Analysis of Unsharp Mask Technique Application on Clinical Stroke MRI Brain Examination using Arterial Spin Labeling (ASL) Sequence in Image Information

Magnetic Resonance Imaging (MRI) with Arterial Spin Labeling (ASL) is a non-invasive technique commonly used to assess cerebral perfusion, especially in stroke patients. However, ASL images often suffer from low contrast and high noise, which can hinder diagnostic accuracy in visualizing perfusion areas and detecting ischemic lesions. Image enhancement techniques, such as the unsharp mask, offer a potential solution to improve image quality. The effectiveness of this enhancement depends on the kernel size used in the unsharp mask filter. This study evaluates the impact of different kernel sizes (3×3, 5×5, and 7×7) on the quality of ASL brain images, focusing on both quantitative and qualitative improvements. A total of 63 ASL brain MRI images from stroke patients were processed using unsharp mask filters with the three kernel sizes. Quantitative analysis measured Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR), while qualitative assessment involved three radiologists independently evaluating five aspects of image quality: perfusion area clarity, grey-white matter contrast, ischemic lesion boundary visibility, noise level, and overall visual quality. Statistical tests, including Friedman and Wilcoxon, were applied to compare results across the kernel sizes. Results revealed that the 3×3 kernel achieved the best results in both quantitative and qualitative assessments, with the highest SNR, CNR, and visual quality scores. Significant differences (p < 0.05) were found between kernel sizes, confirming the superiority of the 3×3 kernel. The 7×7 kernel reduced noise but caused oversmoothing, negatively impacting image sharpness. In conclusion, the 3×3 kernel provides an optimal balance between noise reduction and edge preservation, enhancing ASL brain image quality for stroke diagnosis.