ROI Segmentation in Thyroid Thermography via Deep Learning

• Alessandro C. B. da Silveira UERJ
• Sílvia C. D. Pinto UERJ

DOI: https://doi.org/10.5753/sbcas_estendido.2026.26547

Resumo

Thyroid cancer is the most frequent endocrine tumor in the world, and its traditional diagnosis depends on invasive methods such as ultrasonography and fine-needle aspiration biopsy (FNAB), which are subject to limitations. Infrared thermography emerges as a non-invasive and accessible alternative examination that records the thermal increase caused by tumor hypermetabolism. Thus, this study proposes an automatic approach through the adaptation of U-Net to obtain segmentation of the region of interest in thermographic images of the thyroid from the Visual Lab Thyroid Database. It was concluded that the proposed model achieved good performance (Dice: 0.9066; IoU: 0.8312; Accuracy: 0.8913), but still faces challenges in the delineation of fine contours due to the reduction of convolutional blocks in the encoder and loss of resolution. The results highlight the potential of the solution to automate initial screening, suggesting as next steps the use of K-fold cross-validation and morphological operations to refine the obtained masks.

Referências

Alves, M. L. D. and Gabarra, M. H. C. (2016). Comparison of power doppler and thermography for the selection of thyroid nodules in which fine-needle aspiration biopsy is indicated. Radiologia Brasileira, 49(5):311–315.

Damião, C., Montero, J., Moran, M., Da, R., Cruz Filho, R., Fontes, C., Lima, G., and Conci, A. (2020). On the possibility of using temperature to aid in thyroid nodule investigation. Scientific Reports, 10.

Gonzalez, R. C. and Woods, R. E. (2018). Digital Image Processing. Pearson, New York, 4 edition.

González, J., Damião, C., and Conci, A. (2017). An infrared thermal images database and a new technique for thyroid nodules analysis. Studies in health technology and informatics, 245:384–387.

González, J. R., Damião, C., Moran, M., Pantaleão, C. A., Cruz, R. A., Balarini, G. A., and Conci, A. (2021). A computational study on the role of parameters for identification of thyroid nodules by infrared images (and comparison with real data). Sensors, 21(13).

He, K., Zhang, X., Ren, S., and Sun, J. (2015). Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. IEEE International Conference on Computer Vision (ICCV 2015), 1502.

Ioffe, S. and Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift.

Júnior, R. J. A., Silva, I., Silva, A., and Paiva, A. (2021). Detecção de nódulos da tireoide em exames de termografia utilizando redes neurais convolucionais em cascata. In Anais do XXI Simpósio Brasileiro de Computação Aplicada à Saúde - SBCAS, pages 269–280, Porto Alegre, RS, Brasil. SBC.

Kingma, D. and Ba, J. (2014). Adam: A method for stochastic optimization. International Conference on Learning Representations.

Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., van der Laak, J. A., van Ginneken, B., and Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical Image Analysis, 42:60–88.

Moran, M., Conci, A., and Araújo, A. (2019). Evaluation of quantitative features and convolutional neural networks for nodule identification in thyroid thermographies. pages 747–751.

Moran, M., Conci, A., Gonzalez, J., Araújo, A., Fiirst, W., Damiao, C., Lima, G., and Cruz Filho, R. (2018a). Identification of thyroid nodules in infrared images by convolutional neural networks. pages 1–7.

Moran, M., Conci, A., Gonzalez, J., Saade, D., Fiirst, W., Araújo, A., Damiao, C., Lima, G., Cruz Filho, R., and Resmini, R. (2018b). Monomodal image registration by tensor analysis: Going beyond the brightness constancy assumption. pages 1–8.

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, pages 234–241. Springer International Publishing.

Sebastiao, C. and Ulbricht, L. (2020). Termografia como método para auxilio diagnóstico do câncer de tireoide. Pan American Journal of Medical Thermology, 5:19.

Shorten, C. and Khoshgoftaar, T. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6.

Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1):1929–1958.

Sudre, C., Li, W., Vercauteren, T., Ourselin, S., and Cardoso, M. J. (2017). Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations. volume 2017, pages 240–248.

Vicnesh, J., Salvi, M., Hagiwara, Y., Hah, Y., Mir, H., Barua, P. D., Chakraborty, S., Molinari, F., and Acharya, U. (2024). Application of infrared thermography and artificial intelligence in healthcare: A systematic review of over a decade (2013-2024). IEEE Access, PP:1–1.