Kotti Jayasri, Moovendran Manikandan, Kandasamy Mekala
Department of Information Technology, GMR Institute of Technology, Rajam, Andhra Pradesh, India.
Department of Computational Intelligence, Faculty of Engineering and Technology, School of Computing, SRM Institute of Science and Technology, Chennai, India.
Network. 2024 Apr 26:1-32. doi: 10.1080/0954898X.2024.2343342.
An innovative multi-level BT classification approach based on deep learning has been proposed in this article. Here, classification is accomplished using the SpinalNet, whose structure is optimized by the Hybrid Coot Flamingo Search Optimization Algorithm (CootFSOA). Further, a novel segmentation approach using CootFSOA-LinkNet is devised for isolating the tumour area from the brain image. Here, the input MRI images are fed into the Adaptive Kalman Filter (AKF) to denoise the image. In the segmentation process, LinkNet is used to separate the tumour region from the MRI image. CootFSOA is used to achieve structural optimization of LinkNet. The segmented image is then used to create several features, and the resulting feature vector is fed into SpinalNet to detect BT. CootFSOA is used in this instance as well to adjust the SpinalNet's hyperparameters and achieve high detection accuracy. If a tumour is detected, second-level classification is carried out by employing the CootFSOA-SpinalNet to classify the input image into several types, such as gliomas, pituitary tumours, and meningiomas. Moreover, the efficacy of the CootFSOA-SpinalNet has been examined considering accuracy, True Positive Rate (TPR), and True Negative Rate (TNR) and has recorded superior values of 0.926, 0.931, and 0.925, respectively.
本文提出了一种基于深度学习的创新型多层次脑肿瘤分类方法。在此，使用SpinalNet完成分类，其结构通过混合库特火烈鸟搜索优化算法（CootFSOA）进行了优化。此外，还设计了一种使用CootFSOA-LinkNet的新型分割方法，用于从脑部图像中分离出肿瘤区域。在此，将输入的磁共振成像（MRI）图像输入到自适应卡尔曼滤波器（AKF）中对图像进行去噪。在分割过程中，使用LinkNet从MRI图像中分离出肿瘤区域。CootFSOA用于实现LinkNet的结构优化。然后，将分割后的图像用于创建多个特征，并将得到的特征向量输入到SpinalNet中以检测脑肿瘤。在这种情况下，也使用CootFSOA来调整SpinalNet的超参数并实现高检测准确率。如果检测到肿瘤，则通过使用CootFSOA-SpinalNet对输入图像进行二级分类，将其分为胶质瘤、垂体瘤和脑膜瘤等几种类型。此外，还从准确率、真阳性率（TPR）和真阴性率（TNR）方面对CootFSOA-SpinalNet的效能进行了检验，其分别记录了0.926、0.931和0.925的优异值。
