Braint Tumor Classification & Segmentation Model Using Deep Learning Approach
Domain:
healthcareRecord type:
modelpaper
Creator:
Mek
Editor:
Hus
Publisher:
Nat
Host:
Cancer is one of the most dangerous diseases in the world. The brain tumor is one type of
cancer that rapidly and uncontrollable growth of abnormal tissues that damages the nearby
healthy tissues of the brain. The classification and segmentation of brain tumors from MRI
being performed manually by medical experts. A large quantity of time was spent by
radiologists and doctors for the identification of tumors and segmenting it from different brain
tissues. However, actual labeling brain tumors could be a long task, and segmentation of
tumors is a tedious task for doctors. The brain MRI dataset is collected from a pioneer
diagnosis center located at Addis Ababa Ethiopia. Deep convolutional layers become a
popular algorithm to overcome the problems traditional machine learning algorithms by
automatically feature extraction from the input spaces and accurately detect the brain tumors.
Transfer learning is explored by freezing layers and fine-tuning pre-trained model of
convolutional neural network architectures using MRI slices of brain tumor datasets.
Therefore, accurately predict the brain tumor is significant to give pre-treatment of the patient
and to improve the existence of patients with tumors. So, the convolutional neural network
model is proposed by modifying the VGG16 pre-trained model using a deep learning approach
to classify normal and abnormal brain tumors. The proposed convolutional neural network
model is trained using 3000 MRI, and data augmentation techniques applied to generate more
MRI slices to improve the performance of algorithms and reduce overfitting problems. The
performance of the proposed convolutional neural network model is tested using 800 MRI
datasets. To segment the abnormal region of tumors and the area of the tumor using optimized
Fuzzy C Means algorithms. The mean accuracy of the proposed convolutional neural network
model is found to be 98% for brain tumor classifications with mini-batch size 32 and learning
rate 0.001. The final performance metrics for brain tumor segmentation used the Jaccard
similarity index is 90.81 %, the Dice similarity score is 95.16 %, the sensitivity is 96.54 %, and
the specificity is 94.75% accuracy results using the proposed algorithms. To identify brain
tumor stages and annotate the abnormal brain tumors by using domain experts is
recommended as future work
cancer that rapidly and uncontrollable growth of abnormal tissues that damages the nearby
healthy tissues of the brain. The classification and segmentation of brain tumors from MRI
being performed manually by medical experts. A large quantity of time was spent by
radiologists and doctors for the identification of tumors and segmenting it from different brain
tissues. However, actual labeling brain tumors could be a long task, and segmentation of
tumors is a tedious task for doctors. The brain MRI dataset is collected from a pioneer
diagnosis center located at Addis Ababa Ethiopia. Deep convolutional layers become a
popular algorithm to overcome the problems traditional machine learning algorithms by
automatically feature extraction from the input spaces and accurately detect the brain tumors.
Transfer learning is explored by freezing layers and fine-tuning pre-trained model of
convolutional neural network architectures using MRI slices of brain tumor datasets.
Therefore, accurately predict the brain tumor is significant to give pre-treatment of the patient
and to improve the existence of patients with tumors. So, the convolutional neural network
model is proposed by modifying the VGG16 pre-trained model using a deep learning approach
to classify normal and abnormal brain tumors. The proposed convolutional neural network
model is trained using 3000 MRI, and data augmentation techniques applied to generate more
MRI slices to improve the performance of algorithms and reduce overfitting problems. The
performance of the proposed convolutional neural network model is tested using 800 MRI
datasets. To segment the abnormal region of tumors and the area of the tumor using optimized
Fuzzy C Means algorithms. The mean accuracy of the proposed convolutional neural network
model is found to be 98% for brain tumor classifications with mini-batch size 32 and learning
rate 0.001. The final performance metrics for brain tumor segmentation used the Jaccard
similarity index is 90.81 %, the Dice similarity score is 95.16 %, the sensitivity is 96.54 %, and
the specificity is 94.75% accuracy results using the proposed algorithms. To identify brain
tumor stages and annotate the abnormal brain tumors by using domain experts is
recommended as future work