The Missing Coordinate in Surgery

Surgical Spatial Intelligence Series #1 · Version 1.0 · Published 2026-09-13

Why surgical spatial intelligence may begin with a map that knows when it is still valid

1. When the Coordinate System Changes

Modern surgery is rich in images, instruments, and data. What it often lacks is a persistent spatial relationship between them. Consider an 8-millimeter ground-glass nodule: on a preoperative CT scan, its location may be clear—right upper lobe, 12 millimeters below the pleural surface. Once the lung is collapsed, rotated, and manipulated, the image remains available and the lesion is still there; what has changed is the coordinate system that connected the two.

Surgeons already have ways to recover that information. Percutaneous markers, microcoils, fluorescence, navigational bronchoscopy, intraoperative ultrasound, and cone-beam CT can each restore part of the spatial picture, but every method comes with its own workflow, hardware, timing, and clinical constraints. This points to a broader opportunity: some of the most valuable information in the operating room is not another image but the coordinate that makes an image actionable.

We use the term surgical spatial intelligence to describe the capability to restore a usable spatial relationship between preoperative anatomy and the intraoperative patient—and to maintain that relationship when needed. At full maturity, this capability could become a continuously updating map of deformable anatomy. The first useful product does not need to begin at that endpoint.

2. Start with a Valid Moment

The most ambitious form of soft-tissue navigation is continuous. The organ moves with respiration, tissue deforms under traction, and the field changes as structures are mobilized, divided, or resected. A mature system would track those changes, update the spatial model, recover after temporary loss of view, and estimate when its own state is becoming unreliable. That is the destination; the first useful product can solve a narrower problem.

Before an irreversible step—resection, transection, ablation, or deep dissection—the surgeon needs to know where the target is, where the critical surrounding structures are, and how much confidence to place in that spatial relationship. A system that can provide those answers at the moment of decision already creates clinical value. Once the anatomy has been substantially altered, that spatial state can expire; the system can then reacquire observations, register again, and establish a new state.

This creates a practical maturity path: point-in-time localization → stage-wise re-localization → continuous spatial state estimation. The first stage establishes a reliable state at the moment it matters. The second extends that capability across several important phases of the procedure. The third maintains the relationship continuously. Clinical value can therefore begin before continuous navigation is solved.

3. Keeping the Map Valid

Preoperative imaging provides a rich prior; intraoperative sensing provides current evidence. CT, MRI, and three-dimensional models describe deep anatomy and global spatial relationships. Fluorescence, intraoperative ultrasound, and video provide something different: observations of the anatomy as it exists now. The preoperative model tells the system what it expects to find; intraoperative observations tell it whether those expectations still match reality.

A fluorescent boundary, an ultrasound landmark, or a recognizable surface structure can help confirm a registration. A growing mismatch can indicate that the spatial state is drifting. At that point, the right system behavior may be to lower confidence, request a new observation, re-register, or stop presenting guidance. In that sense, intraoperative sensing does more than inform the surgeon; it can also become an input to the spatial-state estimator.

This is why reliable soft-tissue navigation is fundamentally a state-management problem. The state includes position, deformation, anatomical relationships, uncertainty, and the conditions under which the current estimate should no longer be used. Clinical evidence reflects the difficulty of maintaining such a state: a 2021 systematic review of image-guided laparoscopic liver surgery reported navigation accuracy commonly in the range of 8–15 millimeters, while a 2026 clinical feasibility study using electromagnetic tracking reported a mean target registration error of 6.3 ± 3.8 millimeters and an average registration time of 11 ± 7 minutes.[6,7] The harder problem is not simply how to align a map, but how to know whether that alignment remains valid.

4. Where Today’s Navigation Stops

Navigation is already a mature clinical category, but the term covers several different capabilities. Rigid-anatomy systems in neurosurgery, spine surgery, and orthopedics maintain coordinates using relatively stable structures and fixed references. Instrument-tracking systems localize tools. Intraoperative imaging and fluorescence refresh information at discrete moments. All are useful forms of spatial support.

The harder problem begins when the anatomy itself becomes the moving reference. Instrument motion does not tell us whether the organ model is still correct, and a new image does not automatically maintain alignment with the preoperative model. Soft tissue introduces deformation, respiratory motion, traction, collapse, and eventually structural change. The gap is therefore not the absence of navigation, but the absence of a lightweight way to maintain trusted spatial information as soft-tissue anatomy changes.

