Brain tumor detection remains one of the most pressing challenges in medical image analysis due to the complexity and variability of tumor structures. This study presents an automated approach for analyzing magnetic resonance imaging (MRI) scans, aimed at identifying pathological cases and localizing abnormal regions. A hybrid classification model based on support vector machines and k-nearest neighbors is used to distinguish between normal and pathological images. For segmentation, a superpixel-based method is applied to highlight tumor areas. The system combines traditional image preprocessing with statistical and textural feature extraction to enhance diagnostic accuracy. Experimental results confirm the effectiveness of the proposed two-stage pipeline in supporting early diagnosis and reducing the cognitive workload of clinicians.