Brain tumors, marked by abnormal skull tissue growth, pose a significant threat, causing loss of lives annually. Magnetic Resonance Imaging (MRI) is commonly used technique for brain cancer detection. The crucial step of MRI segmentation serves various clinical purposes in neurology, including quantitative analysis and functional imaging. Image thresholding, a direct and efficient segmentation method, holds significant importance, especially in real-time applications for pixel classification. The selection of threshold values plays a crucial role in determining the accuracy and effectiveness of medical image segmentation. Traditional multilevel thresholding is computationally inefficient, struggling with complex images and lacking adaptability compared to advanced techniques. Recognizing these limitations, the adoption of metaheuristic algorithms enhances the efficiency of multilevel image segmentation. Therefore, this study presents an improved variant of the human-inspired metaheuristic known as the “heap-based optimizer,” integrating a memory-saving strategy called the “Memory Saving Heap-Based Optimizer (MS-HBO).” A new method for multilevel image thresholding is then presented, which combines Type-II Fuzzy sets with the recently developed MS-HBO variant. This integrated method aims to enhance efficiency and effectiveness in image thresholding, showcasing the potential of combining memory-saving strategies with advanced metaheuristic algorithms. The proposed method's performance is benchmarked on eight brain tumor MRI images, measuring PSNR, SSIM, and MSE at levels 2, 4, 6, and 8, and compared to eight standards and newly published metaheuristics. The experimental results confirm the strong optimization of MS-HBO, proving its superiority in multilevel thresholding segmentation over comparison methods across diverse performance metrics.