Holographic-type communication is poised to revolutionize immersive telepresence and extended reality applications by enabling ultra-realistic, volumetric interactions. However, delivering high-fidelity 3D holographic content over bandwidth-constrained and variable wireless links presents significant challenges due to its inherently high data demands andreal-time requirements. This paper proposes a novel utility-aware adaptive streaming framework for segmented holographic video, wherein each frame is decomposed into semantically meaningful components (face, hands, and body pose), encoded at multiple resolution levels using Draco compression. The adaptive selection of segment resolutions is formulated as a 0-1 knapsack optimization problem, aiming to maximize perceived utility under dynamic bandwidth constraints. Segment utilities are modeled using diverse temporal decay functions (linear, exponential, and logarithmic) to capture differential importance over time. We implement and evaluate the full system in Network Simulator 3.40, integrating realistic network traces and application-level utilities. Experimental results demonstrate significant gains in bandwidth utilization, segment delivery completeness, and overall quality of experience, compared to non-adaptive and static strategies. The proposed approach represents a practical and extensible foundation for real-time holographic streaming in future 5G/6G networks.