Segmentation of gliomas in magnetic resonance images using modified U-Net

• Roney Nogueira de Sousa IFCE

DOI: https://doi.org/10.5753/ercas.2024.238512

Abstract

In this study, the performance of a modified U-Net was assessed for the segmentation of gliomas detected in magnetic resonance imaging (MRI) images. A public dataset was utilized, incorporating data augmentation techniques. After 60 training epochs, promising results were achieved, with an accuracy of 99.77%, IOU of 90.21%, and Dice coefficient of 98.59%.

References

Chollet, F. et al. (2015). Keras. [link].

Despotović, I. et al. (2015). Mri segmentation of the human brain: challenges, methods, and applications. Computational and mathematical methods in medicine, 2015.

Gavrikov, P. (2020). Visualkeras. [link].

Hunter, J. D. (2007). Matplotlib: A 2d graphics environment. Computing in Science & Engineering, 9(3):90–95.

Oren, O. et al. (2020). Artificial intelligence in medical imaging: switching from radiographic pathological data to clinically meaningful endpoints. The Lancet Digital Health, 2(9):e486–e488.

Tang, X. (2019). The role of artificial intelligence in medical imaging research. BJR—Open, 2(1):20190031.

Van Rossum, G. and Drake, F. L. (2009). Python 3 Reference Manual. CreateSpace, Scotts Valley, CA.