AmenStyleSPECT — SPECT Brain Render (Amen-style)

A 3D Slicer extension that turns a reconstructed perfusion SPECT volume into two Amen-style 3D renderings:

• Surface scan — a shaded cortical surface at the perfusion threshold; hypoperfusion shows as dents / holes (a hole is below threshold, not zero perfusion). Defaults to a marching-cubes isosurface with the Amen positional spectrum (pink superior → violet inferior); a GPU volume render is also available.
• Active scan — a blue wireframe brain plus the most-active tissue (hyperperfusion) rendered opaque red (≥ 85 %) → white (≥ 92 % of cerebellar max).
• Regional quantification — a per-region table reading each of the 58 Amen regions (29 paired L/R, incl. the precuneus as a default-mode-network hub and the nucleus accumbens) as a % of the cerebellar reference, classified hypoactive / normal / hyperactive, with an ordinal ±4 grade, left/right asymmetry flags and a TSV export. The % Ref cell is tinted with the Amen 20-step "hot" color scale so the table reads like the active render.
• Automatic atlas quantification (v0.7) — register the SPECT to a bundled MNI HMPAO template, warp the AAL3 atlas into patient space, and read all 58 regions from their exact anatomical masks in one click — with a registration-quality verdict (PASS / REVIEW / FAIL) and a manual landmark rescue for atypical cases. The warped atlas is named (hover any region in the slice views to see it) and toggles on/off over the SPECT. No MRI required.

All thresholds are a percent of the Maximum Cerebellar Count (relative, within-patient scaling), following the Amen reading convention. Bands: hot/hyperactive ≥ 85 %, normal 55–85 %, hypoactive < 55 %; surface dents at 55 %.

⚠️ Research / visualization aid only. NOT FDA/CE-cleared and NOT for primary diagnosis. It renders relative within-patient perfusion with fixed percent thresholds; it does not perform normative-database (z-score) statistics. Validate thresholds against reference scans before any interpretive use. SPECT has no intrinsic left/right landmark — confirm orientation on every scan before reading laterality.

---

Requirements

• 3D Slicer 5.x (developed and tested on 5.10). Download: https://download.slicer.org.
• Nothing else. This is a pure scripted (Python) module — no compilation, no CMake build, no extra Python packages. It runs the same on Windows, macOS, and Linux.

---

Install on another PC

You only need the AmenStyleSPECT/SPECTBrainRender/ folder (the rest is for source control / future Extensions-Index packaging). Copy the whole AmenStyleSPECT folder to the target PC — e.g. unzip AmenStyleSPECT-<version>.zip somewhere stable like Documents\Slicer\AmenStyleSPECT (Windows) or ~/Slicer/AmenStyleSPECT (macOS/Linux).

Method A — GUI (recommended, always works)

1. Open Slicer → Edit ▸ Application Settings ▸ Modules.
2. Under Additional module paths, click Add and select the folder that contains SPECTBrainRender.py:
   • Windows: ...\AmenStyleSPECT\SPECTBrainRender
   • macOS/Linux: .../AmenStyleSPECT/SPECTBrainRender (Select the SPECTBrainRender folder itself — not the outer AmenStyleSPECT folder.)
3. (Optional) tick Enable developer mode to get a Reload button while testing.
4. Click OK and restart Slicer when prompted.
5. The module appears under Nuclear Medicine ▸ SPECT Brain Render (Amen-style) (use the module search / magnifier if you don't see it).

Method B — one-paste install helper (convenience)

1. Copy the extension folder to the target PC.
2. Open Slicer → View ▸ Python Console.
3. Open install_helper.py (next to this README), set the MODULE_DIR line at the top to the SPECTBrainRender path on that PC, then paste the whole file into the console and press Enter. It registers the path, persists it, and offers to restart.
4. If for any reason the module doesn't appear after restart, use Method A — it is the authoritative path.

Updating to a newer version

Replace the folder contents with the new files (keep the same location so the module path you registered still points at it). Then either restart Slicer, or — in developer mode — open the module and click Reload.

---

Usage (quick walkthrough)

1. Input — select your already-loaded, reconstructed SPECT scalar volume.

2. Orientation — click Show mid-axial slice, set the L/R · A/P · S/I flips so the patient's left is where you expect, then tick Orientation confirmed.

3. Reference (the 100 % anchor) — keep mode on Cerebellar max and either:

   • type the Maximum Cerebellar Count into Manual reference (counts), or
   • Create cerebellum ROI → drag it over the cerebellum → Find cerebellar max → step through candidates with Jump to # → Accept # as 100 %. The bold label shows the resolved reference. Toggle the candidate crosshair / ROI box off to declutter.

