Multi-encoder self-adaptive hard attention network with maximum intensity projections for lung nodule segmentation
Usman, Muhammad Rehman, Azka Ur Rehman, Abd Shahid, Abdullah Khan, Tariq Mahmood Razzak, Muhammad Imran Chung, Minyoung
Usman, Muhammad
Rehman, Azka
Ur Rehman, Abd
Shahid, Abdullah
Khan, Tariq Mahmood
Razzak, Muhammad Imran
Chung, Minyoung
Supervisor
Department
Computational Biology
Embargo End Date
Type
Journal article
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Accurate lung nodule segmentation is crucial for early-stage lung cancer diagnosis, as it can substantially enhance patient survival rates. Computed tomography (CT) images are widely employed for early diagnosis in lung nodule analysis. However, the heterogeneity of lung nodules, size diversity, and the complexity of the surrounding environment pose challenges for developing robust nodule segmentation methods. In this study, we propose an efficient end-to-end framework, the Multi-Encoder Self-Adaptive Hard Attention Network (MESAHA-Net), which consists of three encoding paths, an attention block, and a decoder block that assimilates CT slice patches with both forward and backward maximum intensity projection (MIP) images. This synergy affords a profound contextual understanding of lung nodules and also results in a deluge of features. To manage the profusion of features generated, we incorporate a self-adaptive hard attention mechanism guided by region of interest (ROI) masks centered on nodular regions, which MESAHA-Net autonomously produces. The network sequentially undertakes slice-by-slice segmentation, emphasizing nodule regions to produce precise three-dimensional (3D) segmentation. The proposed framework has been comprehensively evaluated on the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset, the largest publicly available dataset for lung nodule segmentation. The results demonstrate that our approach is highly robust across various lung nodule types, outperforming previous state-of-the-art techniques in terms of segmentation performance and computational complexity, making it suitable for real-time clinical implementation of artificial intelligence (AI)-driven diagnostic tools.
Citation
M. Usman et al., “Multi-encoder self-adaptive hard attention network with maximum intensity projections for lung nodule segmentation,” Comput Biol Med, vol. 197, p. 111059, Oct. 2025, doi: 10.1016/J.COMPBIOMED.2025.111059.
Source
Computers in Biology and Medicine
Conference
Keywords
Adaptive Hard-attention, Bidirectional Maximum Intensity Projection, Lung Nodule Segmentation, Multi-encoder Architecture, Biological Organs, Computerized Tomography, Diagnosis, Lung Cancer, Signal Encoding, Adaptive Hard-attention, Bidirectional Maximum Intensity Projection, Early-stage Lung Cancers, Encoder Architecture, Image Database, Lung Cancer Diagnosis, Lung Nodule, Lung Nodule Segmentation, Maximum Intensity Projection, Multi-encoder Architecture, Image Segmentation
Subjects
Source
Publisher
Elsevier