Heuristic Transducer Placement¶

The function transducer_positioning identifies heuristic locations for transducer placement. It can be called with transducer_positioning_start.

The function targets brain regions specified by MNI coordinates, converting them to subject-native space using SimNIBS tools. It aims to identify candidate transducer positions where the geometric focus aligns with a target within a user-defined focal distance range. For each position, it computes the intersection proportion between the transducer and skin, mean/variance distances to skin and skull, and geometric focus/exit plane positions. It (optionally removes) areas including ears and selects a heuristic position based on user-defined criteria. Besides target MNI coordinates, the function does not require manual intervention, enabling automatic end-to-end workflows.

For parameters, see the overview.

Candidate positions¶

Convert target from MNI (mm) to subject grid space (voxels)
Find candidate transducer positions on skull (expanding sphere)
Calculate criteria for each location: target distance, intersection with skin, mean distance to skull, variance in distance to skin & skull

Table: Candidate Coordinates¶

The tpars_sub-XXX_target.csv table contains transducer position candidates with the following columns. This information can be used to select a suitable candidate (e.g., by selecting a position with low intersection of the transducer and head tissue (prop_intersect < 0.05) and minimal Euclidean distance to the target; often supported by visual inspection).

Metric	Description	Interpretation
`idx`	index
`trans_x`	Transducer x-coordinate (voxels)
`trans_y`	Transducer y-coordinate (voxels)
`trans_z`	Transducer z-coordinate (voxels)
`targ_x`	Target x-coordinate (voxels)
`targ_y`	Target y-coordinate (voxels)
`targ_z`	Target z-coordinate (voxels)
`dist_to_target`	Euclidean distance transducer-to-target (voxels)
`prop_intersect`	Fraction of voxels in the transducer’s orthogonal plane that intersect with the skull mask.	Low values (≈ 0) indicate minimal skull obstruction; high values (> 0.3) indicate significant bone interference.
`mean_dist_skin`	Mean distance from non‑intersecting voxels to the skin boundary.	Indicates average proximity of the plane to the skin surface.
`var_dist_skin`	Variance of distances from non‑intersecting voxels to the skin boundary.	Higher variance → surface curvature or oblique intersection.
`mean_dist_skull`	Mean distance from all voxels in the plane to the nearest skull boundary.	Shows average clearance between the plane and skull.
`var_dist_skull`	Variance of distances to the skull boundary.	Quantifies irregularity of skull spacing within the beam footprint.

Remove ear locations¶

[Optional]

For practical reasons, transducer locations overlapping with the ears are not desired. Possible locations that overlap with ear positions can optionally be removed within an tp_ear_radius (Radius of the nogo zone, mm) around tp_left_ear_center (approximate coordinates for left ear, voxels) and tp_right_ear_center.

Select heuristic transducer position¶

The desired criterion for intersection with the skin can be defined via tp_criterion_intersection. If no placement is found within this criterion, the intersection will be iteratively expanded by tp_expand_step (default: 1%) until a match is identified.

Criterion	Default	Explanation
`tp_criterion_intersect`	< 0.05	[Fraction] Reduce skull traversal
`tp_criterion_skin_mean`	NaN	[Quantile] Short EP distance to skin
`tp_criterion_skull_mean`	NaN	[Quantile] Short EP distance to skull
`tp_criterion_skin_var`	NaN	[Quantile] Coherent transducer exit plane toward skin (e.g., little curvature)
`tp_criterion_skull_var`	NaN	[Quantile] Coherent transducer exit plane toward skull surface (Irregular bone → phase distortion → incoherent wavefront)

Amongst locations fulfilling the above criteria, the location with a minimum distance to the target is selected.

An an output, this function will create a table that contains the coordinates in voxel (grid space), mm (Localite space), and mm (RAS+ space) (based on the image header of the segmentation image).

IMPORTANT: Localite mm coordinates assume that Localite planning image has canonical_affine_transform applied (see below)! The Localite coordinates are simply rescaled by the specified voxel sizes in the header assuming no affine transformations. Always visually verify coordinates ...

Example coordinate output table:

trans_x 161     
trans_y 138     
trans_z 246     
targ_x  117     
targ_y  139     
targ_z  161     

<...>       

LOCALITE_Transducer_mm  Inf       
loc_trans_x_mm  144.9       
loc_trans_y_mm  123.09      
loc_trans_z_mm  221.02      

LOCALITE_Target_mm  Inf     
loc_targ_x_mm   105.3
loc_targ_y_mm   124.88
loc_targ_z_mm   144.65

RAS_Transducer_mm   Inf
ras_trans_x_mm  43.35
ras_trans_y_mm  -16.17
ras_trans_z_mm  82.3

RAS_Target_mm   Inf
ras_targ_x_mm   5.45
ras_targ_y_mm   -20.71
ras_targ_z_mm   5.19

Plot heuristic transducer position¶

PRESTUS generates an overview of the selected transducer placement:

ex_heuristic_placement

If parameters.localite_path is specified, it will also deposit a copy of the plot there.

A T1 overlay plot (plot_placement_t1_overlay) is generated automatically after the heuristic position is selected. It shows the transducer position and target overlaid on orthogonal T1 slices and is written to the subject output folder alongside the standard placement figure.

Save Localite planning image¶

[Optional]

Localite can struggle with images that contain affine matrices in the image header. PRESTUS offers the optional step (tp_save_localiteT1) of depositing a header-adjusted T1 planning image for Localite (based on the T1.nii.gz in the SimNIBS output m2m directory). For this, parameters.localite_path has to be specified.