Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Vidhushavarshini Sureshkumar has a total of 39 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    ISNM-Net: A Deep Synergic CNN Architecture for Accurate ‎Grape Leaf Disease Classification Using Mobilenet and Res-‎Net

    by Alvin Ancy A., Vidhushavarshini Sureshkumar, M. Kiruthiga Devi, T. Rathi Devi, A. Joshi, Shobana S., C. Sharmila

    Grape leaf diseases cause serious problems for viticulture around the world by having a substantial impact on the output and quality of ‎grapevine cultivation. Conventional manual diagnosis techniques are laborious, subjective, and frequently ineffectual in extensive field set-‎tings. In this study, we present the Inceptive Synergic Network Model (ISNM), a revolutionary deep learning framework for the precise, ‎effective, and scalable categorization of grape leaf diseases. With a Scale-Invariant Feature Learning (SIFL) module to improve spatial in-‎variance and reduce superfluous background noise, ISNM combines the advantages of Mobile-Net and ResNet-50 as dual backbones for ‎reliable multi-scale feature extraction. We also investigate how wavelet-based sub-band decomposition can enhance feature localization in a ‎variety of illumination and disease-spread scenarios. Lightweight convolutional layers optimize the fused deep features, allowing for de-‎ployment on edge devices for real-time monitoring. The suggested ISNM outperforms baseline CNNs and other conventional designs in ‎terms of precision, recall, and computational efficiency, achieving a state-of-the-art accuracy of 98.75% when tested on the Plant Village ‎grape leaf dataset. With potential uses in precision farming, self-sufficient vineyard monitoring, and sustainable crop management, this ‎work provides a scalable, interpretable, and useful approach to early grapevine disease diagnosis‎.

  • Open Access

    HYBRID OPTIMIZATION FOR TEXTURE-BASED NAIL DISEASE PREDICTION USING DEEP LEARNING

    by Vidhushavarshini Sureshkumar, Vinodhini V, Dhayanithi J, Kiruthiga Devi M, Joshi A, Vijayarasu T

    A wide range of possible disorders may be characterized by color changes in nails; changes in shape of nails; or some type of texture irregularity in the nails that is indicative of a nail disorder. Each of these types of disorders could be the result of a number of different factors including a pathologic process, a previously unknown medical condition; stress; or some form of hereditary component. Of all the disorders of the nails that can occur, some of the most common include mycosis; onycholysis; psoriasis; and paronychia. While there are many similarities among the disorders of the nails, they also present some very significant differences in terms of the symptoms and the difficulty in treating them. An essential step in providing effective treatment for a patient with an underlying condition while protecting the health of their nail, includes an appropriate and thorough evaluation and monitoring of the patient's condition. The purpose of this study was to provide a comprehensive review of the study of the diseases of the nails and their diagnosis utilizing advanced optimization techniques and further to document the results of the experimental investigation. We utilized two metaheuristic optimization techniques called the Moth Fly Optimization (MFO) and Fruit Fly Optimization (FFO), in order to optimize the classification models used within this study. Our data set contained a maximum of 560 images used for training; 70 images used for validating the models; and 70 images of text, while the seven different forms of nail disease were illustrated. We have used variety of deep learning architecture and there accuracy such as 90% with EfficientNetB3, 86% with DenseNet121, 87.14% with ResNet50, 80% with InceptionV3, 90% with Xception, and 91% with MobileNetV2. We observed the highest level of precision of 1.0 with EfficientNetB3 and 0.95 with F1-Score. We were able to demonstrate both a global search capability and improved convergence speed through the use of our Adaptive Fruit Fly Optimization Algorithm (AFFOSM), resulting in an overall system accuracy of 91% using MobileNetV2, significantly better than other conventional methods. Our experimental results indicate that optimizing classification models for detecting nail diseases utilizing metaheuristic algorithms can greatly enhance reliability and investigative accuracy.

