Ph.D Theses

Permanent URI for this collection

https://ir.avinuty.ac.in/handle/avu/9

Browse

Browsing Ph.D Theses by Subject "Electronics and Communication Engineering"

Now showing 1 - 3 of 3
  • No Thumbnail Available
    Item
    A Deep Learning Framework for Detection and Segmentation of Multiple Artefacts in Endoscopic Images
    (Avinashilingam, 2023-05) Kirthika N; Dr.B.Sargunam
    Endoscopy is a standard procedure for disease surveillance, monitoring inflammations, detect cancer and tumor. During the procedure the organs are visualized. Artefacts, an artificial effect is found to be present in the resultant images. They play a dominant role in increasing procedure time by more than an hour. Hence an efficient algorithm to detect, segment and restore could assist clinician. The artefacts present in an endoscopic image include saturation, specular reflections, blur, bubbles, contrast, blood, instruments and miscellaneous artefacts. The presence of these artefacts acts as a barrier when investigating the underlying tissue for identifying clinical abnormalities. It also affect post processing steps where most of the images captured are discarded due to the presence of artefacts which in turn affects information storage and extracting useful frame for report generation. Endoscopic artefact detection dataset is the only available public dataset holding endoscopic images with annotations for multiple artefacts. Hence, a custom dataset is annotated using the same annotation protocol of endoscopic artefact detection dataset to maintain homogeneity. The algorithms are trained and tested with images from both public and custom dataset for artefact detection. State of the art object detection algorithms such as YOLOv3, YOLOv4 and faster R-CNN are used for detecting artefacts in endoscopic images. The detection algorithm focusses on three important performance parameters namely mean average precision, intersection over union and inference time. The ensemble model outperformed well across all the performance parameters compared with literature. The inference time is reduced by 8.63%, whereas the mAP and IoU are increased by 61.67% and 63.47% respectively. newlineSegmentation algorithms like U-Net with EfficientNetB3 backbone, Link-Net with EfficientNetB3 backbone and U-Net with SE-ResNeXt101 backbone are used to segment the artefacts. The results are assessed with performance parameters like F2 score and Jaccard score.The results proves a phenomenal increase in Jaccard score by 17.36% and F2 score by 17.42% respectively. An image if found to have artefacts after artefact detection,the affected region will be segmented by the proposed segmentation algorithm.To visualize the scope and need of artefact detection and segmentation a simple application is developed to restore the artefacts. The segmented output contains a binary mask using which fast marching algorithm will restore the segmented area. Hence the resulting restored image gives the clinician a better view of the organ. A simple CNN based classifier is proposed to classify polyp. It is found that the classifier's performance is improved by 3.09% when the artefacts in the images are restored. Thus such outcomes when implemented in real time could effectively have a control over the false diagnosis rate, which is the rate at which the disease is misclassified, procedure time and clinician's fatigue as well.
  • No Thumbnail Available
    Item
    Classification of Diabetic Retinopathy Stages Using Deep Learning Architectures
    (Avinashilingam, 2025-08) Santhiya Lakshmi K; Guide - Dr. B. SARGUNAM
    Diabetic Retinopathy (DR) is a progressive microvascular complication of diabetes and a leading cause of vision impairment and blindness worldwide. Early detection and accurate classification of DR are essential for timely medical intervention and prevention of severe visual deterioration. Traditional diagnostic approaches rely on manual grading by ophthalmologists, which is time-consuming, labor-intensive, and susceptible to inter- observer variability. Recent advancements in Deep Learning (DL) have demonstrated significant potential in automating DR classification; however, existing models face challenges related to overfitting, computational complexity, and generalization across diverse datasets. This research aims to develop a robust and computationally efficient DR classification system capable of accurately detecting all four stages of DR like normal, mild, moderate and severe while addressing these challenges. The proposed methodology follows a three step approach to enhance model performance and adaptability. In the first phase, Transfer Learning (TL) is employed to fine-tune state-of-the-art pretrained models, particularly EfficientNetV2L, using the Asia Pacific Tele-Ophthalmology Society (APTOS) dataset. The objective is to utilize deep feature extraction capabilities while ensuring improved generalization. Data augmentation and scaling techniques are incorporated to mitigate overfitting and enhance model robustness. In the second phase, a custom Convolutional Neural Network (CNN) architecture is designed to optimize feature extraction and computational efficiency. The architecture incorporates structured convolutional layers, batch normalization, and dropout mechanisms to enhance learning stability while reducing the risk of overfitting. In the third phase, a hybrid model is developed by integrating the EfficientNetV2L with the custom CNN architecture, thereby utilising the strengths of both architectures. Extensive hyperparameter tuning is performed, including learning rate scheduling, dropout regularization, batch size optimization, and adaptive optimization using Adam optimizer, to ensure efficient training and convergence. The hybrid model is designed to provide a balance between high classification accuracy and computational feasibility, facilitating its deployment in clinical applications.The proposed system is evaluated using key performance metrics, including accuracy, precision, recall, F1-score, and computational complexity (trainable parameters).The hybrid model (EfficientNetV2L + Custom CNN Model) achieves the highest classification accuracy of 94%, with a precision of 94%, recall of 93%, and F1-score of 94%, utilizing 441,087 trainable parameters. Comparative analysis with benchmark models demonstrates the superiority of the proposed approach, with EfficientNetV2L achieving 92% accuracy, the Custom CNN Model reaching 88% accuracy, while conventional architectures such as VGG16 (85%), ResNet-50 (87%), InceptionV3 (89%), and DenseNet121 (90%) exhibit relatively lower performance. These findings confirm the effectiveness of the hybrid approach in achieving state-of-the-art