An Efficient and Robust Ischemic Stroke Detection Using a Combination of Convolutional Neural Network (CNN) and Kernel K-Means Clustering

This study introduces a combined approach utilizing the widely-used Convolutional Neural Network (CNN) and Kernel K-Means clustering method for the detection of ischemic stroke from Magnetic Resonance Imaging (MRI) images. We propose an efficient and robust alternating classification scheme to overcome the challenges of extensive computation time and noisy ischemic stroke images obtained from Cipto Mangunkusumo Hospital in Indonesia. The method incorporates multiple convolutional layers from the CNN architecture and subsequently vectorizes the matrix output to serve as input for Kernel K-Means clustering. Through a series of experiments, our proposed method has demonstrated highly promising results. Employing 11-fold cross-validation and the RBF kernel function (sigma= 0.05), we achieved exceptional performance metrics, including 99% accuracy, 100% sensitivity, 98% precision, 98.04% specificity, and 98.99% F1-Score. These outcomes underscore the remarkable capabilities of the combined CNN and Kernel K-Means clustering approach in accurately identifying ischemic stroke cases. Furthermore, our method exhibits competitive performance when compared to several other state-of-the-art methods in the field of deep learning. By harnessing the power of CNN's convolutional layers and the clustering capability of Kernel K-Means, we have achieved significant advancements in the domain of ischemic stroke detection from MRI images. The implications of this research are substantial. By enhancing the accuracy and efficiency of ischemic stroke detection, our method has the potential to assist medical professionals in making timely and informed decisions for stroke patients. Early detection and intervention can greatly improve patient outcomes and contribute to more effective treatment strategies.