
Full text loading...
Visual descriptor methods like Local Binary Pattern (LBP) capture anatomical structures in captured images along with their disparities, which can be exploited by suitable methods for the diagnosis of medical anomalies. In our study, we have proposed a Local Mean Gradient Pattern (LMGP), based partly on LBP, a feature extraction algorithm for the classification of Computed Tomography (CT) images of the brain into normal, ischemic, or hemorrhage categories.
The AISD and Kaggle datasets containing patients’ brain CT scan images [acute ischemic stroke, hemorrhage, and normal cases] were taken. Initially, adaptive histogram equalization (AHE) techniques were applied as preprocessing operations to enhance the quality of the CT images. Furthermore, features were extracted from the preprocessed data using several feature extraction techniques, including our proposed LMGP feature descriptor. The features were then scaled using the standard scaling technique. Subsequently, preprocessed images were fed into different classifiers to build models for classifying brain CT scan images.
The effectiveness of our methodology LMGP was determined using different metrics, such as recall, precision, F1 score, logarithmic loss, accuracy (ACC), and area under the curve (ROC). Conclusively, LMGP performed best when the RBF-SVM classifier was used for the classification and gave an accuracy of 94% and 96% in the case of five-fold and ten-fold cross-validation, respectively.
LMGP offers a distinctive and robust method of feature extraction from CT scan images by combining local information along with gradient change in pixels of the image. The efficacy of our proposed methodology (LMGP) was evaluated by using distinct classifiers, and the results were compared with eight different feature extraction methods. Overall, LMGP effectively outperformed all other feature descriptor methods in this study.