4. Preprocess / brain mask — largest-island masking (drops scalp / sinus / salivary uptake) is on by default.

5. Surface scan — pick the technique (isosurface mesh, the default, or volume render), set the threshold (default 55 %), opacity (default 1.0), and smoothness (default 0.3), pick a color scheme (Amen spectrum by default; or solid sculpt, hole-emphasis, perfusion rainbow), then Generate Surface. Use Drop to 45 % for the prognostic view.

6. Active scan — set the active threshold / red point (default 85 %), the white point (default 92 % of cerebellar max, the 100 % ceiling), and wireframe options, then Generate Active. Hyperperfused tissue renders red at the threshold and ramps to white at the white point.

7. Views / export — snap the camera to any of the 6 standard orientations, or pick an Output folder and click Export Surface 6-view PNG / Export Active 6-view PNG to write 6 PNGs + a montage.png.

8. Regional quantification — in Regions to include, tick the lobes (or expand a lobe and tick individual regions) you actually want to read — everything is on by default, so Select none then pick a few, or just untick what you don't need. Click Create selected region ROIs to drop one named Box ROI per ticked region (approximate starting positions) under an Amen Regions folder in the Data module, then drag each box onto its true structure (SPECT has limited anatomical detail, so this step is manual and clinician-judged). Remove unchecked region ROIs (off by default) additionally deletes any region boxes you've unticked, so the scene holds exactly your selection — it only ever touches boxes from this taxonomy, never your cerebellum reference or custom ROIs. Choose the region metric (mean / peak) and reference (% of cerebellar max, or % of the cerebellum-ROI mean), set the hypo / hyper / floor / asymmetry thresholds, and click Quantify regions (it reads whichever region ROIs are in the scene). The table sorts most-hypoactive first, tints each region's % Ref cell with the Amen 20-step hot scale (blue → green → yellow → red → white, so it matches the active render) and shades the Class / Grade cells by severity, shows each region's Hot % (the fraction of the ROI above the hyper threshold), and flags left/right asymmetries; double-click a row to jump the slices to that region. Export regional table (TSV) writes regional_all.tsv and regional_hypo_hyper.tsv to the Views/export output folder.

   Quantification reads the same brain-masked volume the scans render (largest-island mask, when the Preprocess / brain mask option is on), so extracranial uptake — scalp, sinus, salivary, venous — that is stripped out of the render and never shows as red is not counted as region signal. (Turn the brain mask off to quantify the raw volume.) Hypoactive is judged from the region mean (low overall perfusion). Hyperactive is judged from the hot voxels: with Focal hyperactivity on (default), a region is flagged hyperactive when its hot focus reaches Focal: min hot voxels (default 3) or Focal: or min hot fraction % of the ROI (whichever is met first) — so a true focal hot spot is not diluted to "normal" by the surrounding normal tissue in a loose ROI. This reproduces the older "mask at 85 %, average inside the mask" read. The Hot % column shows each region's hot fraction and its hot-voxel count in parentheses — tune the thresholds against those numbers (e.g. if a focus reads 4.8% (24) but isn't flagged, the 24 hot voxels already clear the default count of 3; if noise over-flags, raise the count). Turn Focal hyperactivity off to classify hyperactivity purely from the region mean/peak.

   Each abnormal region is then given an ordinal grade — the Amen 20-step color scale (each step = 5 % of the cerebellar reference) read as signed deviation from the normal window: hyperperfusion +1 … +4 and hypoperfusion −1 … −4, shown in a Grade column and shaded deep-blue (−4) → green (0) → deep-red (+4).

   • Hyper (+1…+4) by the hot-voxel mean, stepping up from the 85 % hot line: +1 85–90 %, +2 90–95 %, +3 95–100 %, +4 ≥ 100 % (at/above the cerebellar ceiling).
   • Hypo (−1…−4) by the region mean, stepping down from the 55 % hypo line: −1 50–55 %, −2 45–50 %, −3 40–45 %, −4 < 40 %.

   The single Grade step (% per color) spin box (default 5) sets the width of every step (so it stays a true 20-step scale). The active-scan render's white point (92 %) is set separately in the Active section and is independent of the grade ladder. The negative grades are the surface-scan (hypoperfusion) read, the positive grades the active-scan (hyperperfusion) read. The TSV carries a signed Grade column.

9. Automatic atlas quantification (optional — replaces the manual box-dragging step) — in Regional quantification, click Register atlas to this SPECT (auto): it registers the scan to the bundled MNI HMPAO template, warps the AAL3 atlas into the patient's space (~30–90 s), auto-fills the cerebellar reference, and reports a PASS / REVIEW / FAIL registration-quality verdict (containment, cerebellum/brain ratio, template-match NCC). The warped atlas is laid over the SPECT as a named colored overlay — hover any region in the slice views to read it in the Data Probe — and the Show SPECT / Show atlas overlay checkboxes + Atlas opacity slider let you blend or isolate either layer. Then click Quantify from atlas (exact masks) to read every region from its precise atlas voxels — the same table / report / TSV as the manual path, with no box-dragging. Verify the auto-set reference, and confirm orientation first.

