Chest X-ray Image Super-Resolution Using Artificial Intelligence

Abstract

Chest X-ray (CXR) imaging is essential for diagnosing respiratory diseases like pneumonia, but low-resolution (LR) images can make detection less accurate. This study explores how deep learning-based super-resolution (SR) techniques can improve CXR image quality and enhance automated pneumonia detection. Real-ESRNet utilized as the generator to restore image details and Real-ESRGAN NetD as the discriminator to refine structures. The model was fine-tuned on 4,500 images, including original and multi-scale variations, randomly selected from the Random sample of NIH Chest X-ray Dataset, which contains 5,606 labeled images. Training was conducted with a batch size of 10 per GPU for four epochs, totalling 3,240 iterations, with checkpoints saved every 810 iterations. To evaluate the impact of SR, The enhanced images were tested using a classification model based on SqueezeNet embeddings and a neural network in Orange Data Mining software. Results showed that SR-improved images led to better pneumonia detection accuracy compared to LR images, though further optimization is needed for better reconstruction quality. These findings highlight how AI-driven super-resolution can play a significant role in improving medical imaging and automated disease diagnosis.

Introduction

Summary:

The research addresses the challenge of low-resolution (LR) X-ray imaging, which impairs accurate diagnosis of diseases like pneumonia due to loss of fine details and edge blurring. To improve diagnostic accuracy, the study evaluates the impact of AI-based super-resolution (SR) techniques on chest X-ray (CXR) images. The approach fine-tunes a super-resolution model using Real-ESRNet as a generator and Real-ESRGAN NetD as a discriminator to enhance four-times downscaled LR images. These enhanced images are then used for pneumonia classification via SqueezeNet embeddings and a neural network implemented in Orange Data Mining software.

The methodology involves fine-tuning the SR model on 4,500 diverse images, upscaling LR images to higher resolution, and comparing pneumonia classification results between LR and SR datasets using metrics such as accuracy, PSNR, SSIM, MSE, and confusion matrices. Real-ESRGAN, an improvement over ESRGAN, is chosen for its ability to handle real-world degradations better, producing stable and high-fidelity reconstructions suitable for clinical use.

The classification model employs SqueezeNet for efficient feature extraction due to its compact size and speed, making it suitable for real-time medical applications with resource constraints. The pneumonia detection neural network uses fully connected layers with ReLU activation, trained separately on LR and SR embeddings to assess performance improvements due to super-resolution.

Overall, the study demonstrates that AI-driven super-resolution enhances CXR image quality, thereby improving automatic pneumonia detection accuracy and potentially reducing healthcare costs.

Conclusion

This study explored super-resolution for pneumonia detection in chest X-rays, showing accuracy improvement over low-resolution images on same training dataset and also for completely unseen dataset. The SR model demonstrated enhanced perceptual quality and structural details. The super-resolved (SR) images improved classification accuracy compared to low-resolution (LR) images, with the SR-based pneumonia detection model achieving higher accuracy. These findings highlight the potential of AI-driven super-resolution in medical imaging and enabling cost-effective diagnostics

