1. Least complex oLSVN-based computer-aided healthcare system for brain tumor detection using MRI images
- Author
-
Razzaq, Saqlain, Asghar, Muhammad Adeel, Wakeel, Abdul, and Bilal, Muhammad
- Abstract
Brain tumors are the most common and vigorous cause of death in the modern era. The medical community is working hard to develop effective methods to detect brain tumors in an early stage. Machine learning-based optimized classifiers can provide an efficient, accurate, and timely solution to detect brain tumors. Herein, a three-step least complex optimal linear support vector network-based computer-aided healthcare system for tumor cell detection using magnetic resonance imaging (MRI) is proposed. In the first step, features obtained from the Handcrafted features (HF) and a 14-layered convolutional neural network (CNN) operating in parallel are concatenated. Initially, these combined features are used for tumor classification. In the second step, to reduce the computational complexity, the bag of feature vector (BoFV) technique followed by principal component analysis (PCA) is introduced to select quality features. As this research focuses on the early-stage detection of brain tumors, an optimized linear support vector network (oLSVN) was introduced for classification in the third step. oLSVN sends tumors-classified images for segmentation to detect the exact area of the tumors, whereas the images in which the tumor is not detected due to poor visibility and noise undergo contrast-limited adaptive histogram equalization (CLAHE) process for noise filtration and image enhancement. These enhanced images are classified again for brain tumor detection in an early stage. A comparative analysis shows that the proposed model outperforms some already existing models. The execution time of the proposed model is 1.32seconds with 98.25%accuracy. As compared to some already existing approaches, the proposed model has an F1-Score of 98.27%,precision of 97.28%, specificity of 97.22%, and a Mathew's Correlation Coefficient of 96.52%.These results validate that the proposed state-of-the-art methodology can thus help the medical industry in the timely and efficient detection of brain tumors.
- Published
- 2023
- Full Text
- View/download PDF