An Empirical Analysis of Anchor Optimization and RPN Sampling  in Faster R-CNN for Pleural Effusion Detection

Muhammad Jafar Sakone; Izzati Muhimmah

doi:10.24252/instek.v11i2.72491

Authors

Muhammad Jafar Sakone Universitas Islam Indonesia
Izzati Muhimmah Universitas Islam Indonesia

DOI:

https://doi.org/10.24252/instek.v11i2.72491

Keywords:

anchor optimization, Faster R-CNN, object detection, pleural effusion, RPN sampling

Abstract

Pleural effusion detection from chest X-ray images remains challenging due to variations in lesion scale and severe foreground-background imbalance within two-stage object detectors. This study empirically investigates the effects of anchor representation and Region Proposal Network (RPN) sampling on the detection process of Faster R-CNN for pleural effusion detection. A Faster R-CNN with ResNet-50 and Feature Pyramid Network (FPN) was evaluated using representative anchors derived from ground-truth object characteristics and adjusted RPN Intersection over Union (IoU) assignment thresholds. The analysis examined anchor representation, RPN sample distribution, proposal quality, detection confidence, and stage-wise detection outcomes. Representative anchors improved geometric correspondence with the target objects, while RPN threshold adjustment substantially increased positive RPN samples from approximately 4 to 110 per image. Proposal analysis showed high spatial accessibility, with proposal IoU coverage at IoU ≥ 0.50 reaching up to 99.35% across the evaluated configurations. However, stage-wise analysis showed that classification was the dominant source of detection failure, accounting for 15.03–16.99% of the evaluated objects, while proposal and localization failures remained comparatively limited. These findings demonstrate that optimization in anchor representation and RPN sampling can enhance upstream proposal characteristics, but such improvements do not necessarily translate proportionally into final detection performance.

Downloads

Download data is not yet available.

References

[1] E. Weldon and J. Williams, "Pleural Disease in the Emergency Department," Emerg. Med. Clin. North Am., vol. 30, no. 2, pp. 475-499, May 2012, doi: 10.1016/J.EMC.2011.10.012.

[2] D. Bates, N. Yang, M. Bailey, and R. Bellomo, "Prevalence, characteristics, drainage and outcome of radiologically diagnosed pleural effusions in critically ill patients," Critical Care and Resuscitation, vol. 22, no. 1, pp. 45-52, Mar. 2020, doi: 10.51893/2020.1.OA5.

[3] E. DeBiasi and J. Puchalski, "Pleural effusions as markers of mortality and disease severity: A state-of-the-art review," Curr. Opin. Pulm. Med., vol. 22, no. 4, pp. 386-391, Jul. 2016, doi: 10.1097/MCP.0000000000000278.

[4] C. A. Aprilia, R. D. Aulia, A. E. Wibowo, E. N. Permatasari, S. Bakhri, and Abdurrahman, "View of Massive pleural effusion due to heart failure in an undetected diabetic hypertensive patient: A case report." Accessed: Dec. 18, 2025. [Online]. Available: https://journal.uii.ac.id/JKKI/article/view/19749/pdf

[5] F. Karatas and O. Dikensoy, "Artificial Intelligence in Pleural Diseases: Current Applications and Next Steps," Jan. 01, 2026, AVES. doi: 10.4274/ThoracResPract.2025.2025-6-2.

[6] Y. Y. Huang et al., "Automated Computer-Assisted Diagnosis of Pleural Effusion in Chest X-Rays via Deep Learning," Diagnostics, vol. 15, no. 18, Sep. 2025, doi: 10.3390/diagnostics15182322.

[7] D. C. Lepcha et al., "Deep Learning in Medical Image Analysis: A Comprehensive Review of Algorithms, Trends, Applications, and Challenges," Computer Modeling in Engineering & Sciences, vol. 145, no. 2, pp. 1487-1573, Nov. 2025, doi: 10.32604/CMES.2025.070964.

[8] S. Matta et al., "A systematic review of generalization research in medical image classification," Dec. 01, 2024, Elsevier Ltd. doi: 10.1016/j.compbiomed.2024.109256.

