A new local tetra pattern in composite planes (LTcP) technique for classifying brain tumors using partial least squares and super-pixel segmentation

The classification of magnetic resonance imaging (MRI) images of brain tumors is challenging, and it is also important, as the number of cases of brain tumors is projected to increase by 1.2% in the year 2030 from the current scenario. This projection makes it necessary to explore medical image classification. The classification involves the extraction of features and the segmentation of the images. The overall accuracy depends upon the type and number of features attained from the image. Thus, we propose a novel descriptor called Local Tetra patterns, which is based on the composite plane (LTcP). The proposed LTcP method helps to extract the maximum number of features from the image. The MRI images are segmented via superpixel segmentation followed by LTcP to obtain the maximum number of features from the image. The 3D directions that are in the cardinalities of xy, yz, and xz are used for primary division of the images into three composite planes. The orientations of the neighborhood and center pixels in both the vertical and horizontal directions were identified via first-order derivatives. We then fused the selected optimal features via the partial least squares (PLS) approach, which in turn helps reduce the number of features required for accurate classification. These features are chosen via the filter method. The proposed approach can classify images with 42 features gained from LTcP, which is better than the existing techniques. The fused features were fed to the SVM classifier, which yielded an accuracy of 91.4%. The computational time for 35 features was observed to be 94.4 s for SVM, which itself proves that the proposed method has a fast computational speed. The F1-score and Recall for the proposed method were observed to be 95.90 and 96.80, respectively. It depicts that the approach can identify a reasonable number of true positives.