Authors :
Yasir Mehmood; Naeem Naseer
Volume/Issue :
Volume 8 - 2023, Issue 9 - September
Google Scholar :
https://tinyurl.com/2s2twxu7
Scribd :
https://tinyurl.com/uvbm3zfu
DOI :
https://doi.org/10.5281/zenodo.10025342
Abstract :
Brain tumors are pathological disorders
characterized by unregulated cell proliferation inside
damaged tissues, demanding early identification to
prevent uncontrollable development. Because of its higher
image quality, magnetic resonance imaging (MRI) is a
commonly used tool for the first diagnosis of brain
tumors. Deep learning, a subset of artificial intelligence,
has recently been integrated, ushering in novel ways to
automate medical picture recognition. Transfer learning
techniques applied to MRI images, this study hopes to
give a reliable and effective methodology for the early
diagnosis of brain tumors. This study uses a deep learning
architecture using sequential Convolutional Neural
Networks (CNNs) and two pre-trained models, VGG16
and EfficientNetB4, from the ImageNet dataset to classify
brain tumor pictures. Image preprocessing methods are
used prior to model training to improve model
performance. The experiments use the BrcH35 dataset
from Kaggle, which has been preprocessed in the MASK
RCNN format. The top-performing transfer learning
models are evaluated using performance criteria such as
accuracy, precision, and F1 score. According to the
results from this work, the EfficientNetB4 model beats
the other models, reaching exceptional accuracy,
precision, and F1 score values of 99.66%, 99.68%, and
100%, respectively. This proposed approach extends
existing research in the field and illustrates its potential
for faster and more reliable brain tumor detection.
Keywords :
Brain tumors; Magnetic Resonance Imaging (MRI);Transfer Learning, Convolutional Neural Networks (CNNs); VGG16, EfficientNetB4.
Brain tumors are pathological disorders
characterized by unregulated cell proliferation inside
damaged tissues, demanding early identification to
prevent uncontrollable development. Because of its higher
image quality, magnetic resonance imaging (MRI) is a
commonly used tool for the first diagnosis of brain
tumors. Deep learning, a subset of artificial intelligence,
has recently been integrated, ushering in novel ways to
automate medical picture recognition. Transfer learning
techniques applied to MRI images, this study hopes to
give a reliable and effective methodology for the early
diagnosis of brain tumors. This study uses a deep learning
architecture using sequential Convolutional Neural
Networks (CNNs) and two pre-trained models, VGG16
and EfficientNetB4, from the ImageNet dataset to classify
brain tumor pictures. Image preprocessing methods are
used prior to model training to improve model
performance. The experiments use the BrcH35 dataset
from Kaggle, which has been preprocessed in the MASK
RCNN format. The top-performing transfer learning
models are evaluated using performance criteria such as
accuracy, precision, and F1 score. According to the
results from this work, the EfficientNetB4 model beats
the other models, reaching exceptional accuracy,
precision, and F1 score values of 99.66%, 99.68%, and
100%, respectively. This proposed approach extends
existing research in the field and illustrates its potential
for faster and more reliable brain tumor detection.
Keywords :
Brain tumors; Magnetic Resonance Imaging (MRI);Transfer Learning, Convolutional Neural Networks (CNNs); VGG16, EfficientNetB4.
