Bregman Divergences applied to Content-Based Image Retrieval of Oral Dysplasia

• Tiago R. M. Soares UFU
• Adriano B. Silva UFU
• Adriano M. Loyola UFU
• Sérgio V. Cardoso UFU
• Paulo R. de Faria UFU
• Leandro A. Neves UNESP
• Marcelo Z. Nascimento UFU
• Humberto Razente UFU

DOI: https://doi.org/10.5753/sbcas.2024.2058

Abstract

In recent years, several works have employed non-metric functions to deal with the semantic gap between query results and users’ perception of similarity in Content-Based Image Retrieval (CBIR) systems. In this study, we investigated the search of images of oral cavity dysplasias, obtained from histological slides that contained induced lesions of mice of the C57Bl/6 lineage. The images were segmented with a method based on the Mask R-CNN neural network to extract morphological and non-morphological descriptors. Bregman divergences (Kullback-Leibler and Mahalanobis) and metrics (Euclidean and Manhattan) were employed in searches, which were evaluated for precision and recall. Bregman divergences proved to be more effective in identifying the dysplasia levels.

References

Banerjee, A., Merugu, S., Dhillon, I. S., and Ghosh, J. (2004). Clustering with bregman divergences. In SIAM Int’l’ Conf. Data Mining, pages 234–245. SIAM.

Faria, P. R., Chammas, R., de Melo, T. L., Hsu, D. K., Liu, F.-T., Nonogaki, S., Cardoso, S. V., and Loyola, A. M. (2011). Absence of galectin-3 does not affect the development of experimental tongue carcinomas in mice. Exp. Mol. Pathol., 90(2):189–193.

Gonzalez, R. C. and Woods, R. C. (2009). Processamento digital de imagens . Pearson.

He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. In IEEE Conf. Comp. Vision and Pat. Recog. (CVPR), pages 770–778.

Junior, J. R. F., Santos, M. K., and Azevedo-Marques, P. M. (2020). Arcabouço para classificação, recuperação por conteúdo e radiômica de imagens médicas: uma investigação de biomarcadores quantitativos para o câncer de pulmão. In Anais Estendidos do XX Simpósio Bras. Computação Aplicada à Saúde, pages 13–18. SBC.

Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2017). Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60(6):84–90.

Liu, M., Vemuri, B. C., Amari, S.-I., and Nielsen, F. (2012). Shape retrieval using hierarchical total bregman soft clustering. IEEE TPAMI, 34(12):2407–2419.

Rocha, B. M., Ferreira, D. P. L., Barcelos, C. A. Z., and Soares, F. (2021). I-generalized and kullback-leibler divergences for content-based image retrieval. In 34th IEEE Canadian Conf. Elect. and Comp. Eng. (CCECE), pages 1–4. IEEE.

Santini, S. and Jain, R. C. (1999). Similarity measures. IEEE TPAMI, 21(9):871–883.

Santos, M. d. O., Lima, F. C. d. S. d., Martins, L. F. L., Oliveira, J. F. P., Almeida, L. M., and Cancela, M. d. C. (2023). Estimativa de incidência de câncer no brasil, 2023-2025. Rev. Bras. Cancerol., 69(1):1–12.

Silva, A., Oliveira, C., Pereira, D., Tosta, T., Martins, A., Loyola, A., Cardoso, S., Faria, P., Neves, L., and Nascimento, M. (2022). Assessment of the association of deep features with a polynomial algorithm for automated oral epithelial dysplasia grading. In SIBGRAPI, pages 264–269. IEEE.

Silva, A. B., Martins, A. S., Tosta, T. A. A., Loyola, A. M., Cardoso, S. V., Neves, L. A., de Faria, P. R., and do Nascimento, M. Z. (2024). OralEpitheliumDB: A dataset for oral epithelial dysplasia image segmentation and classification. Journal of Imaging Informatics in Medicine. pages 1-20.

Skopal, T. (2006). On fast non-metric similarity search by metric access methods. In EDBT, volume 3896 of LNCS, pages 718–736. Springer.

Song, Y., Gu, Y., Zhang, R., and Yu, G. (2022). BrePartition: Optimized high-dimensional ik/iNN search with bregman distances. IEEE TKDE, 34(3):1053–1065.

Sung, H., Ferlay, J., Siegel, R., Laversanne, M., Soerjomataram, I., Jemal, A., and Bray, F. (2021). Global cancer statistics 2020: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin, 71(3):209–249.

Teixeira, M. (2007). Explicação diversa para a origem do câncer, com foco nos cromossomos, e não nos genes, ganha corpo no establishment científico. Revista Latinoamericana de Psicopatologia Fundamental, 10(4):664–676.

Tommasino, C., Merolla, F., Russo, C., Staibano, S., and Rinaldi, A. M. (2023). Histopathological image deep feature representation for CBIR in smart PACS. Journal of Digital Imaging, 36(5):2194–2209.

Tosta, T., Faria, P., Servato, J. P., Neves, L., Roberto, G., Martins, A., and Nascimento, M. Z. (2019). Unsupervised method for normalization of hematoxylin-eosin stain in histological images. Comput. Med. Imaging Graph., 77:101646.

van der Waal, I. (2009). Potentially malignant disorders of the oral and oropharyngeal mucosa; terminology, classification and present concepts of management. Oral Oncology, 45(4–5):317–323.

Zheng, Y., Jiang, Z., Zhang, H., Xie, F., Ma, Y., Shi, H., and Zhao, Y. (2018). Histopathological whole slide image analysis using context-based cbir. IEEE Transactions on Medical Imaging, 37(7):1641–1652.