  • Open Access

    Ensemble-based Heart Disease Diagnosis (EHDD) using Feature Selection and PCA Extraction Methods

    by V. Vinodhini, B. Sathiyabhama, S. Vidhushavarshini, Vinayakumar Ravi, R. Shashidhar

    Introduction: Heart disease is a growing health crisis in India, with mortality rates on the rise alongside the population. Numerous studies have been undertaken to understand, predict, and prevent this critical illness. The dimensionality of the dataset, on the other hand, reduces the prediction's accuracy. Methods: We propose an Ensemble-based Heart Disease Diagnosis (EHDD) model in which the dimension is lowered through filter-based feature selection. The experimental is conducted using the UCI Cleveland dataset with cardiac disease. The precision is achieved through three key steps. The scatter matrix is utilized to divide the distinct class points in the first phase, and the highest eigenvalue and eigenvectors are picked for the new decreased dimension of the dataset. The feature extraction is carried out in the second stage utilizing a statistical approach based on mean, covariance, and standard deviation. Results: The classification component uses the training and test datasets with a smaller sample space. The last stage is to divide the samples into two groups: healthy subjects and diseased subjects. Since a basic binary classifier will not yield the best results, an ensemble strategy using SVM. Conclusion: Random Forest is chosen to create accurate predictions. When compared to existing models, the suggested EHDD model outperforms them by 98%.


  • Open Access

    Multimodal Skin Cancer Prediction: Integrating Dermoscopic Images and Clinical Metadata with Transfer Learning

    by Ramya Panneerselvam, Sathiyabhama Balasubramaniam, Vidhushavarshini Sureshkumar, Vinayakumar Ravi, Siti Sarah Maidin

    Background: Skin cancers exist as the most pervasive cancers in the world; to increase the survival rates, early prediction has become more predominant. Many conventional techniques frequently depend on visual review of clinical information and dermoscopic illustrations. In recent technological developments, the enthralling algorithms of combining modalities are used for increasing diagnosis accuracy in deep learning. Methods: Our research proposes a multi-faceted approach for the prediction of skin cancer that incorporates clinical metadata with dermoscopic visuals. The pre-trained convolutional neural networks, like EfficientNetB3, were used for dermoscopic images along with transfer learning techniques to excavate some of the visual attributes in this study. Moreover, TabNet was used for processing the clinical metadata, including age, gender, and medical history. The features obtained from both fusion techniques were integrated to enhance the prediction accuracy. The benchmark datasets, like ISIC 2018, ISIC 2019, and HAM10000, were used to assess the model. Results: The proposed multi-faceted system achieved 98.69% accuracy in the classification of skin cancer, surpassing the model that used dermoscopic snapshots with clinical data. The convergence of images with clinical metadata has substantially enhanced prediction resilience, demonstrating the importance of multimodal deep learning in skin lesion diagnosis. Conclusion: This research focused mainly on the efficiency of integrating dermoscopic visuals and clinical information using transfer learning for skin cancer prediction. The proposed system offers a promising tool for improving diagnostic accuracy, and further research could explore its application in other medical fields requiring multimodal data integration.


  • Open Access

    Early Prediction of Cardio Vascular Disease (CVD) from Diabetic Retinopathy using improvised deep Belief Network (I-DBN) with Optimum feature selection technique

    by T. K. Revathi, B. Sathiyabhama, S Kaliraj, Vidhushavarshini Sureshkumar

    Cardio Vascular Disease (CVD) is one of the leading causes of mortality and it is estimated that 1 in 4 deaths happens due to it. The disease prevalence rate becomes higher since there is an inadequate system/model for predicting CVD at an earliest. Diabetic Retinopathy (DR) is a kind of eye disease was associated with increasing risk factors for all-causes of CVD events. The early diagnosis of DR plays a significant role in preventing CVD. However, there are many works have been carried out on classification of the disease but they focused less on feature selection and increasing the accuracy of the model. The proposed work introduces Improvised Deep Belief Network named I-DBN to resolve the above mentioned problems and mainly to concentrate on improving the entire performance of the model leading to the unbiased output. We used Principal Component Analysis (PCA) and Particle Swarm Optimization (PSO) algorithm for feature extraction and selection respectively. Five performance metrics have been used to assess the proposed model. The results of I-DBN outperform other state-of-the-art methods. The result validation ensures that I-DBN can deliver trustworthy recommendations to doctors to treat the patients by enhancing the accuracy of CVD prediction up to 98.95%.


