Radha R

@sdnbvc.edu.in

Associate Professor, Computer Science
Shrimathi Devkunvar Nanalal Bhatt Vaishnav College for Women

Radha R


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

RESEARCH, TEACHING, or OTHER INTERESTS

Computer Science, Computer Vision and Pattern Recognition
21

Scopus Publications

Scopus Publications

  • Multi-Level Pelican Optimization with Fuzzy Thresholding and Deep Residual CNN for Enhanced Fatty Liver Classification in Abdominal CT Scans
    R Radha, Blessy Boaz
    2025 International Conference on Emerging Trends in Signal Processing and Computational Intelligence Iccspci 2025, 2025
    Fatty liver disease is one of the most common liver problems in the world; however, it is frequently overlooked in its early stages since the variations in abdominal CT scans tend to be compact. This paper introduces a novel hybrid framework that integrates Multi-Level Pelican Optimization (MLPO), Fuzzy Thresholding, and a Deep Residual Convolutional Neural Network (ResCNN) to facilitate the identification and classification of fatty liver. The proposed framework uses MLPO in two different ways: first, to improve the quality of the image by choosing the best contrast parameters, and second, to find the most distinguishing characteristics after liver segmentation. Fuzzy thresholding is different from typical binary segmentation algorithms since it gives pixels membership values based on their intensity levels. This renders it easier to find soft boundaries, which makes it easier to delineate the liver region even in low-contrast or noisy images. The segmented liver area is achieved through a ResCNN architecture, utilizing deep feature extraction and skip connections to preserve spatial and contextual information. This approach assists the classifier in telling the difference between normal and fatty liver tissues, even when there is a lot of variation in classes and intensities overlap. Many tests have been performed on annotated abdominal CT datasets, and the results were based on accuracy, sensitivity, specificity, F1-score, and the Dice similarity coefficient. The results show that the new MLPO-Fuzzy-ResCNN framework does an excellent job of classifying things better than other methods. This study provides a strong, automated pipeline that can be used in clinical environments to help with early diagnosis and successful monitoring of fatty liver disease utilizing non-invasive imaging methods.
  • B-Vote: A Novel Stratification Framework for Fake Text Analysis
    J. Sahaya Sheeba Mangalam, R. Radha, R. Srividhya
    Proceedings of the 2025 14th International Conference on System Modeling and Advancement in Research Trends Smart 2025, 2025
    Misinformation which has also been referred to as fabricated news, consists of information that is either erroneous or misleading but propagated and presented as genuine journalistic reporting. Psychological factors affect the propensity of people to spread false news: empirical studies show that people have an increased interest in new and unusual events, which, consequently, increases neural activity. Given significant technological advancements, the challenge of rapidly disseminating misinformation has become increasingly salient. This paper addresses these challenges by detailing a machine learning process designed to identify and categorize fake news effectively. To tackle misinformation effectively, this study proposes a novel methodology grounded in Machine Learning and Natural Language Processing, aimed at identifying misinformation, with particular emphasis on news articles originating from sources considered unreliable. This study uses a dataset with equal parts authentic and fabricated news to explore how advanced machine learning algorithms can detect fake news. The methodology for identifying fake news follows a structured process: 1) Data acquisition, 2) Data preprocessing, which includes text cleansing and TF-IDF vectorization, and 3) Application of various classification algorithms. This research advocates for the use of ensemble machine learning methods, explaining how combining multiple algorithms enhances the accuracy of automated news article classification. The study examines a variety of textual features to distinguish fabricated content from genuine information. Using these capabilities, it is possible to train a synthesis of various machine learning algorithms using a variety of ensemble methods and assess its performance. The suggested study will ensure that the adopted methodology defines a sound preprocessing tool to increase precision of classification during the later phases of investigation. Analysis performed is simulated and the results are achieved successfully.
  • Introduction to Transformation of Blockchain-Based Digital Twins
    Blockchain Based Digital Twins Research Trends and Challenges, 2025
  • Automated Vehicle Number Plate (VNP) Detection based on Optimized Segmentation and Machine Learning
    R. Radha, V.R. Viju
    International Conference on Sustainable Computing and Data Communication Systems Icscds 2022 Proceedings, 2022
