Electrical and Electronic Engineering, Computer Networks and Communications, Signal Processing, Computer Vision and Pattern Recognition
7
Scopus Publications
Scopus Publications
Early Detection of Liver Disease and Precise Health Optimization using Advanced Computational Models S Rubin Bose, S. Viswanathan, V. Mahalakshmi, V. Hari, J Angelin Jeba, R Regin Proceedings of the 11th International Conference on Bio Signals Images and Instrumentation Icbsii 2025, 2025 The progressive nature of liver disease makes it a worrying factor globally and is often detected too late, where treatment options are sparse. As with most diseases, early detection makes a significant difference to the outcomes a patient faces and the resources that need to be devoted by the healthcare system. The purpose of this paper is to propose a predictive model for liver disease detection with the implementation of Deep Learning TabNet architecture, which is specialized in analyzing structured data. In contrast to other models, TabNet uses a novel approach that improves feature selection and overall accuracy through the use of attention mechanisms. The system’s non-stop surveillance of patient data enables it to report irregularities followed by real-time alerts for timely action to be taken. Additionally, batch data processing allows for a thorough analysis of the data that captures various trends for proactive disease control. The enhancement of deep learning for the predictive factors of liver disease is expected to strengthen preventive medicine and make further strides in AI diagnostics and patient care. The proposed TabNet model achieves a recall of 0.991, precision of 0.992, and an F1 score of 0.991 at a learning rate of 0.0015, demonstrating exceptional performance in accurately identifying liver disease.
An Efficient Deep Convolutional Neural Network for Automatic Detection of FDG Uptakes in Fused PET/CT Images J. Angelin Jeba, S. Nirmala Devi IETE Journal of Research, 2024 This paper emphasizes on detection and classification of Fluoro-Deoxy-Glucose (FDG) radioactivity uptakes in fused Positron Emission Tomography / Computerized Tomography (PET/CT) images automatically. The deep learning technique using Convolutional Neural Network (CNN) is proposed to reduce the complexity in observation, to solve the problem of low accurateness and the time-consuming process of traditional classification methods. The CNN layers are designed and proposed for the FDG uptakes classification problem in fused PET/CT images. The proposed modified CNN model is trained using different optimizers such as Stochastic Gradient Descent Momentum (SGDM), Adaptive Moment Estimation (ADAM), and Root Mean Square propagation (RMSprop). The deep features extracted from the proposed CNN are classified using different classifiers such as K-Nearest Neighbor (KNN), Decision Trees (DT), Ensemble, Naive Bayes (NB), and multi-class Support Vector Machine (SVM) the results of which are compared. The multi-class SVM classifier trained using SGDM optimizer attains the maximum test accuracy of 98.18% and was found to be superior to pre-trained deep models such as AlexNet, ResNet, and GoogleNet.
Efficient Skin Cancer Diagnosis and Classification via High-Speed Deep Learning Architecture S Rubin Bose, M Izzath Suhail, Shaik Roshni Shabnam, B Hariharan, Angelin Jeba J, Regin R Proceedings of the 2024 10th International Conference on Biosignals Images and Instrumentation Icbsii 2024, 2024 Skin cancer is still a major worldwide health issue, which highlights the necessity for quick and accurate detection methods. This research introduces a robust skin cancer classification system leveraging the capabilities of YOLOv8, An advanced algorithm for object detection. The proposed model processes dermatoscopic images with remarkable speed and accuracy, facilitating the identification of malignant lesions. The YOLOv8 architecture enables classification of various skin cancer types, including melanoma and non-melanoma, by effectively localizing and classifying lesions within the images. A comprehensive dataset, comprising diverse skin lesions, was employed for training and validation, ensuring the model's adaptability to different clinical scenarios. The evaluation of the proposed model demonstrates superior performance compared to traditional methods, exhibiting high sensitivity and specificity.
Enhanced Lung Tumor Detection via Deep Learning Techniques in CT Imaging S Rubin Bose, V Karrthik Kishore, K R Abishek, J Chalwin Ajay, Angelin Jeba J, Regin R Proceedings of the 2024 10th International Conference on Biosignals Images and Instrumentation Icbsii 2024, 2024 Lung cancer is one of the most common causes of cancer-related death worldwide. Early detection is essential for better patient outcomes. Both the YOLOv8 algorithm and Convolutional Neural Networks (CNNs) have demonstrated promise in the field of medical image processing and object detection, respectively. In this research, we present an innovative method for the detection of lung cancer utilising a CNN model and the YOLOv8 algorithm, integrated into a mobile application developed in Kotlin for enhanced accessibility. Our system takes advantage of the YOLOv8 algorithm's real-time object detection capabilities to recognise nodules of lung cancer using CT scans. The CNN model is trained on a dataset of CT scans and is capable of differentiating between benign and malignant nodules. The mobile application provides a user-friendly interface for uploading CT scans and receiving real-time diagnostic results. The proposed system intends to improve lung cancer early detection by providing a convenient and efficient tool for healthcare professionals and patients. By utilising the strength of CNNs and YOLOv8, this technology has the capacity to reduce false negatives and improve overall diagnostic accuracy, ultimately contributing to better patient care and outcomes. Our model performs well overall, with 98.06% precision, 96.57% recall and 97.31% F1 score when test on a benchmark dataset of images of lung cancer.
Modified CNN Architecture for Efficient Classification of Glioma Brain Tumour J. Angelin Jeba, S. Nirmala Devi, M. Meena IETE Journal of Research, 2023 Magnetic Resonance Imaging (MRI) is the medical imaging modality that provides more useful functional data for the diagnosis of pathological conditions in brain tumour than any other modality. Manual observation of MRI data to diagnose the tumour is time-consuming and hence the objective of this work is to classify the Glioma brain tumour using a Convolutional Neural Network (CNN). This proposed work aims to design a new model of the modified CNN architecture for the classification of Gliomas. Various processes were used for the classification of MRI brain tumours, which include image pre-processing, image feature extraction, and subsequent classification of Glioma brain tumours. The proposed modified CNN obtained high classification accuracy of 94.65% compared to the pre-trained AlexNet Model. The traditional machine learning techniques like Support Vector Machine (SVM) and K-Nearest Neighbour (KNN) gain an accuracy of 86.1% and 66.7%, respectively.
Shape Description of FDG uptakes in Pre and Postoperative fused PET/CT Images J. Angelin Jeba, S. Nirmala Devi 2020 6th International Conference on Bio Signals Images and Instrumentation Icbsii 2020, 2020 Automatic object recognition with shape descriptors help to interact with the real-time environment. This paper describes the shape analysis of radioactivity Fluoro Deoxy Glucose (FDG) uptakes present in pre and postoperative stages of Fused PET/CT images using shape feature extraction approach. Shape features are invariant to various affine transformations such as translation, rotational, flipped, scale, etc., are more robust for shape analysis. Geometrical properties of the FDG uptakes are extracted at various levels of spatial resolution by hierarchical Kdabstraction from the segmented PET/CT images. Rotational invariants considered are the Zernike moments magnitudes and the shape signatures of Fourier descriptors are investigated. The discrimination power of features between pre and post-operative images is examined and evaluated.
