This paper investigates the application of Neural Radiance Fields (NeRF) for three-dimensional reconstruction of brain MRI scans. Conventional MRI data is stored as sequential two-dimensional slices, requiring manual interpretation and reconstruction—an approach that introduces inefficiencies and potential diagnostic limitations. To address this, the study follows a two-phase implementation strategy. First, a minimal NeRF model was developed from scratch using a custom multilayer perceptron (MLP) to build a foundational understanding of volumetric reconstruction. In the second phase, the Instant-NGP framework was employed to enable more advanced and efficient neural rendering. MRI data from the BraTS 2021 dataset (FLAIR modality) was converted into PNG slices and pre-processed to ensure spatial coherence and normalized contrast. Training was performed on a Windows-based system with limited hardware, which imposed constraints on resolution and overall model stability. Although the training process achieved convergence in loss metrics, the reconstructed outputs lacked sufficient anatomical detail for clinical interpretation. The paper concludes that while Instant-NGP offers potential for medical volumetric reconstruction, further domain-specific adaptation and access to more robust computational resources are critical for clinical applicability.