A Comparative Study of ResNet and SE-ResNet Architectures on Medical Image Datasets
Abstract views: 10
/ 
PDF downloads: 4
Keywords:
Medical Image Synthesis, Deep Learning, ResNet, Squeeze-and-ExcitationAbstract
In this study, we investigate the effectiveness of different deep learning architectures in the task
of medical image synthesis using convolutional neural networks. Our goal is to compare the performance
of standard ResNet architectures (ResNet-18 and ResNet-50) with their Squeeze-and-Excitation (SE)
enhanced counterparts (SE-ResNet-18 and SE-ResNet-50). The evaluation is conducted on three publicly
available medical datasets: CVC-ClinicDB (colorectal polyp images), Messidor2 (retinal images), and Pap
Smear (cervical cell images). For image synthesis, we employ these architectures as generative backbones
and assess the quality of the generated images using both pixel-level metrics Mean Squared Error (MSE)
and perceptual similarity metrics, namely Fréchet Inception Distance (FID) and Kernel Inception Distance
(KID). Experimental results demonstrate that SE-enhanced ResNet architectures outperform their vanilla
counterparts in generating more realistic and perceptually coherent images. Particularly, SE-ResNet-50
achieves the lowest FID and KID scores across all datasets, indicating superior generative quality. These
findings highlight the impact of channel-wise attention mechanisms in enhancing feature representation
and improving medical image synthesis tasks. Experimental results demonstrate that ResNet50 achieves
the best performance across multiple metrics, including LPIPS, FID, KID, and MSE, confirming its
superiority in both perceptual quality and pixel-level accuracy.
Downloads
References
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770–778.
Hu, J., Shen, L., & Sun, G. (2018). Squeeze-and-Excitation Networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 7132–7141.
Bernal, J., Sánchez, J., & Vilarino, F. (2015). WM-DOVA maps for accurate polyp highlighting in colonoscopy. IEEE Transactions on Medical Imaging, 34(8), 1724–1737.
Decencière, E., Cazuguel, G., Zhang, X., Lay, B., Cochener, B., Trone, C., & Massin, P. (2014). Feedback on a publicly distributed database: the Messidor database. Image Analysis & Stereology, 33(3), 231–234.
Altun, S., & Talu, M. F. (2022). A new approach for Pap-Smear image generation with generative adversarial networks. Journal of the Faculty of Engineering and Architecture of Gazi University, 37(3), 1401–1410.
Zhang, Z., Shen, Y., Xiao, J., Zhang, X., & Li, S. (2018). Image quality assessment: From error visibility to structural similarity. IEEE Transactions on Image Processing, 27(7), 2920–2934.
Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B., & Hochreiter, S. (2017). GANs trained by a two time-scale update rule converge to a Nash equilibrium. Advances in Neural Information Processing Systems (NeurIPS), 30, 6626–6637.
Binkowski, M., Sutherland, D. J., Arbel, M., & Gretton, A. (2018). Demystifying MMD GANs. International Conference on Learning Representations (ICLR).
Zhang, R., Isola, P., Efros, A. A., Shechtman, E., & Wang, O. (2018). The unreasonable effectiveness of deep features as a perceptual metric. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 586–595.
Ovalle-Magallanes, E., Silva, R., Silva, L., Pérez, M., & Rodríguez, J. (2022). Efficient SE-ResNet for coronary angiography classification. Electronics, 11(21), 3570.
Zhang, X., Wang, J., Li, R., Chen, Q., & Huang, Y. (2020). DeepSEED: Deep SE-ResNet for lung nodule detection. Scientific Reports, 10, 15320.
Wang, Y., Liu, C., Li, W., & Yang, H. (2021). CASE-Net: Fetal MRI segmentation with SE blocks and cross-attention. Sensors, 21(13), 4490.
Huang, Y., Zhang, W., Li, M., Yang, F., & Chen, J. (2025). SE-ResNet-50V2 for brain tumor classification on Kaggle dataset. Journal of Medical Imaging and Health Informatics, 15(5), 123–130.
Kadri, A., Perez, M., Santos, L., Chen, G., & Villanueva, O. (2021). CrossViT + Wide ResNet + SE for Alzheimer diagnosis. Health Informatics Journal, 27(3), 1–12.
Kim, D., Park, E., Lee, S., & Choi, M. (2021). LRSE-Net: Lightweight SE-ResNet for patch-based medical imaging. Electronics, 11(21), 3570.
Smith, J., González, M., Patel, V., & Lin, A. (2024). Retinopathy classification with Swish-ResNet-18. Journal of Ophthalmic Machine Learning, 4(2), 45–52.
Brown, A., Nguyen, T., & Roberts, M. (2024). Evaluating ResNet-18 for surgical need prediction in radiographs. Frontiers in Radiology, 8, Article 112.
Chen, L., Wang, W., Zhang, J., & Xu, M. (2022). SERNet: SE-enhanced residual networks for remote sensing segmentation. Remote Sensing, 14(19), 4770.
Yeung, M., Lee, A., Wu, K., Chen, D., & Tan, S. (2021). Focus U-Net for polyp segmentation. arXiv preprint, arXiv:2105.07467.
Fitzgerald, R., & Matuszewski, B. (2023). FCB-SwinV2: A hybrid CNN-Transformer model for polyp segmentation. arXiv preprint, arXiv:2302.01027.
Abd El-Hafez, T., Mohamed, A., & Ali, S. (2022). Improved retinopathy detection using CNNs. Middle East Journal of Engineering & Environmental Research, 5(2), 115–125.
Merlina, F., Laurent, S., Orozco, M., & Svensson, E. (2024). Multi-class Pap Smear classification with transfer learning. Journal of Advanced Diagnostics, 12(4), 200–210.
Liu, W., Zhang, Q., Liu, Y., & Sun, H. (2022). CVM-Cervix: A CNN-Transformer-MLP hybrid model for cytology. arXiv preprint, arXiv:2206.00971.
