Dynamic graph convolution with comprehensive pruning and GNN classification for precise lymph node metastasis detection Chaitra H. N., Shwetha N., Adarsh Rag S., Chandra Singh, Rangaswamy Y. Scientific Reports, 2026 Early and accurate detection of lymph node metastases is crucial for improving breast cancer patient outcomes. However, current clinical practices, including CT, PET imaging, and microscopic examination, are time-consuming and prone to errors due to low tissue contrast, varying lymph node sizes, and complex workflows. To address the limitations of existing approaches in lymph node segmentation, feature embedding, and classification, this study proposes a novel framework Graph-Pruned Lymph Node Detection Framework (GPLN-DF) that integrates a Dynamic Graph Convolution (DGC) autoencoder with Node Attribute-wise Attention (NodeAttri-Attention) for accurate lymph node segmentation. This segmentation is further refined using Comprehensive Graph Gradual Pruning (CGP) to reduce unnecessary parameters and computational costs. After segmentation, Hessian-based Locally Linear Embedding (HLLE) is applied for effective feature extraction and dimensionality reduction, preserving the geometric structure of lymph node regions. Finally, a Graph Neural Network (GNN) classifier enhanced with CGP is used to classify the segmented lymph nodes as metastatic or non-metastatic based on the extracted features. This comprehensive framework addresses challenges such as small lymph node size, shape variability, low contrast in medical imaging, and high computational burden. The model was evaluated on the CAMELYON17 dataset, achieving a classification accuracy of 98.65%, surpassing existing models in segmentation precision and classification performance.
Meta-RL Based Micro-Expression Recognition Framework Using MAML with Calibrated Regression Function N Shwetha, Aravind Jadhav, Chandra Singh, Virupaxi B.Dalal, N Sangeetha, Bhaskar Awadhiya, Yashwanth Nanjappa, Y Rangaswamy International Journal of Computational Intelligence Systems, 2026 Micro-Expression Recognition (MER) plays a crucial role in understanding human emotions, yet its effectiveness is often hindered by the transient and subtle nature of micro-expressions. This article presents a novel MER framework integrating Calibrated Regression with Maximum Mean Discrepancy (MMD), Model-Agnostic Meta-Learning (MAML), and Meta-Reinforcement Learning (Meta-RL) to enhance recognition accuracy and adaptability. The 2D Convolutional Neural Network (2DCNN) is employed as the backbone for feature extraction, capturing fine-grained spatial details of Facial Expressions (FEs). To address challenges in feature alignment, a Heteroscedastic Neural Network (HNN) is introduced for predictive uncertainty estimation. A two-stage learning process is applied, where Negative Log-Likelihood (NLL) optimization refines model parameters, and MMD ensures better alignment between micro- and macro-expressions. Additionally, the Meta-RL framework optimizes feature learning through characterizing the optimal gap of the stationary points achieved using MAML and improving generalization. Extensive experiments on benchmark datasets like SAMM and CASME II shows the advantage of the introduced approach, achieving 96.74% accuracy on SAMM and 98.84% accuracy on CASME II, surpassing state-of-the-art models. The results highlight the model’s robustness, adaptability, and effectiveness, making it well-suited for real-world micro-expression analysis applications.
AI-Driven Waste Segregation and IoT-Enabled Monitoring: A Smart Dustbin for Urban Waste Management Roopesh Ramesh, Nelamangala Nagaraju Shwetha, Mayur Rajendra Badiger, Shreya Santhosh, Tejasvi Bellubbi, Aishwarya Bagi Proceedings of the 2025 International Conference on Artificial Intelligence and Emerging Technology Global AI Summit 2025, 2025 This paper proposes an intelligent AIoT-enabled waste management system that autonomously classifies and segregates waste at the point of disposal while simultaneously monitoring real-time bin fill levels. The system combines edge-deployable Convolutional Neural Networks (CNNs), capacitive moisture sensing, and ultrasonic distance measurement within a modular embedded framework comprising Raspberry Pi and ESP32 microcontrollers. Waste classification is achieved through a lightweight CNN model optimized for resource-constrained environments, enabling near real-time inference with sub-second latency. The classified waste is physically sorted into biodegradable, recyclable, and non-recyclable compartments using a servo-actuated mechanism, thereby minimizing human intervention. IoT connectivity facilitates continuous bin status updates to a cloud-based dashboard (ThingSpeak), with automated alerts upon threshold breaches or system anomalies. Experimental validation demonstrates a classification accuracy exceeding 90%, a sorting success rate of 96%, and fill-level sensing precision within ±2 cm. This scalable, low-cost solution holds significant promise for deployment in smart cities, public infrastructure, and urban sustainability frameworks.
