Metodologia automática para detecção de bacilos de tuberculose utilizando RetinaNet e modelos de cores

• Filipe M. M. Rodrigues UFPI
• Francisco J. S. Reis UFPI
• Mateus A. Veloso UFPI
• João O. B. Diniz IFMA
• Romuere R. Veloso UFPI / UFMA
• Antonio O. C. Filho UFPI / UFMA

DOI: https://doi.org/10.5753/sbcas.2022.222677

Abstract

Tuberculosis is the most deadly bacterial infectious disease globally, with about 1.5 million people dying every year. The disease is caused by Mycobacterium tuberculosis. The main form of diagnosis is sputum bacilloscopy, an exam in which the patient's sputum is analyzed under a microscope searching for bacillus, making this a technique for both diagnosis and monitoring the disease. Therefore, this work aims to develop a methodology for automated detection of the bacillus using RetinaNet. A set of 1218 images was used to evaluate this method. The results were encouraging, with an accuracy of 64.9%, a recall of 70.4%, and an F1 score of 61%. Finally, we believe that our method can act in diagnosing tuberculosis.

References

Araujo, F. H. D., Silva, R. R. V., Medeiros, F. N. S., Parkinson, D. D., Hexemer, A., Carneiro, C. M., and Ushizima, D. M. (2018). Reverse image search for scientific data within and beyond the visible spectrum. Expert Systems with Applications, 109:35-48.

da Saúde., B. M. (2022). Tuberculose. Disponível em: https://www.gov.br/saude/pt-br/assuntos/saude-de-a-a-z/t/tuberculose. Acessado em 15 de fevereiro de 2022.

El-Melegy, M., Mohamed, D., and Elmelegy, T. (2019). Automatic Detection of Tuberculosis Bacilli from Microscopic Sputum Smear Images Using Faster R-CNN, Transfer Learning and Augmentation, pages 270-278.

Gráˆbel, P., Ozkan, O., Crysandt, M., Herwartz, R., Baumann, M., Klinkhammer, B. M., Boor, P., Br ummendorf, T. H., and Merhof, D. (2020). Circular anchors for the detection of hematopoietic cells using retinanet. In 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI), pages 249-253. IEEE.

Jiao, L., Zhang, F., Liu, F., Yang, S., Li, L., Feng, Z., and Qu, R. (2019). A survey of deep learning-based object detection. IEEE Access, 7:128837-128868.

Kaggle (2020). Tuberculosis image dataset. Retrieved from https://www.kaggle.com/saife245/tuberculosis-image-datasets. Accessed June 05, 2021.

Kant, S. and Srivastava, M. M. (2018). Towards automated tuberculosis detection using deep learning. In 2018 IEEE Symposium Series on Computational Intelligence (SSCI), pages 1250-1253.

Lin, T.-Y., Dollár, P., Girshick, R., He, K., Hariharan, B., and Belongie, S. (2017). Feature pyramid networks for object detection. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 936-944.

Lin, T.-Y., Goyal, P., Girshick, R., He, K., and Dollár, P. (2020). Focal loss for dense object detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(2):318-327.

Pho, K., Mohammed Amin, M. K., and Yoshitaka, A. (2018). Segmentationdriven retinanet for protozoa detection. In 2018 IEEE International Symposium on Multimedia (ISM), pages 279-286.

Reza, A. M. (2004). Realization of the contrast limited adaptive histogram equalization (clahe) for real-time image enhancement. Journal of VLSI signal processing systems for signal, image and video technology, 38(1):35-44.

SESA (2022). Tuberculose. Disponível em: https://saude.es.gov.br/nevetuberculose. Acessado em 8 de fevereiro de 2022.

Swetha, K., Sankaragomathi, B., and Thangamalar, J. B. (2020). Convolutional neural network based automated detection of mycobacterium bacillus from sputum images. In 2020 International Conference on Inventive Computation Technologies (ICICT), pages 293-300.

Targ, S., Almeida, D., and Lyman, K. (2016). Resnet in resnet: Generalizing residual architectures. arXiv preprint arXiv:1603.08029.

WHO, W. H. O. (2022). Tuberculosis. Disponível em: https://www.who.int/en/news-room/fact-sheets/detail/tuberculosis. Acessado em 05 de fevereiro de 2022.