classification accuracy while maintaining computational performance. The developed hybrid model is evaluated against an existing architecture that integrates DenseNet121, Xception, and EfficientNetB3, comprising 2,361,860 trainable parameters. The existing model achieves an accuracy of 75%, precision of 73%, recall of 75%, and an F1 score of 71%. In contrast, the proposed model, with only 441,087 parameters, demonstrates superior performance achieving a 19% increase in accuracy, 21% in precision, 18% in recall, and 23% in F1 score highlighting its effectiveness despite a significantly reduced parameter count. This research contributes to the advancement of automated DR classification by integrating fine-tuned pretrained models with a lightweight custom CNN, resulting in a scalable, efficient, and clinically viable diagnostic framework. The system is validated using clinical data to ensure its applicability in practical healthcare settings, particularly in resource-constrained environments where access to high-end computational resources is limited. The proposed model has the potential to significantly enhance early DR detection, improve diagnostic consistency, and ultimately contribute to better patient outcomes in ophthalmic disease management.
  • No Thumbnail Available
    Item
    Energy Efficient Clustering and Routing Techniques In Hierarchical Wireless Sensor Networks
    (Avinashilingam, 2024-03) Makimaa Y P; Dr. R. Sudarmani
    The Wireless Sensor Network (WSN) has become an important field of research in Wireless Communication (WC). Nowadays, many real-time applications are using sensors due to their characteristics such as scalability, small size, lightweight, portable, etc. Several sensor nodes associated with WSNs are randomly dispersed in the surrounding environment with limited sensing, computing, and communication capabilities. All the units in the sensor nodes are low powered battery-operated devices. The energy associated with the batteries is finite and must be replaced when they run out, which raises the expense of maintenance. Self-organization and concurrent processing are the important characteristics of WSN. Once the transmission takes place, the energy gets depleted and the node replacement is impossible even if the single node dies which results in network failure. Energy consumption, network lifetime, and security are the major challenges faced by WSN. Maintaining a balance between energy consumption and network lifetime is always a challenging factor in WSN as both are inversely proportional to each other. Secured clustered routing techniques can be adopted to overcome this challenge. The proposed work aims to provide a secured clustered routing protocol. This research work is divided into 3 stages namely clustering, routing, and security. Clustering is the process of grouping sensor nodes and electing proper Cluster Heads (CHs) to avoid long distance communication to the base station, optimizing energy consumption and providing a better quality of service. The first work is to form a proper cluster and select appropriate Cluster Heads (CHs) using a Genetic Algorithm (GA) and Algorithm for Cluster Establishment (ACE). Data transmission in WSN consumes more energy compared to processing. Long distance nodes consume more energy than nodes with short distance communication, so maintaining energy across the network is difficult due to limited energy resources. To prolong network lifetime and for energy efficient transmission, nodes with sufficient energy are required. In this work, an energy optimization algorithm based on GA overcomes this drawback and finds a suitable CH, which leads to cluster formation. Routing is the process of finding the shortest distance between the nodes for transferring the data. Sometimes the data is lost during the transmission due to long distance communication, node failure, and the presence of malicious nodes. Still, research is going on to overcome these challenges. The second phase of work is to perform routing based on the demand and node behaviour. Routing protocols improve the network lifetime and establish the perfect communication between the nodes. Since node behaviour detection is important, this work mainly focuses on maintaining the connectivity between the nodes and increasing the network lifetime by incorporating two processes such as node behaviour analysis and on- demand secured data transmission. Detecting the node behaviour will monitor the transmission path for continuous message transmission to detect malicious nodes that block the path. Congestion occurs when selfish nodes interfere with overall communication. Nodal behaviour changes create network disruptions. Predicting node behaviour detects malicious nodes in the network. The main advantage in predicting node behaviour is to differentiate between malicious and selfish nodes. Since malicious node causes errors, they are removed from the network. The semi-Markov process predicts the node behaviour, which brings trust to the network, as this method can alter the node formation at the point of detecting malicious nodes and gain trust between the nodes. In the third stage, a secured clustered-based routing protocol establishes a satisfactory path to transmit the data without any loss. An energy-efficient and secured routing protocol called Multi Criteria Based Secured Routing Protocol (MSRP) has been developed. Every node can interact with the server and other nodes. Every time, the control layer is responsible for performing additional processing and stops all network processes if a malicious node is suspected. Several methods have been put forward to increase security and energy efficiency. By the current system, if malicious nodes are engaged in the network, it consumes more energy and produces more unwanted data as well as threatening the data captured, thus increasing computational time and complexity. This MSRP model helps to reduce the unwanted communication and data loss. It also separates the malicious nodes from the network once they are identified. In this research, clustering, routing, and security models have been implemented and compared with the existing protocols. In which, the proposed clustering algorithm based on ACE with GA performs better compared to LEACH protocol. The routing technique based on node behaviour performs better than the existing Trust Management System (TMS) and Reliable Trustworthy Approach (RTA). The protocol for secured routing is based on the multiple criteria, and performed well compared with the ESMR (Energy Efficient and Secure Mobile node Re-authentication scheme), AODV (Ad-hoc On Demand Distance Vector Routing), and TSRF (Trust aware Secure Routing Framework) techniques.