  • Open Access

    Enhancing Medical Image Quality Using Fractional Order Denoising Integrated with Transfer Learning

    by Abirami Annadurai, Vidhushavarshini Sureshkumar, Dhayanithi Jaganathan, Seshathiri Dhanasekaran

    In medical imaging, noise can significantly obscure critical details, complicating diagnosis and treatment. Traditional denoising techniques often struggle to maintain a balance between noise reduction and detail preservation. To address this challenge, we propose an “Efficient Transfer-Learning-Based Fractional Order Image Denoising Approach in Medical Image Analysis (ETLFOD)” method. Our approach uniquely integrates transfer learning with fractional order techniques, leveraging pre-trained models such as DenseNet121 to adapt to the specific needs of medical image denoising. This method enhances denoising performance while preserving essential image details. The ETLFOD model has demonstrated superior performance compared to state-of-the-art (SOTA) techniques. For instance, our DenseNet121 model achieved an accuracy of 98.01%, precision of 98%, and recall of 98%, significantly outperforming traditional denoising methods. Specific results include a 95% accuracy, 98% precision, 99% recall, and 96% F1-score for MRI brain datasets, and an 88% accuracy, 91% precision, 95% recall, and 88% F1-score for COVID-19 lung data. X-ray pneumonia results in the lung CT dataset showed a 92% accuracy, 97% precision, 98% recall, and 93% F1-score. It is important to note that while we report performance metrics in this paper, the primary evaluation of our approach is based on the comparison of original noisy images with the denoised outputs, ensuring a focus on image quality enhancement rather than classification performance.


  • Open Access

    Breast Cancer Detection and Analytics Using Hybrid CNN and Extreme Learning Machine

    by Vidhushavarshini Sureshkumar, Rubesh Sharma Navani Prasad, Sathiyabhama Balasubramaniam, Dhayanithi Jagannathan, Jayanthi Daniel, Seshathiri Dhanasekaran

    Early detection of breast cancer is essential for increasing survival rates, as it is one of the primary causes of death for women globally. Mammograms are extensively used by physicians for diagnosis, but selecting appropriate algorithms for image enhancement, segmentation, feature extraction, and classification remains a significant research challenge. This paper presents a computer-aided diagnosis (CAD)-based hybrid model combining convolutional neural networks (CNN) with a pruned ensembled extreme learning machine (HCPELM) to enhance breast cancer detection, segmentation, feature extraction, and classification. The model employs the rectified linear unit (ReLU) activation function to enhance data analytics after removing artifacts and pectoral muscles, and the HCPELM hybridized with the CNN model improves feature extraction. The hybrid elements are convolutional and fully connected layers. Convolutional layers extract spatial features like edges, textures, and more complex features in deeper layers. The fully connected layers take these features and combine them in a non-linear manner to perform the final classification. ELM performs classification and recognition tasks, aiming for state-of-the-art performance. This hybrid classifier is used for transfer learning by freezing certain layers and modifying the architecture to reduce parameters, easing cancer detection. The HCPELM classifier was trained using the MIAS database and evaluated against benchmark methods. It achieved a breast image recognition accuracy of 86%, outperforming benchmark deep learning models. HCPELM is demonstrating superior performance in early detection and diagnosis, thus aiding healthcare practitioners in breast cancer diagnosis.


  • Open Access

    A Comparative Study on Thyroid Nodule Classification Using Transfer Learning Methods

    by Vidhushavarshini Sureshkumar, Dhayanithi Jaganathan, Vinayakumar Ravi, Vinodhini Velleangiri, Pradeep Ravi

    Introduction: The thyroid is an endocrine gland located in the front of the neck whose main purpose is to produce thyroid hormones necessary for the functioning of the entire body. Thyroid hormones may be produced too little or too much depending on dysfunction. Since the 1990s, there have been an increasing number of thyroid illness cases, and in recent years, thyroid cancer has become the malignancy with the fastest rate of increase. According to recent studies, thyroid dysfunction affects 42 million people in India. Much research has provided solutions for thyroid classification. Methods: In this paper, we survey various transfer learning models to classify thyroid nodules and predict the best accuracy. Our study evaluated several models, including DenseNet169, ResNet101, and various EfficientNet variants, using a comprehensive dataset comprising 7,893 images. DenseNet169 achieved the highest accuracy at 95.96%, followed by ResNet101 and EfficientNetB1, with accuracies of 94.74% and 86.14%, respectively. The models were rigorously tuned and optimized using grid search strategies, with hyperparameters such as learning rate, batch size, optimizer type, and dropout rate carefully selected to enhance performance. The evaluation included precision, recall, and F1 score metrics, ensuring balanced performance across different metrics. Results: Our results demonstrate that advanced transfer learning models can distinguish malignancy from benign conditions with greater accuracy than traditional diagnostic approaches reliant on the human eye. Conclusion: This research highlights the potential of integrating AI techniques in medical diagnostics to improve the accuracy and reliability of thyroid disease detection, ultimately leading to better patient outcomes.”