Detection and classification of brain tumor using hybrid deep learning models
Abstract : Accurately classifying brain tumor types is critical for timely diagnosis and potentially saving lives. Magnetic Resonance Imaging (MRI) is a widely used non-invasive method for obtaining high-contrast grayscale brain images, primarily for tumor diagnosis.
The application of Convolutional Neural Networks (CNNs) in deep learning has revolutionized diagnostic systems, leading to significant advancements in medical imaging interpretation.
In this study, we employ a transfer learning-based fine-tuning approach using EfficientNets to classify brain tumors into three categories: glioma, meningioma, and pituitary tumors. We utilize the publicly accessible CE-MRI Figshare dataset to fine-tune five pre-trained models from the EfficientNets family, ranging from EfficientNetB0 to EfficientNetB4.
Our approach involves a two-step process to refine the pre-trained EfficientNet model. First, we initialize the model with weights from the ImageNet dataset. Then, we add additional layers, including top layers and a fully connected layer, to enable tumor classification.
We conduct various tests to assess the robustness of our fine-tuned EfficientNets in comparison to other pre-trained models. Additionally, we analyze the impact of data augmentation on the model's test accuracy.
To gain insights into the model's decision-making, we employ Grad-CAM visualization to examine the attention maps generated by the most optimal model, effectively highlighting tumor locations within brain images. Our results reveal that using EfficientNetB2 as the underlying framework yields significant performance improvements. Specifically, the overall test accuracy, precision, recall, and F1-score were found to be 99.06%, 98.73%, 99.13%, and 98.79%, respectively.
The author presents a classification method that utilizes a pre-trained GoogLeNet to extract features from brain MRI scans, taking advantage of deep transfer learning. These extracted features are then combined with well-established classifier models.
The MRI dataset is sourced from Figshare, and the experiment employs a five-fold cross-validation procedure at the patient level. The proposed approach significantly outperforms existing methods, achieving a mean classification accuracy of 98%.
The evaluation includes various metrics such as area under the curve (AUC), precision, recall, F-score, and specificity. Additionally, the article addresses the challenge of limited training samples by demonstrating effective system evaluation with fewer samples.
The results emphasize the effectiveness of transfer learning in scenarios with limited available medical imagery, and the paper's analytical discussion explores misclassification cases
The framework introduced by the author utilizes several pre-trained deep convolutional neural networks to extract intricate features from brain MR images. These extracted features are evaluated by diverse machine learning classifiers
To make predictions, the author selects the top three deep features with the best performance across multiple machine learning classifiers and combines them into a feature ensemble.
The proposed work applied a well-established deep learning architecture, the Convolutional Neural Network (CNN), to classify 3064 T1 weighted contrast-enhanced MR images of the brain into three distinct categories: gliomas, meningiomas, and pituitary tumors.
The results of the proposed CNN classifier are impressive, achieving a high accuracy of 98.93% and a sensitivity of 98.18% for cropped lesions, an outstanding 99% accuracy and a sensitivity of 98.52% for uncropped lesions, and a commendable 97.62% accuracy and a sensitivity of 97.40% for segmented lesion images.
In this study, researchers assessed the performance of three distinct convolutional neural network architectures, namely AlexNet, GoogLeNet, and VGGNet, for the classification of brain tumors, including glioma, pituitary tumors, and meningioma. They used the MRI brain tumor dataset available on Figshare and explored various transfer learning approaches, including both fine-tuning and freeze methods.
To improve the dataset's size, mitigate the risk of overfitting, and enhance the generalization of results, they applied data augmentation techniques to the MRI slices.
The results were particularly impressive when using the fine-tuned VGG16 architecture, achieving classification and detection accuracy levels of up to 98.69%.
The author introduced a multi-grade brain tumor classification system based on convolutional neural networks (CNNs). Initially, they employed a deep learning approach to identify tumor locations within an MR image.
|