Explainable benchmarking of U-Net variants for lung segmentation on chest radiographs
Tuberk Toraks · 25.06.2026 · 74 (2) · 84-97
DOI: 10.5578/tt.2026021210 ↗Kurz gefasst
Fachartikel in der Zeitschrift Tuberk Toraks, online erschienen am 26.06.2026. Verfasst von Khalid Hani Abushahla und 1 weiteren Autor:innen. Beteiligte Einrichtung: Biomedical Technologies Application and Research Center (BİYOTAM), Sakarya University of Applied Sciences, Sakarya, Türkiye. Hauptthemen laut MEDLINE: Radiography, Thoracic, Lung, Lung Diseases, Radiographic Image Interpretation, Computer-Assisted.
Abstract (Originalsprache)
Introduction
Accurate lung field segmentation in chest radiographs is essential for reliable computer-aided diagnosis of pulmonary diseases. This study aimed to systematically benchmark baseline, lightweight, and attention-based U-Net architectures while integrating explainable artificial intelligence (AI) to evaluate both segmentation performance and anatomical focus. Materials and Methods: We conducted a comparative evaluation of three U-Net variants-baseline U-Net, attention U-Net, and shallow U-Net-using the chest X-ray Masks and Labels dataset. All models were trained under identical conditions with five-fold cross-validation and evaluated using accuracy, Intersection over Union (IoU), and Dice coefficient. Gradient-weighted Class Activation Mapping (Grad-CAM) was applied to visualize model attention and assess whether network activations were anatomically localized within lung regions. All architectures were implemented in TensorFlow.
Result
U-Net and attention U-Net achieved the highest segmentation performance (Dice ≈ 0.97, IoU ≈ 0.94), with Grad-CAM activations consistently localized to lung fields, indicating reliable anatomical focus. The shallow U-Net showed slightly lower accuracy (Dice= 0.96, IoU= 0.92) but demonstrated faster inference and broader sensitivity to internal parenchymal structures, which may facilitate future disease-focused pulmonary analysis.
Conclusions
This study highlights the trade-offs between segmentation accuracy, computational efficiency, and model interpretability across U-Net variants. By combining quantitative benchmarking with explainable AI-based analysis, our results provide performance insights, supporting the development of trustworthy AI tools for chest radiograph analysis.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Biomedical Technologies Application and Research Center (BİYOTAM), Sakarya University of Applied Sciences, Sakarya, Türkiye.
Zitieren
APA
Abushahla, K. H., & Arslan, H. (2026). Explainable benchmarking of U-Net variants for lung segmentation on chest radiographs. Tuberk Toraks, 74(2), 84-97. https://doi.org/10.5578/tt.2026021210
Vancouver
Abushahla KH, Arslan H. Explainable benchmarking of U-Net variants for lung segmentation on chest radiographs. Tuberk Toraks. 2026;74(2):84-97. doi:10.5578/tt.2026021210
BibTeX
@article{Abushahla2026,
author = {Abushahla, Khalid Hani and Arslan, Halil},
title = {Explainable benchmarking of U-Net variants for lung segmentation on chest radiographs},
journal = {Tuberk Toraks},
year = {2026},
volume = {74},
number = {2},
pages = {84--97},
doi = {10.5578/tt.2026021210}
}
EUVIMED