Aasha Nandhini S
Department of Electronics and Communication Engineering, Sri Sivasubramaniya Nadar College of Engineering, Kalavakkam, India

Published : 4 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 4 Documents
Search

An Explainable Uncertainty-Aware Deep Learning Framework for Robust Skin Disease Diagnosis R Karthick Manoj; Aasha Nandhini S; C Batumalai
Journal of Applied Data Sciences Vol 7, No 3: September 2026
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v7i3.1465

Abstract

Automated skin-disease diagnosis from clinical and dermoscopic images remains challenging because of illumination variability, acquisition artifacts, irregular lesion boundaries, limited feature diversity, and the low interpretability of conventional deep learning models. This study proposes EDLF-Skin, an integrated framework that combines Physio-Color Consistency preprocessing, Attentive Multi-scale U-Trans segmentation, Multi-Modal Feature Fusion, a Hybrid Ensemble with Uncertainty, and a Multi-level Explainability Interface. The preprocessing stage normalizes illumination and removes artifacts while preserving diagnostically relevant color and texture information. The segmentation module integrates convolutional feature extraction with transformer-based attention to improve lesion-boundary delineation. Spatial, frequency, and texture representations are subsequently fused through cross-attention and processed by complementary convolutional and transformer classifiers. Predictive uncertainty is estimated to identify low-confidence cases requiring expert review, while Grad-CAM, SHAP attribution, and textual rationales provide multilevel explanations. Experimental results show that the proposed segmentation model achieved a Dice coefficient of 0.956 and an intersection over union of 0.935. The complete classification framework obtained 96.87% accuracy, an F1-score of 0.953, an area under the receiver operating characteristic curve of 0.981, and an expected calibration error of 0.021. These findings indicate that EDLF-Skin improves segmentation accuracy, classification performance, probability calibration, and decision transparency within a unified diagnostic workflow.
Optimized AI-IoT Solution for Real-Time Pest Identification in Smart Agriculture Aasha Nandhini S; R. Karthick Manoj; M. Batumalay
Journal of Applied Data Sciences Vol 6, No 4: December 2025
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v6i4.810

Abstract

Pest detection and identification play a crucial role in reducing the damage caused by pest, insect and diseases.  Timely detection and response are essential to increase the quality and quantity of crop production. Efficient pest management strategies are important for achieving optimal crop quality and promoting sustainable agricultural practices. This research proposes a framework that can automatically detect pests and offer timely solutions to farmers. The proposed approach integrates intelligent computing methods with connected device networks to identify and classify pests in real time with high precision. The methodology focuses on efficiently segmenting the pest from the captured leaf image using a novel region growing based segmentation algorithm. The threshold for region growing based segmentation algorithm is based on the adaptive local region entropy which contributes to the efficient segmentation. Stacked Ensemble Classifier (SEC) is used for the classification. The metrics used for evaluating the performance of the pest detection framework are accuracy, Area Under the Receiver Operating Characteristic Curve, F1-Score and Mean Average Precision (mAP). The results indicate that the proposed SEC with region growing based segmentation framework achieves 98 % of classification accuracy and mAP of 0.96 proving that it is very effective in both classification and segmentation task. The comparative analysis further reveals that the SEC outperforms the existing machine learning models and ensemble learning models like majority voting and weighted average models for process innovation.
An Explainable Uncertainty-Aware Deep Learning Framework for Robust Skin Disease Diagnosis R Karthick Manoj; Aasha Nandhini S; C Batumalai
Journal of Applied Data Sciences Vol 7, No 3: September 2026
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v7i3.1465

Abstract

Automated skin-disease diagnosis from clinical and dermoscopic images remains challenging because of illumination variability, acquisition artifacts, irregular lesion boundaries, limited feature diversity, and the low interpretability of conventional deep learning models. This study proposes EDLF-Skin, an integrated framework that combines Physio-Color Consistency preprocessing, Attentive Multi-scale U-Trans segmentation, Multi-Modal Feature Fusion, a Hybrid Ensemble with Uncertainty, and a Multi-level Explainability Interface. The preprocessing stage normalizes illumination and removes artifacts while preserving diagnostically relevant color and texture information. The segmentation module integrates convolutional feature extraction with transformer-based attention to improve lesion-boundary delineation. Spatial, frequency, and texture representations are subsequently fused through cross-attention and processed by complementary convolutional and transformer classifiers. Predictive uncertainty is estimated to identify low-confidence cases requiring expert review, while Grad-CAM, SHAP attribution, and textual rationales provide multilevel explanations. Experimental results show that the proposed segmentation model achieved a Dice coefficient of 0.956 and an intersection over union of 0.935. The complete classification framework obtained 96.87% accuracy, an F1-score of 0.953, an area under the receiver operating characteristic curve of 0.981, and an expected calibration error of 0.021. These findings indicate that EDLF-Skin improves segmentation accuracy, classification performance, probability calibration, and decision transparency within a unified diagnostic workflow.
A Hybrid CNN-Transformer Model with Quantum-Inspired Fourier Transform for Accurate Skin Disease Classification Aasha Nandhini S; R. Karthick Manoj; M. Batumalay
Journal of Applied Data Sciences Vol 6, No 3: September 2025
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v6i3.782

Abstract

Skin disease classification is a complex task that requires robust feature extraction, efficient classification, and interpretability. Artificial intelligence-based technologies offer effective solutions for developing a framework for skin disease classification while ensuring explainability for healthcare professionals. This study proposes a novel Hybrid Transformer model comprising of Convolutional Neural Network (CNN) architecture infused with a Quantum-Inspired Fourier Transform (QIFT) to enhance classification accuracy. QIFT is incorporated to emphasize frequency-domain information alongside the spatial features captured by CNNs, potentially improving feature representation and model generalization. For demonstration, a dataset containing four different classes of dermatological images is used. Data augmentation techniques and adaptive learning rate scheduling are employed to optimize the dataset. A weighted cross-entropy loss function is used to address class imbalances in the dataset. In this research, explainability is implemented using a standard attribution technique like Integrated Gradients providing insights into model decision-making, and enhancing trust in medical applications. Performance evaluation involves validating the proposed framework using metrics such as confusion matrix analysis, classification reports, and training-validation curves. Experimental results demonstrate a high classification accuracy of 92.5% across skin disease categories. The findings indicate that integrating QIFT and CNN-based feature extraction with transformer-driven attention mechanisms enhances skin disease classification performance while ensuring interpretability as process innovation.