CT-driven detector motion¶

Modern CZT cameras like GE StarGuide do not orbit on a fixed circle. Each detector column moves in radially to hug the patient's body contour, then the gantry rotates and each column sweeps, keeping the detectors as close to the patient as possible for the best resolution and sensitivity. The make_ring_motion.py tool reproduces that behavior: it reads a patient's body contour and writes a OpenTOPAS motion deck in which each detector docks against the contour and moves without ever colliding with the body.

What it produces¶

Given a body contour, the tool computes, for each detector, a docking radius and a motion sequence, and emits an OpenTOPAS deck driven by time features:

HOME  →  APPROACH (each detector moves in to hug the contour)
      →  ACQUIRE  (the gantry rotates and detectors sweep)
      →  RETRACT  (detectors return home)

It supports two detector layouts:

starguide, the 12-column CZT ring, with the true StarGuide column geometry.
ring, a generic ring of N detectors you specify.

Getting the contour¶

You can supply the body contour three ways:

Analytic ellipsePatient CT/PET (DICOM)Measured radii

The simplest option: approximate the cross-section as an ellipse with semi-axes A and B (mm):

python3 tools/make_ring_motion.py --system starguide --ellipse 150 100 --out motion.txt

Extract the real body contour from a DICOM series. The tool builds a clean body mask ( thresholding, filling internal air, removing the scanner table and detached arms, and keeping the largest connected component), then casts rays from the body centroid to measure the radial profile:

python3 tools/make_ring_motion.py --system ring --ndet 16 \
        --dicom /path/to/CT_series --hu-threshold -300 --out motion.txt

If you have already extracted the contour as a list of radii (for example, from a de-identified, on-server extraction that returns only geometry), pass them directly:

python3 tools/make_ring_motion.py --system starguide \
        --radii 91 94 140 157 138 121 101 114 120 153 153 115 --out motion.txt

Why the docking is collision-safe¶

A detector that docks against the contour at its starting azimuth would collide with the patient as the gantry rotates it into a wider part of the body. The tool avoids this: each detector docks at the maximum contour distance over the entire angular range it will sweep (its azimuth through the gantry rotation, plus the panel's tangential extent), so it clears the body throughout the motion. This is checked automatically: the generated geometry builds with zero overlaps, and the regression tests assert that no detector penetrates the body at any gantry angle.

Visualizing and running¶

The generated deck is an ordinary OpenTOPAS parameter file. Open it in the Qt viewer to watch the motion, or include it in an acquisition. To run it headless, strip the viewer lines and set Ts/PauseBeforeQuit = "False".

Verifying it yourself¶

The contour extraction and docking geometry are covered by ground-truth regression tests (tests/test_ring_motion.py) that need no patient data: they generate synthetic phantoms of known size and confirm the tool recovers them:

python3 tests/test_ring_motion.py

The tests check that a known circle and ellipse are recovered to within one pixel, that a scanner table and a detached arm are correctly removed, that each detector docks at contour-plus-clearance and faces the patient, and that no detector collides during rotation.