Edge impulse-based convolutional neural network for brain tumor classification

Brain tumor, a type of intracranial tumor, poses significant threats to human health. In an effort to improve the accuracy and efficiency of tumor recognition and classification, this paper used convolutional neural network to train the model through the Edge Impulse platform. In this paper, BrainChip's Keras-based classification model was selected as the learning block. The initial data set was broken down into training set and test set categories for data processing. Data in the training set made up 88% of the overall data set, whereas data in the test set made up 12% of the total data set. In this paper, the number of training epochs was set as 10. In addition, under the condition that other conditions remain unchanged, this study conducted the experiment with the learning rates of 0.0005, 0.001, 0.002 and 0.003 respectively. After training models with different learning rates through convolutional neural network, four groups of results with different accuracy were obtained in this study. Finally, this study's experiment produced the most accurate results (87.0%) when the learning rate was 0.002. The findings of this study suggest that training a CNN model can effectively and accurately identify different types of brain tumors in a large dataset.