Stroke detection represents a critical challenge in modern neuroimaging, with significant implications for timely medical intervention. This study introduces StrokeNet, a novel multi-path 3D-2D Convolutional Neural Network (CNN) architecture designed to enhance stroke detection through comprehensive volumetric and multi-view analysis of Computed Tomography (CT) scans. By integrating three-dimensional feature extraction with multi-perspective 2D representations, our approach addresses fundamental limitations in existing diagnostic methodologies. The proposed framework demonstrates superior performance in binary stroke classification, achieving 90.59% accuracy. Experimental validation across diverse patient datasets reveals StrokeNet's robust generalization capabilities, offering a promising approach for automated stroke detection.