Stereotactic Targeting

Evidence status. Technical principles are mature, but platform performance depends on the complete workflow. Published accuracy values are not interchangeable when studies use different targets, error definitions, imaging, and verification methods.

1.AC-PC coordinates are a language, not the anatomy

The anterior commissure-posterior commissure line standardizes orientation and permits proportional or fixed-offset coordinates. The midsagittal plane provides the third reference. This is invaluable for communication and a defensible fallback when a structure is not visible, but interindividual variability makes atlas coordinates a starting hypothesis.

Record exactly how AC and PC were defined, whether the line connects centers or borders, which image set supplied the landmarks, and how the plan was transformed. Name the origin explicitly: the Schaltenbrand convention places it at the midcommissural point, the Talairach convention at the anterior commissure, and a y-coordinate quoted without its origin is uninterpretable, because the two differ by half the intercommissural distance. State the sign convention for laterality, anterior-posterior, and superior-inferior in the same breath. Small convention differences become millimeters at the target; a mistaken origin becomes a centimeter.

2.Atlas, indirect, direct, and connectivity-based targeting

Indirect targeting places a point relative to AC-PC or ventricular landmarks. Direct targeting identifies the nucleus or tract on patient-specific MRI, which means naming the sequence, because direct visualization depends on the contrast mechanism and not on the word MRI. A program should be able to state its field strength, its sequence, its voxel size, and who reads it. Atlas overlays provide context but inherit the atlas specimen, normalization method, and deformation error. Connectivity-based targeting estimates a network relationship and adds a new layer of uncertainty.

Indirect targeting is only usable if you can state a starting point. The conventional subthalamic set is x = 11–13 mm from midline, y = 2–4 mm posterior to the midcommissural point, z = 4–5 mm below the intercommissural plane. It is a starting hypothesis to be reconciled with direct visualization, not a destination, and meaningless if quoted without that origin. For GPi and Vim, document the institution’s starting convention and patient-specific adjustments. Published coordinates differ by atlas, origin, and target subregion, so a coordinate must always be reconciled with anatomy.

3.Frame-based stereotaxy

A rigid frame couples the skull to a coordinate system and constrains the arc or guide. Strengths include a short transformation chain, continuous rigid fixation, extensive clinical experience, and compatibility with awake microelectrode workflows. Costs include patient discomfort, pin risks, imaging logistics, limited repositioning, and a compact planning window.

Frame accuracy still depends on image acquisition, fiducial localization, arc calibration, target definition, and instrument behavior. Frame placement should preserve the intended entry and avoid collision with shoulders, scanner, or arc.

4.Frameless and robotic workflows

Frameless platforms register the patient's skull to preoperative imaging using bone fiducials, surface matching, or other localization. Robots then orient an instrument guide along the trajectory; they do not autonomously choose the target. Advantages include flexible scheduling, multiple trajectories, ergonomic access, and efficient sEEG implantation. The price is a longer and sometimes less visible registration chain.

Bone fiducials usually outperform skin fiducials for rigid cranial work; Widmann and colleagues' survey of frameless targeting devices is the clearest account of where in each platform the error actually enters, and of why published millimeters are not comparable across devices. Surface matching can be accurate when acquisition and exposure are disciplined, but it deserves an explicit acceptance threshold and rescue plan. New fiducial-less systems should be judged by independently measured clinical error at the target, not by convenience alone; their published evidence is still largely single-center experience, such as one six-year series of frameless, fiducial-less DBS in Parkinson disease. Across platforms, the meta-analysis of 6056 trajectories by Philipp and colleagues reported lower target error with robotic assistance, but its included studies use different error definitions and different verification imaging, so read the pooled figure as a summary of heterogeneous measurements rather than as one quantity.

5.Trajectory is part of the target

Select an entry that avoids sulci, cortical veins, ventricles, eloquent cortex, and visible vessels while respecting lead geometry and hardware routing. Avoid sulci because the vessels are in them: a transsulcal trajectory is the classic route to a symptomatic hemorrhage. Avoid ventricles for different reasons: CSF egress and brain shift, and deflection of the cannula at the ependymal interface, which degrade accuracy without necessarily causing bleeding. When the entry point is constrained, weigh the two separately. For DBS, trajectory also determines which structures are sampled by microelectrode recording and which contacts span the target.

Physiology answers a question anatomy cannot: it identifies the sensorimotor territory within the nucleus, which is where a lesion or lead belongs, and it does so along the trajectory you actually chose. Firing rate and pattern transitions mark nuclear borders, and a transition well away from the expected depth is information about your trajectory rather than noise. Macrostimulation through the same track then defines the side-effect margin before anything permanent is done.

Use vascular imaging appropriate to the operation, inspect the entire path in orthogonal and probe-eye views, and document the minimum vessel clearance policy. A safe path should remain safe after plausible fusion and registration error.

6.Accuracy has named measurements

Target registration error estimates registration performance at a point not necessarily used to fit the transformation. Entry-point error and target-point error compare the postoperative device with the plan. Radial error is perpendicular to the planned path; depth error is parallel to it. Euclidean error combines them but can hide the clinically relevant direction.

7.Close the loop

Asleep and awake are two verification strategies, not two tribes. Awake workflows commonly use clinical testing and physiology; asleep workflows rely heavily on imaging and may also use electrophysiology. Imaging verification can complement either approach. Holl and colleagues showed what a disciplined frame-based, image-verified chain achieves without physiological mapping, and pooled comparisons to date have not established the superiority of either strategy for motor outcome, limited as they are by heterogeneous targets, error definitions, and follow-up. Choose the strategy your program can audit, and state which verification method you are relying on before you quote your accuracy.

Verify hardware with intraoperative imaging (intraoperative CT or flat-panel imaging fused to the plan, or interventional MRI, which permits updated targeting and trajectory verification during surgery), postoperative CT/MRI, electrophysiology, clinical testing, or a combination matched to risk. Fuse postoperative imaging to the preoperative plan, inspect hemorrhage and pneumocephalus, localize contacts, and explain discrepancies. Verification is not merely complication screening; it is how a program learns its systematic bias.

Maintain a prospective accuracy log by platform, procedure, surgeon, registration method, trajectory, and imaging workflow. Review outliers and trends. Calibration without clinical audit tests the machine; audit tests the system.