Deep Learning-Based Generative Al for Segmenting Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning

G. Indumathi, Sudharsan T. S., Ashok I., Tharunkumar M.

2025

Abstract

Segmenting necrotic lung lesions in CT images plays a vital role in diagnosing and managing pulmonary diseases. However, traditional methods often struggle with the complex shapes and varying appearances of lesions while relying heavily on manually annotated datasets. To overcome these limitations, we introduce an innovative framework that combines Variational Autoencoders (VAEs) and self-supervised contrastive learning for more accurate and efficient segmentation. The VAE helps the model learn compact and meaningful representations of CT images, while contrastive pretraining enhances these features using unlabeled data, improving generalization across different datasets. This approach not only reduces dependency on manual annotations but also excels in capturing fine lesion boundaries and handling diverse lesion appearances. By advancing medical image segmentation, our method provides a robust, scalable, and efficient solution to key clinical challenges, ultimately aiding in early detection and treatment planning for lung diseases.

Paper Citation

in Harvard Style

Indumathi G., S. S., I. A. and M. T. (2025). Deep Learning-Based Generative Al for Segmenting Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 512-518. DOI: 10.5220/0013932200004919

in Bibtex Style

@conference{icrdicct`2525,
author={G. Indumathi and Sudharsan S. and Ashok I. and Tharunkumar M.},
title={Deep Learning-Based Generative Al for Segmenting Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={512-518},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013932200004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - Deep Learning-Based Generative Al for Segmenting Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning
SN - 978-989-758-777-1
AU - Indumathi G.
AU - S. S.
AU - I. A.
AU - M. T.
PY - 2025
SP - 512
EP - 518
DO - 10.5220/0013932200004919
PB - SciTePress