N Arul

@ajiet.edu.in

Assistant Professor, Computer Science and Engineering
AJ Institute of Engineering and Technology

N Arul


               Download Compact PDF  
https://researchid.co/arul_hi2000

EDUCATION

Master of Engineering - Computer Science and Engineering

RESEARCH, TEACHING, or OTHER INTERESTS

Engineering
6

Scopus Publications

Scopus Publications

  • An explainable artificial intelligence framework for ischemic heart disease prediction using enhanced squirrel search feature selection
    D. Cenitta, N. Arul, R. Vijaya Arjunan, Krishnaraj Chadaga, J. Andrew
    Scientific Reports, 2026
    Clinical decision-making depends on the ability to anticipate ischemic heart disease (IHD) accurately and interpretably. Despite their excellent accuracy, machine learning models are not often accepted in medical applications due to their black-box nature. To diagnose IHD, this study suggests an Explainable Artificial Intelligence (XAI) architecture that combines explainable models with the Enhanced Squirrel Search Optimization (ESSO) method for feature selection. The proposed Enhanced Squirrel Search Optimization (ESSO) introduces adaptive exploration mechanisms to efficiently identify an optimal subset of clinically relevant features. By integrating ESSO-based feature selection with Random Forest classification and explainable AI techniques (SHAP and LIME), the framework simultaneously improves predictive performance and provides interpretable insights for clinical decision support. We use SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) methodologies to offer clear explanations into the model’s predictions, building on earlier research that showed how effective squirrel search is at selecting the best clinical features. With KNN-based imputation, normalization, categorical encoding, outlier handling, and SMOTE-based class balancing for managing missing data, the UCI heart disease dataset is utilized for both training and validation. Random Forest is used to classify the chosen features, and domain experts assess the derived explanations for clinical relevance. With precise visual and numeric explanations of contributing features including cp., oldpeak, thal, and ca., the suggested model achieves an accuracy of 98.4%. By bridging the gap between interpretability and high-performance machine learning, our research makes the model appropriate for practical use in clinical settings.
  • A bio-inspired approach to feature optimization for ischemic heart disease detection
    D. Cenitta, N. Arul, T. Praveen Pai, R. Vijaya Arjunan, Tanuja Shailesh
    Healthcare Analytics, 2025
    Ischemic Heart Disease (IHD) stands as one of the primary contributors to worldwide deaths, therefore requiring precise and efficient predictive models. Standard machine learning techniques encounter hurdles, including excessive feature dimensions and unbalanced data distribution together with inappropriate feature group choice that negatively affect model effectiveness. The research introduces an optimized feature selection method by employing an Improved Squirrel Search Algorithm (ISSA) to raise the predictive capacity for IHD classification. The ISSA implements adaptive search features to automatically optimize feature selection, through which it maintains important attributes while eliminating redundant information. The selected features are evaluated using a Random Forest classifier, known for its robustness and interpretability in medical prediction tasks. Experimental results on the University of California Irvine (UCI) Heart Disease dataset show that the Improved Squirrel Search Algorithm–Random Forest (ISSA-RF) model achieves a classification accuracy of 98.12%, outperforming existing feature selection techniques while reducing computational overhead. Bio-inspired optimization proves effective in medical diagnostics through recent research findings that lead to more efficient predictive healthcare models with interpretable properties. • Introduce a bio-inspired optimization method for improved feature selection in heart disease prediction. • Apply adaptive learning to eliminate redundant data and enhance diagnostic accuracy. • Integrate intelligent search with ensemble learning for efficient healthcare analytics. • Demonstrate improved model interpretability with reduced computational cost. • Validate a dynamic optimization strategy that enhances reliability in clinical decision support.
  • An Explainable and Optimal Deep Transfer Learning Approach for Forest Fire Detection Using Grad-CAM Visualizations
    Sameer Mansoori, Aishwarya Pitchumani, Nila Sangamitra Arul, S Vignesh, Arti Anuragi
    Ised 2025 13th International Conference on Intelligent Systems and Embedded Design Proceedings, 2025
    Forest fires pose severe environmental, economic, and human threats. Traditional detection methods such as satellite imagery, manual observation, and sensor-based technologies, suffer from high false alarm rates, slow response times, and high installation costs. Thus, this study proposes an Advanced Deep Learning-based Detection Framework that uses Transfer learning to enhance fire detection and response in real-time. The framework consists of several pretrained models-MobileNet, InceptionV3, ResNet50, VGG16, EfficientNet, and Xception-that are fine-tuned through hyperparameter tuning and LwF for maximized classification accuracy with minimized false positives. We used 3 datasets, each with 3 classes - Fire, Normal and Smoke. Performance evaluation was carried out using accuracy, precision, recall and confusion matrices. It was then established that MobileNet provided the highest classification accuracy of 98.2% with low computational cost while LwF greatly advanced generalization over complex datasets. Furthermore, to boost model interpretability and ensure trustworthy decisionmaking, Grad-CAM visualization was also put into play. The results suggest that integration of deep learning based wildfire detection with focus on explainability can result in significant improvement in early warning systems. Future research will focus on incorporating Vision Transformers (ViTs), multimodal sensor fusion, and edge computing architectures. This study indicates the advantages of pretrained models to enhance disaster preparedness by mitigating the impacts of climate change.
