| dc.description.abstract | Brain tumors are one of the most serious and potentially life-threatening medical conditions.
Each year, thousands of people die from brain tumors due to the rapidly growing tumor cells.
In order to save lives worldwide, automatic brain tumor detection is critically important.
Quick and accurate identification of the tumor type is essential for determining the most
effective treatment approach and improving outcomes. Currently, histopathological
examination of biopsy samples by medical professionals is used for diagnosis and classification
of brain tumors. However, this process can be time-consuming and depends on the assessing
doctor's expertise.
These limitations underscore the need for fully automated deep learning methods for brain
tumor classification. Convolutional neural networks (CNNs) are highly suitable for analyzing
medical images like brain scans. CNNs are a type of deep learning that is able to learn visual
features through convolutional and max pooling layers.
Recent research has frequently utilized transfer learning with pre-trained models to detect and
classify brain tumors. Deep transfer learning techniques show strong potential for
automatically identifying all brain tumor types using medical images. Automated analysis
through transfer learning could streamline diagnosis and help optimize treatment planning.
In this thesis, my aims to improve the classification of brain tumors by combining the power
of two pretrained convolutional neural network (CNN) architectures which are MobileNetV3
and InceptionV3. Leveraging a dataset obtained from Kaggle, consisting of brain images with
and without tumors. This research seeks to improve the accuracy and robustness of brain tumor
classification models. The proposed methodology involves fine-tuning and integrating the
combine pretrained model to harness their complementary strengths in feature extraction and
classification. | en_US |