   • On REVIEW / FAIL: rescue it — Place patient landmarks (6 named points appear at approximate spots, each carrying a placement hint), drag each onto its true location. Five are silhouette edges (frontal/occipital poles, vertex, L/R temporal); the cerebellum point is the centre of the bright cerebellar blob and has the most leverage. Then Re-align atlas from landmarks. If it still fails, fall back to the manual Create selected region ROIs workflow — the verdict is telling you the registration can't be trusted for this scan.
   • Seed boxes from atlas positions the manual region boxes at their atlas centroids (better starting points than the fixed scaffold) if you prefer the adjustable-box workflow. The atlas and template are bundled with the extension — nothing is downloaded.

The Status / calibration log prints the resolved reference and the absolute count value behind each percent threshold on every render, and the full regional table.

Nodes this creates in the scene

BrainMasked (Surface) and BrainMasked (Active) (the two scans render from separate masked clones, each with its own volume-rendering display/property node, so they never overwrite each other), BlueBrain + BlueNodes (active wireframe), the AmenSurface model (the marching-cubes isosurface surface scan), the named region Box ROIs under the Amen Regions folder, and — for the atlas pipeline — the warped AAL3 (in patient) labelmap (with an AAL3 region names color table so it hover-labels each region) and the Patient landmarks fiducials. Your original volume is never modified.

---

License & attribution

This extension is free software under the GNU General Public License v3 (see the bundled LICENSE). It redistributes two third-party assets under their own terms:

• AAL3 atlas (Resources/Atlas/AAL3v1.nii + AAL3v1.nii.txt) — Automated Anatomical Labelling atlas 3, GIN-IMN (Bordeaux), released as copyright freeware under the GNU GPL. Cite: Rolls, Huang, Lin, Feng & Joliot, "Automated anatomical labelling atlas 3," NeuroImage 2020;206:116189; original AAL: Tzourio-Mazoyer et al., NeuroImage 2002;15:273–289.
• MNI HMPAO perfusion template (Resources/Atlas/SPECT_HMPAO_template.nii.gz) — the Tc-99m HMPAO SPECT template distributed with SPM (Statistical Parametric Mapping; GPL), converted to NIfTI, used as the spatial-normalization target.

The AAL3 → Amen-taxonomy region grouping is a draft clinical mapping (e.g. Broca's = pars opercularis only; the "Deep Limbic / Thalamic" region pools several small midbrain nuclei). Review it against your own conventions before any interpretive use.

---

Packaging a distributable zip (maintainer)

From the project root (slicerAmen/):

./AmenStyleSPECT/package.sh           # writes AmenStyleSPECT-<version>.zip

The script zips the AmenStyleSPECT folder while excluding .DS_Store, __pycache__, and any local junk. Copy the resulting zip to other PCs and install per the steps above.

Going further (optional)

The repo keeps a root CMakeLists.txt and per-module CMakeLists.txt so the extension can later be built/submitted to the Slicer Extensions Index for one-click install via the Extensions Manager. That requires building against a matching Slicer and is not needed for the manual install above.

Files

AmenStyleSPECT/
  CMakeLists.txt                         extension metadata (for Extensions Index only)
  LICENSE                                GNU GPL v3
  README.md                              this file
  package.sh                             build a distributable zip
  install_helper.py                      one-paste path registration for a new PC
  SPECTBrainRender/
    CMakeLists.txt                       scripted-module build rules
    SPECTBrainRender.py                  module (Widget + Logic + self-tests)
    Resources/
      Icons/SPECTBrainRender.png
      UI/SPECTBrainRender.ui
      Presets/AmenColor.vp.json          reference copy of the clinic color preset
      AAL3v1.nii.txt                     AAL3 region lookup (label -> name)
      Atlas/AAL3v1.nii                   AAL3 atlas (MNI, GPL)
      Atlas/SPECT_HMPAO_template.nii.gz  MNI HMPAO perfusion template (from SPM)

Version: 0.8.0 (open-source, GPL v3). Aligns the grading and taxonomy to the published Amen reading convention: the ordinal grade is now the 20-step / 5 %-per-color scale (bands hypo < 55 %, normal 55–85 %, hot ≥ 85 %), the regional table's % Ref cell is tinted with the Amen hot color LUT, the nucleus accumbens is restored as its own region (58 regions / 29 paired), the warped atlas now hover-labels each region and toggles on/off over the SPECT with an opacity slider, and the landmark-rescue points carry placement hints. Builds on 0.7.0 (automatic AAL3 atlas pipeline: in-module registration to a bundled MNI HMPAO template, exact-mask quantification, PASS/REVIEW/FAIL QC, landmark rescue) and 0.6.0 (marching-cubes surface scan + Amen positional spectrum, active scan, cerebellar reference, 6-view montage export). DICOM NM import and CALIBRATION.md are still to come.