add_body_envelope: add step-8 table detection to remove the CT table

[Can-Zhao]

The find-air-invert steps can leak the air-density CT table into the body (air trapped between patient and table is a separate component from the exterior air, so it reads as "not air" -> body). Step 8 detects it as the largest connected component of body voxels that are actually air (CT < hu_threshold) and drops it when that component is >= table_frac_thresh (default 0.05) of the body. Since the seg labels the lungs, no legitimate air region is table-sized (a table is ~16-28% of body vs <0.3% clean).

[greptile-apps]

Greptile Summary

This PR adds a step-8 "table detection" safety net to add_body_envelope in scripts/utils.py. When the find-air-invert pipeline leaks the CT table into the body mask (air trapped between patient and table is isolated from exterior air, so it reads as "not air" → body), step 8 identifies it as the largest connected component of body voxels with air-density HU and removes it when that component meets or exceeds table_frac_thresh (default 5%) of the total body.

• Adds a single new parameter table_frac_thresh: float = 0.05 at the function signature, inserted before device — all visible callers use only the first two positional arguments, so there is no backward-compatibility break.
• The detection logic correctly scopes to (out == body_label) & air_hu, so organ-labeled voxels (including lungs) are never candidates for removal; the fraction math (n_table / n_body) is computed before removal, which is the right baseline; and the print statement (previously flagged as missing observability) is now present and will surface in logs when the safety net fires.

Confidence Score: 5/5

Safe to merge — the change is a targeted safety net that only fires when a large air-density connected component is found inside the body mask, and it correctly scopes removal to unlabeled body voxels so organ labels are never touched.

The detection logic is well-bounded: it operates on (out == body_label) & air_hu, so labeled organ voxels (including lungs) can never be candidates for removal. The fraction comparison uses the pre-removal body count as the denominator, which is the correct baseline. All visible callers use only the first two positional arguments, so the new parameter causes no call-site breakage. The two concerns raised in the previous review round are both addressed.

No files require special attention.

Important Files Changed

Filename	Overview
scripts/utils.py	Adds step-8 table-detection safety net to add_body_envelope; logic, scoping, and fraction math are correct; previous review concerns (np.asarray wrap, observability print) are addressed.

_{Reviews (2): Last reviewed commit: "ruff format" | Re-trigger Greptile}