Connectome-Informed Single-Isocenter Multi-Target SRS Planning: A Pilot Feasibility Study
Abstract
Purpose
To test the feasibility of connectome-informed single-isocenter multi-target SRS planning, incorporating an individual’s functional networks and major white-matter pathways as avoidance structures while maintaining target coverage and standard anatomic OAR constraints.
Methods
Four adults with multiple brain metastases, without prior history or radiation nor surgery, treated with cranial fSRS (9Gy×3) were enrolled in the study. Each patient underwent baseline and post-treatment (4–6 months) imaging on the same 3T scanner. MRI included T1-MPRAGE, T2-FLAIR, resting-state fMRI (~13 min), and DTI (b=0/1000, 30 directions). Patient-specific structural and functional connectomes were derived using DSI Studio, CONN toolbox, and the Quicktome platform (Omniscient Neurotechnology). Seven canonical functional networks (default mode, salience, central executive, dorsal attention, sensorimotor, visual, and limbic) were generated and co-registered to the planning CT. Two plans were created for each patient: (A) standard anatomy-based OAR planning and (B) connectome-informed planning, using network hubs and major white-matter pathways as additional avoidance structures. Patients were treated with the standard plan; comparative dosimetry was extracted from both plans. All connectome-guided plans underwent physics QA and multidisciplinary tumor board review to confirm deliverability and oncologic criteria. Longitudinal graph-theory metrics (global, nodal, and network-specific) were computed relative to the baseline to quantify post-RT network alterations.
Results
Feasibility was demonstrated in a representative case with two metastases adjacent to default mode network (DMN) pathways. Incorporation of individualized DMN structures reduced DMN dose while maintaining plan quality: PTV1_D99.9%=98.68%/99.19%, PTV2_D99.9% = 97.89/98.92%, brain_V20Gy=5.7/5.8cc, brainstem_D0.5cc=4.57/4.91Gy, DMN_D0.2cc=34.7/31.7Gy, and DMN_Dmean=3.9/3.4Gy (standard/connectome-guided), with preferential dose fall-off along DMN tracts. Early longitudinal connectomics suggested reduced global efficiency (0.556/0.507) and degree (34/19) with increased path length (2.068/2.195), clustering (0.399/0.567), and local efficiency (0.695/0.783) from baseline to follow-up.
Conclusion
Patient-specific connectome-informed SRS planning was technically feasible, QA-deliverable, and clinically acceptable, and it enabled selective sparing of functional networks beyond anatomy-based OAR optimization.