NerveAlign
Private

NERVE-FACE ALIGNMENT

Bring the faces into register.

SYNTHETIC TESTSame fascicles, two scales. Face B is rotated 37° clockwise.

Prepare faces

Face A
Synthetic face A

Calibrated test image. Choose a tool to adjust the mask.

01

Calibrate & center

40.0 px/mm
02

Fascicles

Threshold segmentation within the nerve circle.

12 px

Alignment

B rotates clockwise over A
Fascicle overlap Dice—
Current rotation0.0°
Exact optimum—
CENTERED · PHYSICAL SCALE
Face AFace BOverlap

Choose the highest mean Dice across the angular window. Default ±5°: 11 mask measurements, 1° apart, including the center.

Off by default

Ready to compare the test pair.

Exact optimal alignment

Angle · clockwise
—
Maximum Dice overlap
—

Surgically robust alignment

Angle · clockwise
—
Center Dice overlap
—
Mean Dice across ±5°
—
Minimum Dice across ±5°
—

Enable surgical tolerance optimization, then find both alignments.

Overlap details & rotation curve
IoU —A covered —B covered —
Mask DiceExact optimumRobust optimum±5° window

Dice = 2 × intersection / total fascicle area. IoU = intersection / union. Exact search: 1° steps, then 0.1° refinement. Robust search: every 1° center, scoring actual masks at every 1° offset within the window. Highest mean wins; ties favor the higher minimum, then center Dice, then smaller angle. Scores describe geometry, not axonal identity.

GIF and overlay exports use the current rotation. Apply either alignment before exporting.

A few precise steps.

  1. Load each face. Photograph face-on with a 1 cm bar in the same plane as the nerve face. Use JPEG, PNG, WebP, or a format your browser can decode.
  2. Calibrate independently. Select “Set 1 cm bar” and tap both ends of the reference on each image. Then select “Set nerve circle”: tap the center followed by the outer edge.
  3. Build the fascicle mask. Choose lighter or darker fascicles and a threshold, then segment. Paint, erase, or trace closed outlines to correct the result. Ultrasound mode uses median speckle filtering and a darker-fascicle default; every mask needs visual review.
  4. Compare. Centered faces keep their measured physical sizes. B rotates clockwise in the displayed image plane. Mirroring is optional; use consistent face orientation and confirm anatomical landmarks.
  5. Allow for positioning tolerance. Enable “Surgical tolerance optimization” and set a whole-number tolerance from 0° to 180° (default ±5°). Find both alignments to compare the exact maximum with the highest mean overlap across the window. The robust angle is applied automatically; use either result’s Apply button to switch. The graph marks both optima and shades the robust window, including across 0°.
  6. Export. Apply either optimum, then export a GIF that rotates B from 0° to the current angle and holds there. The overlay and GIF include the scale, angle, and overlap score.

The test pair uses a known 37° clockwise offset and different pixels/mm; the expected correction is about 323° clockwise (37° counterclockwise). The result is a geometric match, not a prediction of motor–sensory correspondence or PEG-fusion outcome.

On iPhone: open in Safari, sign in with your authorized ChatGPT account, then Share → Add to Home Screen. Private sign-in remains required; images are never uploaded or saved by the app.