Loganayagi T, Sravani Meesala, Maram Balajee, Rao Telu Venkata Madhusudhana
Department of Electronics and Communication Engineering, Paavai Engineering College, Pachal, Namakkal, Tamilnadu, India.
Department of Computer Science and Engineering, GMR Institute of Technology, Rajam, Andhra Pradesh, India.
J Biotechnol. 2025 Sep;405:124-138. doi: 10.1016/j.jbiotec.2025.05.009. Epub 2025 May 9.
Brain tumor detection is essential to identify tumors at an early stage, allowing for more effective treatment. The patient's chances of recovery and survival can be improved by early detection. The existing methods for detecting brain tumour have several limitations, including limited accessibility, exposure to radiation, high costs and potential for false negatives. To overcome the issues, a Deep Maxout-Visual Geometry Group-16 (DM-VGG-16) model is devised for detecting tumour in brain from Magnetic Resonance Imaging (MRI). Initially, MRI image is sent for pre-processing as input. Here, Non-Local Mean (NLM) filter performs pre-processing. The pre-processed image is subjected to segmentation stage, which is accomplished by Template-based K-means and improved Fuzzy C Means algorithm (TKFCM). Moreover, in feature extraction stage, various features, like area, cluster prominence, Hybrid PCA- Normalized GIST (NGIST) and Improved Median binary Pattern (IMBP) are extracted. Lastly, proposed DM-VGG-16 model is utilized for detection of brain tumors from extracted features. The DM-VGG-16 is the integration of Deep Maxout Network (DMN) and Visual Geometry Group-16 (VGG-16). The DM-VGG-16 outperformed superior results than conventional techniques with performance metrics, including accuracy, True Negative Rate (TNR) and True Positive Rate (TPR) of 90.76 %, 90.65 % and 90.75 % correspondingly.
脑肿瘤检测对于早期识别肿瘤至关重要，有助于实现更有效的治疗。早期检测可提高患者的康复和生存几率。现有的脑肿瘤检测方法存在若干局限性，包括可及性有限、辐射暴露、成本高昂以及存在假阴性的可能性。为克服这些问题，设计了一种深度最大池化-视觉几何组16（DM-VGG-16）模型，用于从磁共振成像（MRI）中检测脑肿瘤。首先，将MRI图像作为输入进行预处理。在此，非局部均值（NLM）滤波器执行预处理。预处理后的图像进入分割阶段，该阶段通过基于模板的K均值和改进的模糊C均值算法（TKFCM）完成。此外，在特征提取阶段，提取各种特征，如面积、聚类显著性、混合主成分分析-归一化全局描述符（NGIST）和改进的中值二进制模式（IMBP）。最后，利用所提出的DM-VGG-16模型从提取的特征中检测脑肿瘤。DM-VGG-16是深度最大池化网络（DMN）和视觉几何组16（VGG-16）的集成。DM-VGG-16在性能指标方面优于传统技术，其准确率、真阴性率（TNR）和真阳性率（TPR）分别为90.76%、90.65%和90.75%。
