Accurate measurement of brain tumor volume plays a vital role in diagnosis, treatment planning, and monitoring therapeutic response. In this study, a hybrid framework is presented that integrates deep learning-based segmentation with mathematical modeling for improved tumor volume estimation. Tumor regions are first segmented from MRI scans using a pre-trained 3D Enhanced U-Net, which provides reliable boundary delineation. The segmented output is then processed using an Ellipsoid Approximation (EA) method enhanced with a Shape Correction Factor (SCF). Unlike conventional ellipsoid-based estimations that assume regular geometry, the SCF incorporates ratios of the orthogonal dimensions of the tumor to account for asymmetry and irregular morphology. This adjustment enables a more realistic and accurate volume estimation, particularly for tumors with complex shapes commonly observed in clinical imaging. Experimental results demonstrate that the integration of the 3D Enhanced U-Net model for segmentation with the EA-SCF method yields superior accuracy compared to conventional methods, especially for tumors with complex morphologies.