EfficientBacillus: uma arquitetura profunda para detecção dos bacilos de Koch

• Patrick Ryan Sales dos Santos UFPI
• Vitória de Carvalho Brito UFPI
• Antonio Oseas de Carvalho Filho UFPI
• Alcilene Dalília de Sousa UFPI
• João Otávio Bandeira Diniz IFMA
• Daniel de Sousa Luz IFPI

DOI: https://doi.org/10.5753/sbcas.2023.229608

Resumo

A tuberculose é uma infecção bacteriana causada pelo bacilo de Koch e é transmitida pelo ar. A doença afeta principalmente os pulmões e é considerada a segunda principal causa de morte por infecção no mundo. Apesar disso, a tuberculose tem cura e o diagnóstico precoce é de suma importância para o sucesso do tratamento e para evitar a disseminação da doença. Tradicionalmente, a baciloscopia de esfregaço de escarro tem sido o principal método para o diagnóstico e monitoramento do tratamento da tuberculose. Nesse contexto, diversas abordagens computacionais têm sido desenvolvidas para auxiliar no diagnóstico da tuberculose, através da análise das imagens de baciloscopia. Neste trabalho, propomos a utilização da EfficientDet, explorando cada um dos seus backbones na tarefa de identificação dos bacilos. Também avaliamos 4 representações de cor diferentes e aplicamos uma validação cruzada com k-fold = 5. Os resultados mostraram-se promissores, com IoU de 0,523, recall de 0,925, precision de 0,694 e f1-score de 0,774. Os resultados alcançados evidenciaram o potencial do método na detecção dos bacilos, podendo auxiliar no diagnóstico da tuberculose.

Referências

An, L., Peng, K., Yang, X., Feng, P., and Huang, P. (2022). Automated detection of tuberculosis bacilli using deep neural networks with sputum smear images. In 2022 5th International Conference on Pattern Recognition and Artificial Intelligence (PRAI), pages 1040–1045. IEEE.

Díaz-Huerta, J. L., Téllez-Anguiano, A. C., Gutiérrez-Gnecchi, J. A., Arellano-Calderón, S., and Olivares-Rojas, J. C. (2022). Analysis and comparison of image processing and artificial intelligence algorithms to detect afb in pulmonary tuberculosis images. Tuberculosis, 134:102196.

Kaggle (2020). Tuberculosis image dataset. Retrieved from [link]. Accessed November 13, 2022.

Kotei, E. and Thirunavukarasu, R. (2022). Computational techniques for the automated detection of mycobacterium tuberculosis from digitized sputum smear microscopic images: A systematic review. Progress in Biophysics and Molecular Biology.

Marques, G., Agarwal, D., and de la Torre Díez, I. (2020). Automated medical diagnosis of covid-19 through efficientnet convolutional neural network. Applied soft computing, 96:106691.

Mekhalfi, M. L., Nicolò, C., Bazi, Y., Al Rahhal, M. M., Alsharif, N. A., and Al Maghayreh, E. (2021). Contrasting yolov5, transformer, and efficientdet detectors for crop circle detection in desert. IEEE Geoscience and Remote Sensing Letters, 19:1–5.

Reist, F. J. D. S., Veloso, M. A., Rodrigues, F. M. M., Brito, V. D. C., Dos Santos, P. R. S., Araújo, J. D. L., Rabelo, R. D. A. L., and De Carvalho Filho, A. O. (2021). Bacillusnet: An automated approach using retinanet for segmentation of pulmonary tuberculosis bacillus. In 2021 IEEE Symposium on Computers and Communications (ISCC), pages 1–4. IEEE.

Rico-Garcia, M., Salazar, A., Madrigal, C.-A., Morantes-Guzman, L.-J., and Cortes-Mancera, F. (2015). Detection of mycobcaterium tuberculosis in microscopic images of ziehl-neelsen-stained sputum smears lacnem 2015. In 6th Latin-American Conference on Networked and Electronic Media (LACNEM 2015), pages 1–6. IET.

Shah, M. I., Mishra, S., Sarkar, M., and Rout, C. (2016). Automatic detection and classification of tuberculosis bacilli from zn-stained sputum smear images using watershed segmentation. In International Conference on Signal Processing (ICSP 2016), pages 1–4.

Shwetha, V., Prasad, K., Mukhopadhyay, C., Banerjee, B., and Chakrabarti, A. (2021). Automatic detection of bacilli bacteria from ziehl-neelsen sputum smear images. In 2021 2nd International Conference on Communication, Computing and Industry 4.0 (C2I4), pages 1–5. IEEE.

Song, S., Jing, J., Huang, Y., and Shi, M. (2021). Efficientdet for fabric defect detection based on edge computing. Journal of Engineered Fibers and Fabrics, 16:15589250211008346.

Tan, M. and Le, Q. (2019). Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning, pages 6105–6114. PMLR.

Tan, M., Pang, R., and Le, Q. V. (2020). Efficientdet: Scalable and efficient object detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10781–10790.

WHO, W. H. O. (2021). Tuberculosis. Retrieved from [link]. Accessed March 05, 2023.

Wu, M., Heng, C., Zhu, H., and Cai, H. (2022). Covid-19 detection based on computer vision and big data. In 2022 7th International Conference on Intelligent Computing and Signal Processing (ICSP), pages 474–477. IEEE.

Yousefi, H., Mohammadi, F., Mirian, N., and Amini, N. (2020). Tuberculosis bacilli identification: a novel feature extraction approach via statistical shape and color models. In 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), pages 366–371. IEEE.

Zachariou, M., Arandjelović, O., Sabiiti, W., Mtafya, B., and Sloan, D. (2022). Tuberculosis bacteria detection and counting in fluorescence microscopy images using a multi-stage deep learning pipeline. Information, 13(2):96.