References

[1] Yue, L., Shen, H., Li, J., Yuan, Q., Zhang, H., & Zhang, L. (2016). Image super-resolution: The techniques, applications, and future. Signal Processing, 128, 389–408. https://doi.org/10.1016/j.sigpro.2016.05.002 [2] Fritsche, M., Gu, S., & Timofte, R. (2019). Frequency separation for real-world super-resolution. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (pp. 3599–3608). [3] Nijhawan, R., Verma, R., Bhushan, S., Dua, R., & Mittal, A. (2017). An integrated deep learning framework approach for nail disease identification. In 2017 13th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS) (pp. 197–202). IEEE. https://doi.org/10.1109/SITIS.2017.41 [4] C. Ledig, L. Theis, F. Huszár, J. Caballero, A. Cunningham, A. Acosta, A. Aitken, A. Tejani, J. Totz, Z. Wang, et al., \"Photorealistic single image super-resolution using a generative adversarial network,\" arXiv preprint, 2016. [5] Y. Zhang, Y. Tian, Y. Kong, B. Zhong, Y. Fu, \"Residual dense network for image super-resolution,\" in The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018. [6] Dawit Kiros Redie, Abdulhakim Edao Sirko, Tensaie Melkamu Demissie, Semagn Sisay Teferi, Vimal Kumar Shrivastava, Om Prakash Verma, Tarun Kumar Sharma, Diagnosis of COVID-19 using chest X-ray images based on modified DarkCovidNet model, Evol. Intell. 16 (3) (2023) 729–738. [7] Yibo Feng, Xu Yang, Dawei Qiu, Huan Zhang, Dejian Wei, Jing Liu, Pcxrnet: pneumonia diagnosis from chest x-ray images using condense attention block and multiconvolution attention block, IEEE J. Biomed. Health Inform. 26 (4) (2022) 1484–1495. [8] Y. Li, Bruno Sixou, F. Peyrin, A review of the deep learning methods for medical images super-resolution problems, Irbm 42 (2) (2021) 120–133. [9] Jing Tian, Kai-Kuang Ma, A survey on super-resolution imaging, Signal. Image Video Process. 5 (2011) 329–342. [10] Hujun Yang, Zhongyang Wang, Xinyao Liu, Chuangang Li, Junchang Xin, Zhiqiong Wang, Deep learning in medical image super resolution: a review, Appl. Intell. (2023) 1–26. [11] I. Sirazitdinov, M. Kholiavchenko, T. Mustafaev, Y. Yixuan, R. Kuleev, and B. Ibragimov, \"Deep neural network ensemble for pneumonia localization from a large-scale chest x-ray database,\" Computers & Electrical Engineering, vol. 78, pp. 388-399, 2019. [12] K. Hammoudi et al., \"Deep Learning on Chest X-ray Images to Detect and Evaluate Pneumonia Cases at the Era of COVID-19,\" arXiv preprint arXiv:2004.03399, 2020. [13] A. K. Jaiswal, P. Tiwari, S. Kumar, D. Gupta, A. Khanna, and J. J. Rodrigues, \"Identtifying pneumonia in chest X-rays: A deep learning approach,\" Measurement, vol. 145, pp. 511-518, 2019. [14] C. Dong, C. C. Loy, K. He, and X. Tang, “Image super-resolution using deep convolutional networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, no. 2, pp. 295–307, Feb. 2016, doi: 10.1109/TPAMI.2015.2439281. [15] L. Xu, X. Zeng, Z. Huang, W. Li, and H. Zhang, “Low-dose chest X-ray image super-resolution using generative adversarial nets with spectral normalization,” Biomedical Signal Processing and Control, vol. 55, p. 101600, Jan. 2020, doi: 10.1016/j.bspc.2019.101600. [16] A. Khishigdelger, A. Salem, and H. S. Kang, “Elevating chest X-ray image super-resolution with residual network enhancement,” Journal of Imaging, vol. 10, no. 3, p. 64, Mar. 2024, doi: 10.3390/jimaging10030064. [17] K. Ahmadian and H. R. Alikhani, “X-ray medical image super-resolution via self-organization neural networks and geometric directional gradient,” IET Image Processing, vol. 16, no. 14, pp. 3910–3928, Dec. 2022, doi: 10.1049/ipr2.12603. [18] A. Mittal et al., “Detecting pneumonia using convolutions and dynamic capsule routing for chest X-ray images,” Sensors (Switzerland), vol. 20, no. 4, pp. 1–30, 2020, doi: 10.3390/s20041068. [19] B. Imran and L. D. Bakti, “Implementation of Machine Learning Model for Pneumonia Classification Based on X-Ray Images,” J. Mantik, vol. 5, no. 3, pp. 2101–2107, 2021. [20] K. R. Swetha, M. Niranjanamurthy, M. P. Amulya, and M. Y. Manu, “Prediction of Pneumonia Using Big Data, Deep Learning and Machine Learning Techniques,” Proc. 6th Int. Conf. Commun. Electron. Syst. ICCES 2021, no. August, pp. 1697–1700, 2021, doi: 10.1109/ICCES51350.2021.9489188. [21] X. Wang, L. Xie, C. Dong, and Y. Shan, “Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data,” arXiv preprint arXiv:2107.10833, 2021. Available: https://arxiv.org/abs/2107.10833. [22] C. Dong, C. C. Loy, K. He, and X. Tang, “Image super-resolution using deep convolutional networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, no. 2, pp. 295–307, 2016. doi: 10.1109/TPAMI.2015.2439281. [23] X. Wang, Y. Yu, K. Gu, S. Wu, J. Gu, C. Dong, Y. Qiao, and C. C. Loy, “ESRGAN: Enhanced super-resolution generative adversarial networks,” in Proceedings of the European Conference on Computer Vision (ECCV) Workshops, 2018. Available: https://arxiv.org/abs/1809.00219. [24] Moneebah. (n.d.). GitHub - Moneebah/Finetuning_Real-ESRGAN_Pneumonia: Finetuning pretrained model of Real-ESRGAN onto chest xrays (with and without pneumonia). GitHub. https://github.com/Moneebah/Finetuning_Real-ESRGAN_Pneumonia [25] Random sample of NIH Chest X-ray Dataset. (2017, November 23). Kaggle. https://www.kaggle.com/datasets/nih-chest-xrays/sample [26] Chest X-Ray images (Pneumonia). (2018, March 24). Kaggle. https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia

Copyright

Copyright © 2025 Ayush Patel, Prof. (Dr.) Tejas Shah. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.