Loading Your Library
Preparing your reading experience...
Loading content...



Ariston Publication
3167 Riverview Place
Columbus, OH 43202
United States of America
Vidya Kamma, S. Gopinath, B. Buvaneswari, S. Asha, Maliuk Andrei, Loybenitez Jorge Rafael, S. M. Riazul Islam, Nur Alam MD
Vidya Kamma ¹,*
S. Gopinath ²
B. Buvaneswari ³
S. Asha ⁴
Maliuk Andrei 5
Loybenitez Jorge Rafael 5
S. M. Riazul Islam 6
Nur Alam MD 5
¹ Department of Computer Science and Engineering, Neil Gogte Institute of Technology, Rangareddy, Hyderabad, Telangana, India
² Department of Information Technology, Gnanamani College of Technology, Namakkal, Tamil Nadu, India
³ Department of Information Technology, Panimalar Engineering College, Chennai, Tamil Nadu, India
⁴ Department of Information Technology, S.A. Engineering College, Chennai, Tamil Nadu, India
5 Department of Artificial Intelligence, Kyungdong University Global, South Korea,
6 Department of School of Natural & Computing Sciences, Computing Science, UK
* Author to whom correspondence should be addressed: kammavidya@gmail.com (Vidya Kamma)
Medical image segmentation is essential for accurate clinical diagnosis by enabling precise delineation of anatomical structures and pathological regions. This paper proposes a Hybrid Graph Neural Network (HGNN) framework that integrates Convolutional Neural Networks (CNNs) for local spatial feature extraction with Graph Convolutional Networks (GCNs) for modeling structural and relational dependencies among image regions. The framework converts spatial feature maps into graph-based representations, where nodes represent image regions and edges capture anatomical relationships. A feature fusion mechanism combines CNN and GCN embeddings to generate accurate segmentation predictions. Experimental evaluation on publicly available MRI datasets demonstrates that the proposed model outperforms conventional segmentation methods in terms of segmentation accuracy, overlap measures, sensitivity, and specificity. By effectively capturing both local image features and global contextual information, the HGNN framework improves segmentation performance and model interpretability. The proposed approach provides a robust and scalable solution for medical image analysis with potential applications across diverse imaging modalities and clinical decision-support systems.

This research advances medical image analysis by introducing a Hybrid Graph Neural Network framework that combines convolutional and graph-based learning to improve segmentation accuracy and interpretability. The integration of local spatial feature extraction with global relational modeling enables precise delineation of complex anatomical structures and pathological regions. The proposed framework supports reliable clinical diagnosis, treatment planning, and decision support while maintaining computational efficiency. Its scalable architecture offers broad applicability across diverse medical imaging modalities and intelligent healthcare systems.
This study proposes a Hybrid Graph Neural Network (HGNN) framework for medical image segmentation by integrating Convolutional Neural Networks with Graph Convolutional Networks. CNNs extract local spatial features, while GCNs model structural relationships among anatomical regions using graph representations. Experimental evaluation on MRI datasets demonstrates superior segmentation accuracy, Dice coefficient, sensitivity, specificity, and Intersection-over-Union compared with conventional methods. The proposed framework effectively captures local and global contextual information, providing an accurate, interpretable, and scalable solution for advanced medical image analysis and clinical decision support.
Article type: Research Article
Submitted: 3 November 2025
Revised: 9 February 2026
Accepted: 12 March 2026
First published: 2 April 2026
DOI: https://doi.org/10.69626/cai.2026.0118
Volume: 03
Issue: 02
Pages: 118–131
Citation: CompSci & AI Advances, 3(2), 118–131 (2026).
Kamma, V., Gopinath, S., Buvaneswari, B., Asha, S., Andrei, M., Rafael, L. J., Islam, S. M. R. and MD, N. A. 2026. A Hybrid Graph Neural Network Framework for Enhanced Medical Image Segmentation. CompSci & AI Advances, 3(2), 118–131. https://doi.org/10.69626/cai.2026.0118
Number of downloads: 9
Number of views: 99