    Vehicle number plate (VNP) detection is a rather difficult operation unless we assume the use of a static camera, fluctuations in illumination, known VNP templates, ensured color patterns, and other simple assumptions. Practical applications require robust and generalized VNP detection methods to meet complex situations. By treating the vehicle VNP as an object, this research presents an innovative solution to this problem. The primary purpose of this study is to address the following VNP detection challenges: (1) VNP detection in each frame of an image sequence, (2) partial VNP detection, and (3) VNP detection by moving cameras and cars. This research compares a segmentation method for Artificial Neural Fuzzy Inference System classification against a variety of traditional methods and state-of. the-art object identification approaches (ANFIS). The expectation maximum (EM) approach can be used to compute the ANFIS parameters. A high recognition rate can be attained with this strategy. Extensive tests and comparisons show that the experimental results outperform standard methods.
  • Heavy-Vehicle Detection Based on YOLOv4 featuring Data Augmentation and Transfer-Learning Techniques
    V Sowmya, R Radha
    Journal of Physics Conference Series, 2021
    Real-time Vehicle detection is crucial in today’s era for our complex interconnected transportation ecosystem built on an advanced technological network of intelligent systems encompassing a wide range of applications such as autonomous vehicles, traffic Surveillance, advanced driver assistance systems, and etcetera. The significance of its application to digital transportation infrastructure embarks upon a distinct framework for heavy-vehicle detection integrated with the YOLOv4 algorithm for real-time detection. In this proposed work, two entities of heavy vehicles such as buses, trucks are considered. The crux of the model, an algorithmic computational mechanism incorporates Mosaic Data augmentation and Transfer-learning techniques that are applied to avoid over-fitting and to improve the optimal speed during training. Subsequently, a fine-tuning YOLOv4 algorithm is implemented for detecting the heavy vehicle. The algorithm is tested for real-time situations in various traffic densities through Computer Vision. Experimental results show that the proposed system achieves higher detection accuracy of 96.54% mAP. More specifically, the performance of the proposed algorithm with the COCO test set and PASCAL VOC 2007 test set demonstrates improvement when compared with other state-of-the-art approaches.
  • Comparative analysis of various feature extracting algorithms using satellite images
    Hema Malini G.B.
    Journal of Advanced Research in Dynamical and Control Systems, 2020
  • A sentiment classification on Indian government schemes using pyspark
    International Journal on Emerging Technologies, 2020
  • A novel approach to segment leaf region from plant leaf image using automatic enhanced grabcut algorithm
    Compusoft, 2019
  • Preprocessingpubmed abstracts
    S. Vijaya, R. Radha
    Proceedings of the 2017 2nd IEEE International Conference on Electrical Computer and Communication Technologies Icecct 2017, 2017
    Vast growth of biomedical databases has increased most of the researchers focus on the field of Text Mining. The documents appear in unstructured format. To process and discover knowledge from these data, the unstructured databases must be converted to structured format. For this task Text mining plays a vital role. Text preprocessing is an essential step in text mining. The common preprocessing tasks in text mining are Tokenizing, Removing Stop words and Stemming. In this paper we have discussed the implementation steps we have done on PubMed abstract using Rapidminer.
  • Review of automatic detection and grading of diabetic maculopathy
    International Journal of Applied Engineering Research, 2015
  • Data duplication using ensemble classification
    International Journal of Applied Engineering Research, 2015
  • Detecting the optic disc boundary and macula region in digital fundus images using bit-plane slicing, edge direction, and wavelet transform techniques
    Arpn Journal of Engineering and Applied Sciences, 2015
  • Using K-means clustering technique to study of breast cancer
    R. Radha, P. Rajendiran
    Proceedings 2014 World Congress on Computing and Communication Technologies Wccct 2014, 2014
  • Automatic extraction, segmentation and recognition of multi-font Indian Pincode
    R. Radha, R.R. Aparna
    International Journal of Computational Vision and Robotics, 2014
  • A novel approach for Postal address block and Pincode segmentation
    International Journal of Imaging and Robotics, 2014
  • An effective algorithm for edges and veins detection in leaf images
    R. Radha, S. Jeyalakshmi
    Proceedings 2014 World Congress on Computing and Communication Technologies Wccct 2014, 2014
  • Digit recognition using hybrid classifier
    R. Radha, R. R. Aparna
    Proceedings 2014 World Congress on Computing and Communication Technologies Wccct 2014, 2014
  • Identification of retinal image features using bitplane separation and mathematical morphology
    R. Radha, Bijee Lakshman
    Proceedings 2014 World Congress on Computing and Communication Technologies Wccct 2014, 2014
  • Analysis of gene expression profile to select patient samples for outcome prediction
    R. Radha, S. Jayalakshm, S.P. Rajagopala
    Information Technology Journal, 2008
  • Assessing the interestingness of discovered knowledge using a hybrid approach based on fuzzy concepts
    International Journal of Soft Computing, 2007
  • Fuzzy logic approach for diagnosis of Diabetics
    R. Radha, S.P. Rajagopala
    Information Technology Journal, 2007