The growing adoption of telemedicine creates urgent demand for reliable, high-performance network infrastructure capable of supporting heterogeneous real-time traffic - including HD video consultations, DICOM medical imaging, IoT telemonitoring streams, and electronic health records. This paper addresses the structural optimization of telemedicine networks deployed over Software-Defined Networking (SDN) infrastructure, where conventional maximum-flow algorithms prove inadequate due to high computational complexity and incompatibility with dynamic, free-oriented topologies. The aim of this work is to create the structural optimisation method using the Modified Iterative Algorithm (MIA), originally developed for flow distribution on free-oriented graphs, for the specific requirements of SDN-based telemedicine networks. The proposed method incorporates the following key enhancements: 1) optimized recursive path traversal prioritizing shortest ST-paths via a hop-growth strategy; 2) dynamic residual-capacity updates to prevent real-time channel overload; 3) automated detection and resolution of edge-intersection conflicts between concurrent paths. The verification was done in a Mininet-emulated SDN environment under two operational scenarios: normal telemedicine center load and stress test simulating urgent DICOM transfer. Theoretically, the study extends flow optimization methodology to free-oriented SDN graphs, eliminating the need for artificial edge duplication required by classical directed-graph approaches. Practically, the adapted MIA enables dynamic, controller-independent bandwidth allocation in TMN, ensuring QoS compliance for time-critical medical applications and opening pathways for integration with OpenFlow-compatible SDN controllers such as Ryu. Comparative evaluation against Ford-Fulkerson, ACO, and GA-based methods shows 7.8% throughput improvement over ACO with 58% lower control latency, while telemedicine-specific QoS assessment identifies both the method's suitability for video/telemonitoring/DICOM and its current limitation for haptic telesurgery (0.12% packet loss during redistribution vs. required <0.01%).