EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

HYBRID TRANSFORMER–CNN FRAMEWORK FOR EXPLAINABLE MULTIMODAL BRAIN TUMOR CLASSIFICATION USING MULTI-SEQUENCE MRI

Kafui Tsoeke Agbevanu, Mohamed Merza Khalil Ebrahim Al.Rayes

Qubahan Journal of Medical Sciences · 2024

DOI: 10.48161/qjms.v1a17 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Qubahan Journal of Medical Sciences, 2024. Verfasst von Kafui Tsoeke Agbevanu und 1 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

There has been significant progress in the field of medical image analysis due to advancements in artificial intelligence (AI) and deep learning technologies, enabling automated, more accurate, and more efficient disease diagnosis. An AI-based model, Hybrid CNN-Transformer, is introduced for explainable brain tumor classification using MRI brain images. The proposed framework leverages the feature-extraction capabilities of CNNs and the context-learning capabilities of Transformers to achieve better diagnostic outcomes and increased interpretability. For this purpose, a freely available MRI brain tumor dataset, comprising glioma, meningioma, pituitary, and no-tumor categories, was used to develop and validate the proposed framework. Prior to developing the model, preprocessing steps, such as resizing, normalization, and data augmentation, were applied to MRI images to enhance image quality and improve the model's generalization capacity. The DenseNet121 CNN was used to extract features in the spatial domain, whereas the Transformer encoder was used to capture contextual relationships via self-attention mechanisms. Experimental evaluation has been conducted using accuracy, precision, recall, F1 score, specificity, the confusion matrix, and ROC-AUC. Comparison with CNN and Transformer models shows that the proposed approach achieves competitive diagnostic accuracy while offering better explainability. The results clearly indicate considerable promise in using hybrid architectures for AI-based radiological diagnostics

Zitieren

APA

Agbevanu, K. T., & Al.Rayes, M. M. K. E. (2024). HYBRID TRANSFORMER–CNN FRAMEWORK FOR EXPLAINABLE MULTIMODAL BRAIN TUMOR CLASSIFICATION USING MULTI-SEQUENCE MRI. Qubahan Journal of Medical Sciences. https://doi.org/10.48161/qjms.v1a17

Vancouver

Agbevanu KT, Al.Rayes MMKE. HYBRID TRANSFORMER–CNN FRAMEWORK FOR EXPLAINABLE MULTIMODAL BRAIN TUMOR CLASSIFICATION USING MULTI-SEQUENCE MRI. Qubahan Journal of Medical Sciences. 2024. doi:10.48161/qjms.v1a17

BibTeX
@article{Agbevanu2024,
  author = {Agbevanu, Kafui Tsoeke and Al.Rayes, Mohamed Merza Khalil Ebrahim},
  title = {HYBRID TRANSFORMER–CNN FRAMEWORK FOR EXPLAINABLE MULTIMODAL BRAIN TUMOR CLASSIFICATION USING MULTI-SEQUENCE MRI},
  journal = {Qubahan Journal of Medical Sciences},
  year = {2024},
  doi = {10.48161/qjms.v1a17}
}
Kostenlos starten →