Investigation of Deep Learning Architectures for Lesion Segmentation in Histological Images

• Domingos L. L. de Oliveira IFSP / UFU
• Adriano B. Silva UFU
• Leandro A. Neves UNESP
• Thaína A. A. Tosta UNIFESP
• Alessandro S. Martins IFMT
• Paulo R. de Faria UFU
• Marcelo Z. do Nascimento UFU

DOI: https://doi.org/10.5753/sbcas.2025.7688

Abstract

This study evaluates the performance of convolutional neural networks (CNN)-based models and Transformers for histological image segmentation using public datasets: OCDC and GlaS. Four models were analyzed: UNet, UNet++, SharpUNet, and TransUNet, with and without data augmentation. Data augmentation proved crucial for improving model generalization, with SharpUNet and UNet++ achieving the best results in Dice coefficient and accuracy. TransUNet, despite its hybrid architecture, underperformed, possibly due to its complexity and need for large training data. The results indicate that CNN-based models, such as UNet++ and SharpUNet, are effective for histological image segmentation, especially when combined with data augmentation.

References

Banerji, S. and Mitra, S. (2022). Deep learning in histopathology: A review. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 12(1):e1439.

Basu, A., Senapati, P., Deb, M., Rai, R., and Dhal, K. G. (2024). A survey on recent trends in deep learning for nucleus segmentation from histopathology images. Evolving Systems, 15(1):203–248.

Chen, J., Lu, Y., Yu, Q., Luo, X., Adeli, E., Wang, Y., Lu, L., Yuille, A. L., and Zhou, Y. (2021). Transunet: Transformers make strong encoders for medical image segmentation. CoRR, abs/2102.04306.

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., and Houlsby, N. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. CoRR, abs/2010.11929.

Maurício, J., Domingues, I., and Bernardino, J. (2023). Comparing vision transformers and convolutional neural networks for image classification: A literature review. Applied Sciences, 13(9):5521.

Moscalu, M., Moscalu, R., Dascălu, C. G., T, arcă, V., Cojocaru, E., Costin, I. M., T, arcă, E., and S, erban, I. L. (2023). Histopathological images analysis and predictive modeling implemented in digital pathology—current affairs and perspectives. Diagnostics, 13(14):2379.

Ronneberger, O., Fischer, P., and Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, pages 234–241. Springer.

Santos, D. F., de Faria, P. R., Travencolo, B. A., and do Nascimento, M. Z. (2021). Automated detection of tumor regions from oral histological whole slide images using fully convolutional neural networks. Biomedical Signal Processing and Control, 69:102921.

Santos, D. F., de Faria, P. R., Travençolo, B. A., and do Nascimento, M. Z. (2023a). Influence of data augmentation strategies on the segmentation of oral histological images using fully convolutional neural networks. Journal of Digital Imaging, 36(4):1608–1623.

Santos, D. F. D., de Faria, P. R., Loyola, A. M., Cardoso, S. V., Travençolo, B. A. N., and do Nascimento, M. Z. (2023b). Hematoxylin and eosin stained oral squamous cell carcinoma histological images dataset.

Silva, A. B., Martins, A. S., Tosta, T. A. A., Neves, L. A., Servato, J. P. S., de Araújo, M. S., de Faria, P. R., and do Nascimento, M. Z. (2022). Computational analysis of histological images from hematoxylin and eosin-stained oral epithelial dysplasia tissue sections. Expert Systems with Applications, 193:116456.

Silva, A. B., Rozendo, G. B., Tosta, T. A., Martins, A. S., Loyola, A. M., Cardoso, S. V., Lumini, A., Neves, L. A., de Faria, P. R., and do Nascimento, M. Z. (2023). Cnn ensembles for nuclei segmentation on histological images of oed. In 2023 IEEE 36th International Symposium on Computer-Based Medical Systems (CBMS), pages 601–604. IEEE.

Sirinukunwattana, K., Pluim, J. P., Chen, H., Qi, X., Heng, P.-A., Guo, Y. B., Wang, L. Y., Matuszewski, B. J., Bruni, E., Sanchez, U., et al. (2017). Gland segmentation in colon histology images: The glas challenge contest. Medical image analysis, 35:489–502.

Springenberg, M., Frommholz, A., Wenzel, M., Weicken, E., Ma, J., and Strodthoff, N. (2023). From modern cnns to vision transformers: Assessing the performance, robustness, and classification strategies of deep learning models in histopathology. Medical Image Analysis, 87:102809.

Srinidhi, C. L., Ciga, O., and Martel, A. L. (2021). Deep neural network models for computational histopathology: A survey. Medical image analysis, 67:101813.

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., and Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.

Xu, H., Xu, Q., Cong, F., Kang, J., Han, C., Liu, Z., Madabhushi, A., and Lu, C. (2023). Vision transformers for computational histopathology. IEEE Reviews in Biomedical Engineering.

Xu, Y., Quan, R., Xu, W., Huang, Y., Chen, X., and Liu, F. (2024). Advances in medical image segmentation: a comprehensive review of traditional, deep learning and hybrid approaches. Bioengineering, 11(10):1034.

Yuan, Y. and Cheng, Y. (2024). Medical image segmentation with unet-based multi-scale context fusion. Scientific Reports, 14(1):15687.

Zhou, Z., Rahman Siddiquee, M. M., Tajbakhsh, N., and Liang, J. (2018). Unet++: A nested u-net architecture for medical image segmentation. In Deep learning in medical image analysis and multimodal learning for clinical decision support: 4th international workshop, DLMIA 2018, and 8th international workshop, ML-CDS 2018, held in conjunction with MICCAI 2018, Granada, Spain, September 20, 2018, proceedings 4, pages 3–11. Springer.

Zunair, H. and Hamza, A. B. (2021). Sharp u-net: Depthwise convolutional network for biomedical image segmentation. Computers in biology and medicine, 136:104699.