  • Explainable Transfer Learning with Residual Attention BiLSTM for Prognosis of Ischemic Heart Disease.
    Cenitta D, Arul N, Praveen Pai T, VIijaya Arjunan Ranganathan, Tanuja Shailesh, Andrew J
    F1000research, 2025
    Background Early and accurate prediction of Ischemic Heart Disease (IHD) is critical to reducing cardiovascular mortality through timely intervention. While deep learning (DL) models have shown promise in disease prediction, many lack interpretability, generalizability, and fairness—particularly when deployed across demographically diverse populations. These shortcomings limit clinical adoption and risk reinforcing healthcare disparities. Methods This study proposes a novel model: X-TLRABiLSTM (Explainable Transfer Learning–based Residual Attention Bidirectional LSTM). The architecture integrates transfer learning from pre-trained cardiovascular models into a BiLSTM framework with residual attention layers to improve temporal feature extraction and convergence. To ensure transparency, the model incorporates SHAP (SHapley Additive exPlanations) to quantify the contribution of each clinical feature to the final prediction. Additionally, a demographic reweighting strategy is applied to the training process to reduce bias across subgroups defined by age, gender, and ethnicity. The model was evaluated on the UCI Heart Disease dataset using 10-fold cross-validation. Results The X-TLRABiLSTM model achieved a classification accuracy of 98.2%, with an F1-score of 98.1% and an AUC of 99.1%, outperforming standard ML classifiers and state-of-the-art DL baselines. SHAP-based interpretability analysis highlighted clinically relevant predictors such as chest pain type, ST depression, and thalassemia. A fairness-aware reweighting strategy was applied during training, and fairness evaluation revealed minimal performance disparity across demographic subgroups, with F1-score gaps ≤ 0.6% and error rate gaps ≤ 0.4%. Confusion matrix analysis demonstrated low false-positive and false-negative rates, reinforcing the model’s reliability for clinical deployment. Conclusions X-TLRABiLSTM offers a highly accurate, interpretable, and demographically fair framework for IHD prognosis. By combining transfer learning, residual attention, explainable AI, and fairness-aware optimization, this model advances trustworthy AI in healthcare. Its successful performance on benchmark clinical data supports its potential for real-world integration in ethical, AI-assisted cardiovascular diagnostics.
  • Ischemic Heart Disease Prognosis: A Hybrid Residual Attention-Enhanced LSTM Model
    D. Cenitta, R. Vijaya Arjunan, Ganesh Paramasivam, N. Arul, Anisha Palkar, Krishnaraj Chadaga
    IEEE Access, 2025
    Well-timed prediction and an accurate diagnosis of Ischemic Heart disease (IHD) can reduce the risk of death, whereas an inaccurate diagnosis can prove fatal. So, there is a need to develop an optimal heart disease prediction model to avoid inaccurate ischemic heart disease diagnosis and further treatment. Recently, researchers have developed several deep learning techniques that take input from medical practitioners, automatically find hidden patterns in enormous volumes of data, and predict heart diseases without human intervention. Further, the deep learning model can help doctors to classify the severity of heart disease and choose appropriate treatment accordingly. These deep learning models can be improved to achieve greater accuracy and stability. Creating a hybrid model that combines attention residual learning with a Long Short-Term Memory (LSTM) is one method to prove it. Our suggested Hybrid Residual Attention-Enhanced LSTM (HRAE-LSTM) approach improves accuracy and stability by combining attention residual learning with an LSTM. For evaluating the effectiveness of the proposed HRAE-LSTM model, realistic datasets of 303 instances from the heart disease dataset (UCI), were used. The proposed HRAE-LSTM outperforms existing cardiac disease prediction systems by 97.7 % with the UCI dataset, respectively.
  • Explainable AI With Homomorphic Encryption for Secure Cloud-Based ECG Analysis in Heart Disease Diagnosis
    D. Cenitta, G. K. Shwetha, K. P. Vyshali Rao, Srividya Ramisetty, N. Arul, R. Vijaya Arjunan
    IEEE Access, 2025
    Electrocardiogram (ECG) analysis is widely used for early detection of cardiac abnormalities, yet the deployment of deep learning models in cloud environments raises concerns regarding data privacy and clinical interpretability. To address these challenges, this work presents a novel framework that integrates homomorphic encryption with explainable deep learning for secure and interpretable ECG classification in the cloud. This paper presents a novel framework that integrates homomorphic encryption with explainable deep learning for secure ECG-based heart disease diagnosis in the cloud. Explainable AI (XAI) was employed to enhance clinician and patient trust, while homomorphic encryption (HE) ensures confidentiality of sensitive ECG signals during cloud-based processing. The originality of this work lies in jointly addressing three critical requirements—data privacy, interpretability, and computational efficiency—within a single diagnostic pipeline. The proposed method employs a convolutional neural network (CNN) optimized for encrypted computation and applies SHapley Additive exPlanations (SHAP) to provide interpretable results aligned with clinical decision-making. Experimental validation on the MIT-BIH dataset demonstrates that the model achieves 94.2% classification accuracy, 92.0% F1-score, and 91% agreement with cardiologists, while maintaining an average encrypted inference latency of 420 ms, demonstrating its practicality for secure cloud deployment. These results confirm that the framework offers a practical and trustworthy solution for secure, cloud-based ECG diagnostics.