution for the future development of Green IoT communication systems.

Accurate and early identification of plant diseases is essential for ensuring sustainable agriculture and maximizing crop productivity. This paper presents a hybrid deep learning framework integrating Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks for multi-class plant disease detection, classification, and Plant Health Value (PHV) estimation. The proposed framework begins with a comprehensive data preprocessing pipeline involving image resizing, normalization, and augmentation to improve model robustness. The CNN module extracts critical spatial and visual features such as lesion shape, leaf texture, and color intensity, while the BiLSTM model captures temporal and sequential feature correlations to accurately learn disease progression patterns. A Decision Support System (DSS) is incorporated to compute the Plant Health Value (PHV), where PHV ranges from 0% (Healthy) to 100% (Severely Unhealthy), indicating the severity of disease infection. Additionally, the DSS generates actionable recommendations to assist in early intervention and treatment planning. Experimental results on a multi-species plant dataset demonstrate that the proposed CNN–BiLSTM hybrid model significantly improves accuracy, interpretability, and early disease prediction compared to conventional CNN based methods, offering a robust and intelligent framework for automated plant health monitoring.

The advancement of trustworthy diagnostic tools in dermatological automation is hindered by the limited transparency of current deep learning systems, which function as opaque models and impede clinical acceptance. This research presents a novel intelligent framework for skin lesion analysis that unites deep learning methodologies with explainable artificial intelligence (XAI) principles to address this interpretability deficit. The proposed approach utilizes a transfer-learned ResNet50 architecture for robust image classification, coupled with Local Interpretable Model-agnostic Explanations (LIME) to furnish clear, visual justifications for the model’s outputs. Performance assessment on the HAM10000 benchmark yielded a classification accuracy of 94.3%, with a validation accuracy of 91.8%. Concurrently, the LIME framework effectively identified and visualized diagnostically critical features in the lesion images, thereby elucidating the model’s reasoning process for medical practitioners. These findings confirm that augmenting high-performance deep learning with post-hoc explanatory techniques yields a credible and understandable diagnostic instrument, thereby promoting clinician trust and facilitating data-informed medical judgments. Subsequent developments will prioritize scalable cloud implementation, interoperability with healthcare information systems, extension to underrepresented lesion categories, and rigorous evaluation in diverse clinical environments.

Multiple moving object detection and tracking are challenging roles in many computer vision applications such as object navigation and human identification. Object tracking is one of the key challenges for securing against crime, supporting public safety, and enabling effective traffic management systems. In video surveillance applications, detection of multiple moving vehicles from video is the major task for tracking and understanding the behavior of the detected objects. Performance of object detection algorithms is degraded by factors such as fog or haze, occlusion, dynamic background, poor illumination, and low resolution. Fog is one of the major bottlenecks of video surveillance applications. The proposed Dark Channel Prior algorithm using guided filter (GDCP) is adapted for fog removal. The Gaussian Mixture Model (GMM) is proposed for detecting multiple moving objects, and features are extracted from the detected objects using Motion Vector Estimation. The K-Nearest Neighbor algorithm is used for tracking the moving objects (vehicles) using the detected features. Efficiency is improved due to the adoption of the proposed fog-removal algorithm and feature extraction for effective tracking. There are wide varieties of applications in moving object detection and tracking.