Paper Information | SBC BOOKS

Fill in your paper information

Language (*)

Title (*)

Keywords (*)

Abstract (*)

As úlceras do pé diabético representam uma complicação grave do diabetes mellitus, gerando um impacto significativo no Sistema Único de Saúde (SUS) brasileiro. Este trabalho propõe um pipeline de visão computacional orientado à saúde pública para a classificação e segmentação automatizada dessas lesões, priorizando a eficiência computacional (Green AI) e a robustez em cenários de dados limitados. A metodologia integra arquiteturas leves (EfficientNet-B0 e MobileNetV3-Large) com estratégias de adaptação eficiente, incluindo Batch Normalization-only (BN-only), técnicas de aumento de dados (MixUp/CutMix), FixRes e inferência adaptativa guiada por entropia. O consumo energético e a pegada de carbono (CO2) foram auditados via CodeCarbon utilizando os benchmarks DFUC2020 e DFUC2021. Os resultados demonstram que os modelos BN-only alcançam métricas competitivas (AUC, F1-macro e AUPRC) com um custo energético significativamente inferior ao full fine-tune, viabilizando a triagem em dispositivos móveis e ambientes com restrições de infraestrutura. Este estudo contribui com um protocolo experimental reproduzível e discute as implicações da IA sustentável e ética no fortalecimento da atenção primária e na redução de desigualdades no acesso ao diagnóstico especializado no Brasil.

Pages (*)

File Link

English Information

Title

Keywords

Abstract

Diabetic foot ulcers (DFU) are a severe complication of diabetes mellitus, imposing a significant burden on Brazil's public health system (SUS). This study proposes a public-health-oriented computer vision pipeline for the automated classification and segmentation of these lesions, prioritizing computational efficiency (Green AI) and robustness under data scarcity. The methodology integrates lightweight architectures (EfficientNet-B0 and MobileNetV3-Large) with efficient adaptation strategies, including Batch Normalization-only (BN-only), advanced data augmentation (MixUp/CutMix), FixRes, and entropy-guided adaptive inference. Energy consumption and CO2 emissions were rigorously audited using the CodeCarbon framework on the DFUC2020/DFUC2021 datasets. Experimental results demonstrate that BN-only models achieve competitive metrics (AUC, F1-macro, and AUPRC) at a significantly lower energy cost than full fine-tuning, enabling scalable screening on mobile devices and infrastructure-constrained environments. This work provides a reproducible experimental protocol and discusses the implications of sustainable and ethical AI for strengthening primary care and reducing diagnostic inequities in Brazil.

(*) To change the order drag the item to the new position.

Authors

#	Name
1	Danielle Monteiro(danimonteirodba@gmail.com)
2	Marcio Teixeira(mjt@unicamp.br)
3	Pedro Corrêa(pedro.correa@usp.br)

(*) To change the order drag the item to the new position.

