EfficientNet-Based Myocarditis Detection in T2-weighted CMR Images: A Patient-Level Evaluation Strategy

• Robert G. V. de Souza UFJF
• Joventino O. Campos UFJF
• Luiz M. S. Maciel UFJF

DOI: https://doi.org/10.5753/eniac.2026.31335

Resumo

Myocarditis diagnosis via Cardiac Magnetic Resonance is challenged by subtle tissue alterations and observer subjectivity. This study evaluates four EfficientNet architectures for automated classification using exclusively non-contrast T2-weighted images. Ensuring methodological rigor, we implemented a 5-fold cross-validation with strict patient-level separation on 45 participants, applying data augmentation to balance classes. EfficientNetB1 achieved the best performance across 10 independent repetitions, with 96.73% average image-level. Aggregating slice predictions via majority voting yielded 99.56% patient-level accuracy. Results demonstrate a reliable non-invasive diagnostic support tool that simplifies clinical protocols by eliminating contrast agents.

Referências

Baeßler, B. et al. (2019). Radiomics for distinguishing myocardial infarction from myocarditis at late gadolinium enhancement at mri: Comparison with subjective visual analysis. Radiology: Cardiothoracic Imaging, 1(3):e180026.

Ferreira, V. M. et al. (2018). Cardiovascular magnetic resonance in nonischemic myocardial inflammation: expert recommendations. Journal of the American College of Cardiology, 72(24):3158–3176.

Friedrich, M. G. et al. (2009). Cardiovascular magnetic resonance in myocarditis: A jacc white paper. Journal of the American College of Cardiology, 53(17):1475–1487.

Hatfaludi, C.-A. et al. (2024). Deep learning automatically distinguishes myocarditis patients from normal subjects based on mri. The International Journal of Cardiovascular Imaging, 40(12):2617–2629.

Kailani, N. M. S. (2025). Comparison between resnet and efficientnet for image classification: An analytical study of performance and efficiency. North African Journal of Scientific Publishing (NAJSP), 3(4):18–25.

Kasmaee, A. M. M., Ataei, A., Moravvej, S. V., Alizadehsani, R., Gorriz, J. M., Zhang, Y.-D., Tan, R.-S., and Acharya, U. R. (2024). Elrl-md: a deep learning approach for myocarditis diagnosis using cardiac magnetic resonance images with ensemble and reinforcement learning integration. Physiological Measurement, 45(5):055002.

Łajczak, P. M. and Jóźwik, K. (2024). Artificial intelligence and myocarditis: a systematic review of current applications. Heart Failure Reviews, 29(6):1217–1234.

Łajczak, P. M., Sahin, O. K., Matyja, J., et al. (2025). Machine learning for myocarditis diagnosis using cardiovascular magnetic resonance: a systematic review, diagnostic test accuracy meta-analysis, and comparison with human physicians. The International Journal of Cardiovascular Imaging.

Loshchilov, I. and Hutter, F. (2019). Decoupled weight decay regularization. International Conference on Learning Representations.

Oh, K., Chung, Y.-C., Kim, K., Kim, W., and Oh, I. (2019). Deep learning for schizophrenia classification. IEEE International Conference on Big Data and Smart Computing (BigComp), pages 113–116.

Sharifrazi, D., Alizadehsani, R., Joloudari, J. H., Band, S. S., Hussain, S., Sani, Z. A., Hasanzadeh, F., Shoeibi, A., Dehzangi, A., Sookhak, M., et al. (2022). Cnn-kcl: Automatic myocarditis diagnosis using convolutional neural network combined with k-means clustering. Mathematical Biosciences and Engineering, 19(3):2381–2402.

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

Yagis, E., Atnafu, S. W., and Garcia Seco de Herrera, A. (2021). Effect of data leakage in brain mri classification using 2d convolutional neural networks. Scientific Reports, 11(1):22544.

Zech, J. R., Badgeley, M. A., Liu, M., Costa, A. B., Titano, J. J., and Oermann, E. K. (2018). Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study. PLoS medicine, 15(11):e1002683.