AeroGrain:Advanced Hot-Air Preservation with Real-Time Moisture Sensing N Shwetha, Virupaxi Dalal, R Shreyas, B U Thanuj Gowda, Suprith S Raykar, Sharanabasappa 4th IEEE North Karnataka Subsection Flagship International Conference Holistic Engineering for Sustainable Development Nkcon 2025, 2025
Deep Learning Integration for Indoor Navigation and Aerial Tracking Via Siamese Architectures and Multi-Scale Feature Fusion Nelamangala Nagaraju Shwetha, Virupaxi Dalal, Chandra Singh, Aravind Jadhav, Shobha V Patil, Sowmya Gadag 2025 IEEE International Conference on Intelligent Signal Processing and Effective Communication Technologies Inspect 2025, 2025 This work presents an integrated deep learning framework for autonomous indoor navigation and aerial target tracking. The system employs a dual-domain approach combining semantic scene understanding and real-time object localization. For ground navigation, a Siamese Deep Convolutional Neural Network (SiCNN) with a shared backbone is used, branching into semantic segmentation and scene classification. The segmentation pipeline uses Atrous Convolutions and a lightweight ASPP module for high-resolution path planning, while the classification branch predicts control policies based on human-inspired strategies. For aerial tracking, a YOLOv5-based model is enhanced with a Convolutional Attention Mechanism and Multi-Scale Feature Fusion to improve small object detection and background suppression. A novel loss function, Effective Intersection over Union (EIoU), improves bounding box accuracy and training stability. The system is trained using imitation learning, supervised fine-tuning, and domain-adaptive transfer learning. Evaluation on standard datasets shows strong performance in accuracy, segmentation IoU, and real-time tracking, demonstrating robustness and scalability for future robotic systems.
Adaptive channel equalisation using different hybrid metaheuristic algorithms in digital communication N. Shwetha, Manoj Priyatham, N. Gangadhar International Journal of Autonomous and Adaptive Communications Systems, 2024 An adaptive channel equalisation concept is used to reduce the effects of inter-symbol interference (ISI) in digital communication. The equalisation process is considered an optimisation issue to minimise the mean square error (MSE) between the transmitted signal and the output of the equaliser. Therefore, metaheuristic algorithms are widely adopted to enhance the function of adaptive channel equalisers. In this paper, a bio-inspired emperor penguin optimisation (EPO) algorithm is hybridised with five different algorithms to optimise the finite impulse response (FIR) channel for reducing the effects of ISI. The main role of these algorithms is to optimise the weights or coefficients of the equaliser to reduce the effect of ISI. Finally, the performance of each algorithm in channel equalisation is assessed, and it is observed that EPO incorporated with both manta ray foraging and tunicate swarm algorithm has obtained relatively better equalisation results than other hybrid optimisation algorithms.
Hybridization of Artificial Neural Network with Spotted Hyena Optimization (SHO) Algorithm for Heart Disease Detection N. Shwetha, N. Gangadhar, Mahesh B. Neelagar, N. Sangeetha, Virupaxi Dalal Modeling and Optimization of Signals Using Machine Learning Techniques, 2024 Heart-related illnesses are the leading cause of mortality globally, which causes a high number of deaths in poor- and middle-income nations like India. Large amounts of data are constantly being generated by medical professionals. The generated data can be used to diagnose heart disease in advance, which can efficiently diminish the incidence of various heart-related diseases. Predictions can be done effectively by improving the knowledge identification needed to detect previously unknown patterns. Effective predictions can be made and hidden patterns can be detected by accessing data and concerns collected from healthcare industries. In this work, machine learning technique is used on cardiac disease-related data to try to find out the potential for heart disease before suffering from serious problems. Therefore, an artificial neural network (ANN) is used to predict a coronary illness. Additionally, the spotted hyena optimization (SHO) algorithm is hybridized with ANN to update the weights in an ANN. The implementation is carried out on the MATLAB platform. The proposed method's effectiveness is verified by different patients considering 13 constraints as the dataset. These constraints are evaluated for training and testing each data in the dataset. The efficiency of the proposed approach is shown in comparison with different methods, namely, social learning algorithm (SLO), particle swarm optimization (PSO), gaussian discriminant analysis (GDA), and genetic algorithm (GA).
Wireless Detection Systems Using Matrix-Oriented Diffusion Akshatha Bhat, Lavanya M S, Niranjan L, Rakheeba Taseen, Haseeba Yaseen, N Shwetha International Conference on Smart Systems for Applications in Electrical Sciences Icsses 2023, 2023