Reference

#	Reference
1	Alhasson, H. F. and Alharbi, S. S. (2025). Advances in image-based diagnosis of diabetic foot ulcers using deep learning and machine learning: a systematic review. Biomedicines, 13(12):2928.
2	Alkhalefah, S., Alturaiki, I., and Altwaijry, N. (2025). Advancing diabetic foot ulcer care: Ai and generative ai approaches for classification, prediction, segmentation, and detection. Healthcare, 13(6):648.
3	Armstrong, D. G., Boulton, A. J. M., and Bus, S. A. (2017). Diabetic foot ulcers and their recurrence. New England Journal of Medicine, 376(24):2367–2375.
4	Brasil (2018). Lei nº 13.709, de 14 de agosto de 2018. lei geral de proteção de dados pessoais (lgpd). Diario Oficial da União, Brasília, DF, 15 ago. 2018. Disponível em: http://www.planalto.gov.br/ccivil_03/ _ato2015-2018/2018/lei/L13709.htm. Acesso em: 10 mai. 2026
5	Cassidy, B., Kendrick, C., Reeves, N. D., et al. (2022). Diabetic foot ulcer grand challenge 2021: evaluation and summary. In Diabetic Foot Ulcers Grand Challenge (DFUC 2021), Lecture Notes in Computer Science, volume 13183. Springer, Cham.
6	DATASUS (2025). Morbidade hospitalar do sus (sih/sus): procedimentos relacionados a amputações de membros inferiores. Departamento de Informática do SUS, Ministério da Saude. Disponível em: http://tabnet.datasus.gov.br. Acesso em: jan. 2025.
7	Frankle, J., Schwab, D. J., and Morcos, A. S. (2021). Training BatchNorm and only BatchNorm. In International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=vYeQQ29Tbvx.
8	Galdran, A., Carneiro, G., and Ballester, M. A. G. (2022). Convolutional nets versus vision transformers for diabetic foot ulcer classification. In Diabetic Foot Ulcers Grand Challenge (DFUC 2021), Lecture Notes in Computer Science, volume 13183. Springer.
9	Guo, C., Pleiss, G., Sun, Y., and Weinberger, K. Q. (2017). On calibration of modern neural networks. In Proceedings of the 34th International Conference on Machine Learning (ICML), PMLR 70, pages 1321–1330. https://proceedings.mlr.press/v70/guo17a.html.
10	Howard, A., Sandler, M., Chu, G., et al. (2019). Searching for MobileNetV3. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 1314–1324.
11	IBGE (2020). Pesquisa nacional de saúde 2019: informações sobre domicílios, acesso e utilização dos serviços de saúde. Technical report, Instituto Brasileiro de Geografia e Estatı́stica, Rio de Janeiro. https://biblioteca.ibge.gov.br/index.php/biblioteca-catalogo?view=detalhes&id=2101748.
12	IBGE (2022). Censo demográfico 2022: resultados do universo — características da população. Technical report, Instituto Brasileiro de Geografia e Estatística, Rio de Janeiro. Disponível em: https://www.ibge.gov.br/estatisticas/sociais/populacao/22827-censo-demografico-2022.html. Acesso em: 10 mai. 2026.
13	IDF (2019). Idf diabetes atlas. Technical report, International Diabetes Federation. https://www.diabetesatlas.org.
14	Mehta, S. and Rastegari, M. (2022). Mobilevit: Light-weight, general-purpose, and mobile-friendly vision transformer. In International Conference on Learning Representations (ICLR).
15	Ministério da Saúde (2020). Vigitel brasil 2019: vigilância de fatores de risco e proteção para doenças crônicas por inquérito telefônico. Technical report, Ministério da Saúde, Brasília. https://bvsms.saude.gov.br/bvs/publicacoes/vigitel_brasil_2019_vigilancia_fatores_risco.pdf.
16	Niri, R., Zahia, S., Stefanelli, A., et al. (2025). Wound segmentation with u-net using a dual attention mechanism and transfer learning. Journal of Imaging Informatics in Medicine.
17	Oliveira, T. C. D., Oliveira, A. F., Araújo, L. C., et al. (2025). Digital health technologies for diabetic foot ulcers: a systematic review of clinical evidence, access inequities, and public health integration. International Journal of Environmental Research and Public Health, 22(9):1430.
18	Schmidt, V., Goyal, K., Joshi, A., et al. (2021). CodeCarbon: estimate and track carbon emissions from machine learning computing.
19	Schwartz, R., Dodge, J., Smith, N. A., and Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12):54–63.
20	Strubell, E., Ganesh, A., and McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL), pages 3645–3650.
21	Tan, M. and Le, Q. V. (2019). EfficientNet: rethinking model scaling for convolutional neural networks. In Proceedings of the 36th International Conference on Machine Learning (ICML), PMLR 97, pages 6105–6114. https://proceedings.mlr.press/v97/tan19a.html.
22	Touvron, H., Vedaldi, A., Douze, M., et al. (2019). Fixing the train-test resolution discrepancy. In Advances in Neural Information Processing Systems (NeurIPS), volume 32, pages 8250–8260. https://proceedings.neurips.cc/paper/2019/hash/dc6a7e655d7e5840e66733e9ee67cc69-Abstract.html.
23	Yang, L., Mehta, N., Chen, D., et al. (2020). Resolution adaptive networks for efficient inference. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2369–2378.
24	Yap, M. H., Cassidy, B., Pappachan, J. M., et al. (2021). Analysis towards classification of infection and ischaemia of diabetic foot ulcers. In 2021 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI), pages 1–4.
25	Yun, S., Han, D., Oh, S. J., et al. (2019). CutMix: regularization strategy to train strong classifiers with localizable features. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 6023–6032.
26	Zhang, H., Cisse, M., Dauphin, Y. N., and Lopez-Paz, D. (2018). mixup: beyond empirical risk minimization. In International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=r1Ddp1-Rb.

Paper Registration

Fill in your paper information

English Information

Authors

Reference