DEVELOPMENT AND EVALUATION OF PNEUMFC NET: A NOVEL AUTOMATED LIGHTWEIGHT FULLY CONVOLUTIONAL NEURAL NETWORK MODEL FOR PNEUMONIA DETECTION
- Title
- DEVELOPMENT AND EVALUATION OF PNEUMFC NET: A NOVEL AUTOMATED LIGHTWEIGHT FULLY CONVOLUTIONAL NEURAL NETWORK MODEL FOR PNEUMONIA DETECTION
- Creator
- Prakash S.; Ramamurthy B.
- Description
- The aim of this study is to address the challenges of pneumonia diagnosis under constraint resources and the need for quick decision making. We present the PneumFC Net, a novel architectural solution where our approach focuses on minimizing the number of trainable parameters by incorporating transition blocks that efficiently manage channel dimensions and reduce number of channels. In contrast to using fully connected layers, which disregard the spatial structure of feature maps and substantially increase parameter counts, we exclusively employ only convolutional layer approach. In the study, X-ray image dataset is used to train and evaluate the proposed Convolutional Neural Network model. By carefully designing the architecture, the model achieves a balance between parameters and accuracy while maintaining comparable performance to pre-trained models. The results demonstrate the model's effectiveness in detecting pneumonia images reliably. In addition, the study examines the decision-making process of the model using Grad-CAM, which helps to identify important aspects of radiographic images that contribute to the positive pneumonia prediction. Furthermore, the study shows that the proposed model, Pneum FC Net not only has the highest accuracy of 98%, but the total trainable model parameters is only 0.02% of the next best model VGG-16, thus establishing the potential of this new robust Deep Learning model. This research primarily addresses concerns related to mitigating significant computational requirements, with a specific focus on implementing lightweight networks. The contribution of this work involves the development of resource-efficient and scalable solution for pneumonia detection. 2024 Little Lion Scientific. All rights reserved.
- Source
- Journal of Theoretical and Applied Information Technology, Vol-102, No. 3, pp. 1037-1048.
- Date
- 2024-01-01
- Publisher
- Little Lion Scientific
- Subject
- Computer Aided Diagnosis; Fully Convolutional Neural Network; PneumFC Net; Pneumonia Detection
- Coverage
- Prakash S., Department of Computer Science, Christ University, Karnataka, Bengaluru, India; Ramamurthy B., Department of Computer Science, Christ University, Karnataka, Bengaluru, India
- Rights
- Restricted Access
- Relation
- ISSN: 19928645
- Format
- Online
- Language
- English
- Type
- Article
Collection
Citation
Prakash S.; Ramamurthy B., “DEVELOPMENT AND EVALUATION OF PNEUMFC NET: A NOVEL AUTOMATED LIGHTWEIGHT FULLY CONVOLUTIONAL NEURAL NETWORK MODEL FOR PNEUMONIA DETECTION,” CHRIST (Deemed To Be University) Institutional Repository, accessed February 27, 2025, https://archives.christuniversity.in/items/show/13341.