Gokapay Dilip Kumar, Mohanty Sachi Nandan
School of Computer Science & Engineering (SCOPE), VIT-AP University, Amaravati, Andhra Pradesh, India.
Digit Health. 2024 Dec 18;10:20552076241305282. doi: 10.1177/20552076241305282. eCollection 2024 Jan-Dec.
Brain tumors are abnormal growths of brain cells that are typically diagnosed via magnetic resonance imaging (MRI), which helps to discriminate between malignant and benign tumors. Using MRI image analysis, tumor sites have been identified and classified into four distinct tumor categories: meningioma, glioma, not tumor, and pituitary. If a brain tumor is not detected in its early stages, it could progress to a severe level or cause death. Therefore, to address these issues, the proposed approach uses an efficient classifier based on deep learning for brain tumor detection.
This article describes the classification and detection of brain tumor by an efficient two-channel convolutional neural network. The input image is initially rotated during the augmentation stage. Morphological operations, thresholding, and region filling are then used in the pre-processing stage. The output is then segmented using the Berkeley Wavelet Transform. A two-channel convolutional neural network is used to extract features from segmented objects. In the end, the most effective deep neural network is employed to determine the features of brain tumors. The classifier will utilize the Enhanced Serval Optimization Algorithm to determine the optimal gain parameters. MATLAB serves as the platform of choice for implementing the suggested model.
Several performance metrics are calculated to assess the proposed brain tumor detection method, such as accuracy, F measures, kappa, precision, sensitivity, and specificity. The proposed model has a 98.8% detection accuracy for brain tumors.
The evaluation shows that the suggested strategy has produced the best results.
脑肿瘤是脑细胞的异常生长，通常通过磁共振成像（MRI）进行诊断，MRI有助于区分恶性和良性肿瘤。通过MRI图像分析，已识别出肿瘤部位并将其分为四种不同的肿瘤类别：脑膜瘤、胶质瘤、非肿瘤和垂体瘤。如果脑肿瘤在早期未被检测到，可能会发展到严重程度或导致死亡。因此，为了解决这些问题，所提出的方法使用基于深度学习的高效分类器进行脑肿瘤检测。
本文描述了一种通过高效的双通道卷积神经网络对脑肿瘤进行分类和检测的方法。在增强阶段，输入图像首先进行旋转。然后在预处理阶段使用形态学操作、阈值处理和区域填充。接着使用伯克利小波变换对输出进行分割。使用双通道卷积神经网络从分割后的对象中提取特征。最后，采用最有效的深度神经网络来确定脑肿瘤的特征。分类器将利用增强的薮猫优化算法来确定最佳增益参数。MATLAB是实现所建议模型的首选平台。
计算了几个性能指标来评估所提出的脑肿瘤检测方法，如准确率、F值、kappa值、精确率、灵敏度和特异性。所提出的模型对脑肿瘤的检测准确率为98.8%。
评估表明所建议的策略产生了最佳结果。
