Liu Tingting, Yuan Zhi, Wu Li, Badami Benjamin
Department of Oncology - Cardiology, Affiliated Tumor Hospital, Xinjiang Medical University, Urumqi, Xinjiang, China.
Engineering Research Center of Renewable Energy Power Generation and Grid-Connected Control, Ministry of Education, Xinjiang University, Urumqi, Xinjiang, China.
Proc Inst Mech Eng H. 2021 Apr;235(4):459-469. doi: 10.1177/0954411920987964. Epub 2021 Jan 13.
Precise and timely detection of brain tumor area has a very high effect on the selection of medical care, its success rate and following the disease process during treatment. Existing algorithms for brain tumor diagnosis have problems in terms of better performance on various brain images with different qualities, low sensitivity of the results to the parameters introduced in the algorithm and also reliable diagnosis of tumors in the early stages of formation. A computer aided system is proposed in this research for automatic brain tumors diagnosis. The method includes four main parts: pre-processing and segmentation techniques, features extraction and final categorization. Gray-level co-occurrence matrix (GLCM) and Discrete Wavelet Transform (DWT) were applied for characteristic extraction of the MR images which are then injected to an optimized convolutional neural network (CNN) for the final diagnosis. The CNN is optimized by a new design of Sparrow Search Algorithm classification (ESSA). Finally, a comparison of the results of the method with three state of the art technique on the Whole Brain Atlas (WBA) database to show its higher efficiency.
精确且及时地检测脑肿瘤区域对于医疗护理的选择、其成功率以及治疗过程中对疾病进程的跟踪具有非常高的影响。现有的脑肿瘤诊断算法在对具有不同质量的各种脑图像表现出更好的性能、结果对算法中引入的参数的低敏感性以及在肿瘤形成早期阶段的可靠诊断方面存在问题。本研究提出了一种用于自动诊断脑肿瘤的计算机辅助系统。该方法包括四个主要部分：预处理和分割技术、特征提取以及最终分类。灰度共生矩阵（GLCM）和离散小波变换（DWT）被应用于磁共振图像的特征提取，然后将其输入到优化的卷积神经网络（CNN）进行最终诊断。通过麻雀搜索算法分类（ESSA）的新设计对CNN进行优化。最后，在全脑图谱（WBA）数据库上，将该方法的结果与三种现有技术进行比较，以显示其更高的效率。
