Brain Tumor Detection and Classification Using Convolutional Neural Network (CNN)

Abstract

According to the World Health Organization, brain tumors are one of the leading causes of mortality globally. Early identification of this disease is difficult due to its intricacy and quiet character. Chronic brain tumor disease is linked to the risk of clinical occurrences, making it a serious public health issue around the world. Despite the fact that it is commonly acknowledged that chronic brain tumor disease has significant associations with increased risks of end-stage excretory organ disease, vascular occurrences, and all-cause mortality, there is still a scarcity of reliable data on individual individuals. For this brain tumor prediction challenge, we will utilize the deep learning-based Convolutional Neural Network (CNN) technique, which no one has used before in the research, especially on Image Dataset. CNN has been a popular method and highly sought-after model classification today. With input, neurons, hidden layers, and output, the CNN-based expert system works similarly to the human brain. Chronic brain tumor photos of healthy and unhealthy photographs were taken in good lighting circumstances for this study to discover any hidden features. The image samples are then processed using techniques such as Grayscale, B&W, Complement, Robert, Resize, and Power Transform. The chronic is then run through a Convolutional Neural Network texture feature extraction technique (CNN). Contrast, Correlation, Energy, Homogeneity, Entropy, Mean, Standard deviation, Variance, Skewness, and Kurtosis are the characteristics. The data is organized on a spreadsheet, which acts as a record, after feature extraction. Finally, there is one hidden layer, 16 input neurons, and two healthy or not outputs in a convolutional neural network. The data is divided into train and test datasets, with 70% of the data used for training, 10% for validation, and 20% for testing. The detection accuracy was 92.78 percent, with a 5.33-second execution time depending only on the number of iterations or epochs. For the confusion matrix of brain tumor detection and classification, an accuracy of 97.9% was recorded, a precision of 98.3% was accounted with a recall of 98.5%, and an AUC of 99.7% was calculated for this dedicated research work.