5. Beyond Case Volume

The size of the opportunity depends on two different numbers: how many procedures could use spatial guidance, and how densely those procedures are distributed across hospitals. Using a deliberately narrow model focused on two non-overlapping tasks—localization of occult thoracic lesions and localization of deep abdominal lesions—we estimate a global opportunity boundary of approximately 532,000 procedures per year. This number describes the potential task pool, not a revenue forecast.

The second number—case density—may matter more for adoption. Surgical technologies with high fixed costs work best when enough procedures are concentrated within an institution to justify capital, training, maintenance, and workflow change. A 2026 systematic review of robotic-surgery economics found minimum annual caseload thresholds ranging from roughly 40 to 300 procedures, depending on the platform and procedure.[1] The exact thresholds do not define spatial-navigation economics, but they illustrate the underlying mechanism: fixed cost creates a volume requirement.

A system used 30 times a year is a very different purchasing proposition from one used every day. This changes what “low cost” should mean. The relevant cost structure includes capital requirements, per-case costs, registration time, specialized staff, dedicated sensors, and limits on reuse across procedures. Lowering those burdens can reduce the case density required to justify deployment. That is how lower cost can expand adoption.

6. Making Spatial Guidance Lighter

Much of the conversation around AI in surgery focuses on autonomy, while clinical adoption remains more incremental. A systematic review of 49 FDA-cleared surgical robotic systems between 2015 and 2023 found that 86% were Level 1 robot-assistance systems, while 6% reached Level 3 conditional autonomy.[2] For spatial guidance, the more immediate opportunity may be to reduce the work and hardware required to establish a trustworthy state.

Learning-based methods are beginning to replace selected elements of the navigation workflow: manual landmarks, explicit surface digitization, some registration steps, and, in selected settings, parts of the external sensing stack.[3,4,5] A 2026 study of navigated arthroscopy, for example, showed that SLAM could remove the need for a surface-digitizing probe, while an anatomical reference marker was still required to meet the navigation objective.[5] This points to a more practical design question than whether navigation can become fully sensorless: what is the minimum sufficient sensor set for the task?

The temporal requirement matters as well. If temporal coverage is limited to a critical decision window, the minimum sufficient sensor set may also be smaller than the one required for continuous non-rigid tracking. Together, these define two dimensions of system simplification: minimum sufficient sensing and minimum sufficient temporal coverage. The near-term opportunity for AI is therefore to reduce the cost and workflow burden of obtaining a trustworthy spatial state.

Conclusion

High-end navigation has already shown that hospitals value trusted spatial information. The next step may be to make trustworthy spatial states easier and cheaper to establish at the moments that matter, rather than solve continuous soft-tissue navigation in one leap.

Surgical spatial intelligence can begin with one valid moment, expand to several, and eventually become continuous. The first generation does not need a map that is always right. It needs a map that knows when it is still right. Surgical spatial intelligence can begin here.

References

1. Garavini AB, et al. Minimum caseload for cost-effective robotic-assisted surgery: a systematic review. J Robot Surg. 2026;20(1):216. PMID: 41634236.

2. Lee A, Baker TS, Bederson JB, Rapoport BI. Levels of autonomy in FDA-cleared surgical robots: a systematic review. NPJ Digit Med. 2024;7(1):103. PMID: 38671232.

3. Zhou J, Gao B, Wang K, Pei J, Heng PA, Qin J. Landmark-free preoperative-to-intraoperative registration in laparoscopic liver resection. IEEE Trans Med Imaging. 2025;44(11):4350–4362. PMID: 40424105.

4. Yang Y, et al. Markerless augmented reality registration for surgical guidance: a multi-anatomy clinical accuracy study. Int J Comput Assist Radiol Surg. 2026;21(7):1525–1532. PMID: 42250192.

5. Baptista T, Raposo C, Marques M, Vaz D, Antunes M, Barreto JP. Adapting monocular SLAM for surface-trace free 3D registration in navigated arthroscopy. Int J Comput Assist Radiol Surg. 2026;21(5):1099–1107. PMID: 42062640.

6. Schneider C, et al. Performance of image guided navigation in laparoscopic liver surgery — a systematic review. Surg Oncol. 2021;38:101637. PMID: 34358880.

7. Olthof KA, et al. Toward augmented reality in laparoscopic liver surgery using electromagnetic tracking: a clinical feasibility study. Surg Endosc. 2026;40(4):3538–3547. PMID: 42377473.

Revision History

Evidence update record: This is the anchor essay of the series. Sources and links will be added as Parts 02–05 are published; if any core judgment changes, it will be recorded here rather than silently revised.