[9] Y. Zhong, J. Wang, J. Peng, and L. Zhang, "Anchor Box Optimization for Object Detection," Jan. 2020, [Online]. Available: http://arxiv.org/abs/1812.00469

[10] W. Ma, T. Tian, H. Xu, Y. Huang, and Z. Li, "AABO: Adaptive Anchor Box Optimization for Object Detection via Bayesian Sub-sampling," Jul. 2020, [Online]. Available: http://arxiv.org/abs/2007.09336

[11] D. Miller, G. Goode, C. Bennie, P. Moghadam, and R. Jurdak, "Why Object Detectors Fail: Investigating the Influence of the Dataset."

[12] S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks," Jan. 2016, Accessed: Dec. 08, 2025. [Online]. Available: http://image-net.org/challenges/LSVRC/2015/results

[13] J. Wang, K. Chen, S. Yang, C. C. Loy, and D. Lin, "Region Proposal by Guided Anchoring," Apr. 2019, [Online]. Available: http://arxiv.org/abs/1901.03278

[14] L. Zhu, Z. Xie, L. Liu, B. Tao, and W. Tao, "IoU-uniform R-CNN: Breaking Through the Limitations of RPN," Dec. 2019, [Online]. Available: http://arxiv.org/abs/1912.05190

[15] T. Vu, H. Jang, T. X. Pham, and C. D. Yoo, "Cascade RPN: Delving into High-Quality Region Proposal Network with Adaptive Convolution," Dec. 2019, [Online]. Available: http://arxiv.org/abs/1909.06720

[16] Y. Zhao et al., "Pulmonary Nodule Detection Based on Multiscale Feature Fusion," Comput. Math. Methods Med., vol. 2022, 2022, doi: 10.1155/2022/8903037.

[17] S. Yao, Y. Chen, X. Tian, and R. Jiang, "Pneumonia Detection Using an Improved Algorithm Based on Faster R-CNN," Comput. Math. Methods Med., vol. 2021, 2021, doi: 10.1155/2021/8854892.

[18] W. Fan et al., "A deep-learning-based framework for identifying and localizing multiple abnormalities and assessing cardiomegaly in chest X-ray," Nature Communications , vol. 15, no. 1, Dec. 2024, doi: 10.1038/s41467-024-45599-z.

[19] A. Nogueira-Rodriguez, D. Glez-Pena, M. Reboiro-Jato, and H. Lopez-Fernandez, "Negative Samples for Improving Object Detection-A Case Study in AI-Assisted Colonoscopy for Polyp Detection," Diagnostics, vol. 13, no. 5, Mar. 2023, doi: 10.3390/diagnostics13050966.

[20] D. Miller, G. Goode, C. Bennie, P. Moghadam, and R. Jurdak, "Why Object Detectors Fail: Investigating the Influence of the Dataset."

[21] J. Huang et al., "Speed/accuracy trade-offs for modern convolutional object detectors," Apr. 2017, [Online]. Available: http://arxiv.org/abs/1611.10012

[22] X. Wang, Y. Peng, L. Lu, Z. Lu, M. Bagheri, and R. M. Summers, "ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases," Dec. 2017, doi: 10.1109/CVPR.2017.369.

[23] Y. Wu, A. Kirillov, F. Massa, W.-Y. Lo, and R. Girshick, "Detectron2," 2019, Facebook AI Research: 0.6.

[24] K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 2016-December, pp. 770-778, Dec. 2015, doi: 10.1109/CVPR.2016.90.

[25] T.-Y. Lin, P. Dollar, R. Girshick, K. He, B. Hariharan, and S. Belongie, "Feature Pyramid Networks for Object Detection," Apr. 2017.

[26] R. Faisal, R. A. Kambau, and M. H. H, "Perhitungan Otomatis Ukuran Udang Vaname Menerapkan Object Detection dengan Convolutional Neural Network Menggunakan Metode Bidirectional Feature Pyramid Network," INSTEK, no. Volume 9, Apr. 2024, Accessed: Sep. 05, 2026. [Online]. Available: https://doi.org/10.24252/instek.v9i1.46420