Brain strokes are serious health conditions when blood flow to a specific brain region is blocked or disrupted. This obstructs the delivery of oxygen and nutrients to the impacted brain cells. Accelerated neuronal degeneration may occur consequently, and if the problem is not promptly addressed, it might exacerbate, leading to significant impairment or mortality. Computed tomography (CT) scan is a crucial instrument for stroke diagnosis as it enables rapid visualization of the brain for physicians. Various image pre-processing techniques could significantly enhance CT scans, hence facilitating more precise stroke detection. To determine how various image pre-processing methods affect the precision of CT-scan stroke identification, this study compares them. Various methods are tested on a dataset consisting of CT images. These methods encompass noise reduction, contrast enhancement, edge detection, and picture normalization, among others. Stroke detection accuracy is improved by evaluating each pre-processing strategy using modern deep learning models. Results show that deep learning models perform better when it comes to stroke identification when certain pre-processing approaches are used to improve the clarity and diagnostic utility of CT images. To improve clinical outcomes, this comparative study sheds light on the best pre-processing procedures for improving stroke recognition in CT imaging.