.png){kind=icon}
.png){kind=icon}
Return
Brain tumor detection using deep ensemble model with wavelet features
.png){kind=icon}
PRE
AI
DOI：10.1007/s12553-022-00699-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Purpose A person's healthy activities are determined by the state of his or her brain. The brain is in charge of all of a person's activities. If a small abnormality develops in the brain, it will have a negative impact on the person regardless of whether the other organs are in good condition. As a result, early detection of any abnormal growth in the brain is essential. Methods In this work, the authors have utilized data pre-processing using discrete wavelet transform (DWT) and segmentation, whereas, for detection, an ensemble learning technique is proposed. DWT and segmentation help in increasing the dataset size that is used to train the deep learning model. Segmentation using supervised Auto-encoder (AE) is used for data enhancement to strengthen the training process. The original data, outputs of DWT, and segmented images are utilized for the training of the ensemble model designed with three parallel-connected convolutional neural networks (CNNs). Results The detection results obtained from the ensemble of these recurrent models are then passed through the Multilayer Perceptron (MLP) for final detection. Kaggle brain MRI image dataset is used to complete the proposed method. Test accuracy, F1-score, precision, sensitivity, and specificity provided by this method are 98.08%, 0.9836, 1.0000, 0.9677, and 1.0000 respectively. In comparison to state-of-the-art models, the proposed model produces competitive outcomes. Conclusion In time detection of the tumor may lead to the survival of the patient. Automatic and accurate detection is another perspective of this field. For this purpose, we have proposed a deep ensemble model with wavelet features. The ensemble model provides increased performance in comparison to single models due to the parallel training.
Keywords:
Brain tumor detection
Discrete wavelet transform
Supervised autoencoder
Convolutional neural network
Ensemble learning
Journal
H
IF:
4
Papers:
1.8K
Citations:
5.6K