Intraoperative tracer guided surgery extends surgical perception beyond the limits of direct visualization. Yet, the spatial value of these signals remains largely unrealized. Fluorescent and radioactive tracers are currently displayed as near flat overlays or numbers. Surgeons compensate by maintaining a mental model of the surgical cavity, constructed from preoperative imaging and carried through the entire procedure from memory. This cognitive burden is both substantial and structurally unresolved, with positive surgical margins, the presence of tumor cells at the edge of resected tissue (indicating incomplete removal), described as frequent and margin relocation during procedures characterized as “almost impossible.”
This project identifies the absence of a real-time geometric model of the surgical cavity as the foundational design problem underlying current tracer visualization limitations. Two interconnected challenges are addressed. First, a reconstruction pipeline was developed combining monocular depth estimation via the Depth Anything V2 AI with metric grounding from a time of flight sensor, producing a 3D surface model. Second, a set of visualization concepts was developed to register tracer signals within that spatial model, replacing mental reconstruction with designed spatial representations.
The selected visualization output applies distance corrected fluorescence intensity normalization, using per-pixel depth measurements from the reconstruction to remove the systematic intensity bias introduced by variable camera distance. Together, these outputs constitute an initial validated prototype of a design pipeline that treats spatial legibility of intraoperative tracer data as a representation problem, and proposes geometry as its solution.
... Read more