Glioblastomas brain tumour segmentation based on convolutional neural networks

Moh'd Rasoul AlHadidi, Bayan AlSaaidah, Mohammed Al-Gawagzeh

Abstract


Brain tumour segmentation can improve diagnostics efficiency, rise the prediction rate and treatment planning. This will help the doctors and experts in their work. Where many types of brain tumour may be classified easily, the gliomas tumour is challenging to be segmented because of the diffusion between the tumour and the surrounding edema. Another important challenge with this type of brain tumour is that the tumour may grow anywhere in the brain with different shape and size. Brain cancer presents one of the most famous diseases over the world, which encourage the researchers to find a high-throughput system for tumour detection and classification. Several approaches have been proposed to design automatic detection and classification systems.
This paper presents an integratedframework to segment the gliomas brain tumour automatically using pixel clustering for the MRI images foreground and background and classify its type based on deep learning mechanism, which is the Convolutional Neural Network (CNN). In this work, a novel segmentation and classification system is proposed to detect the tumour cells and classify the brain image if it is healthy or not. After collecting data for healthy and non-healthy brain images, satisfactory results are found and registered using computer vision approaches. This approach can be used as a part of a bigger diagnosis system for breast tumour detection and manipulation.

Keywords


Brain tumour; Convolutional neural networks; Superpixel; Image segmentation; Pixel clustering



DOI: http://doi.org/10.11591/ijece.v10i5.pp%25p
Total views : 51 times

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.