Fix make_gaussian_kernel truncated parameter for kernel_size > 3

[atharvajoshi01]

Fixes #8780.

make_gaussian_kernel passed kernel_size // 2 as the truncated parameter to gaussian_1d. But gaussian_1d interprets truncated as the number of standard deviations, not the half-width in samples. It computes tail = int(sigma * truncated).

With kernel_size=15, sigma=5.0, truncated=7:

• tail = 5.0 * 7 = 35 → kernel has 71 elements
• Slicing to [:15] keeps only the far-left tail where values are near zero
• LNCC of identical images returns ~0.0 instead of -1.0

Fix: set truncated = (kernel_size // 2) / sigma so that tail equals kernel_size // 2, producing a correctly-sized kernel.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: b622c50a-8aec-49f4-a320-6658dd7c1d7f

📥 Commits

Reviewing files that changed from the base of the PR and between 8d39519 and 0a097c0.

📒 Files selected for processing (1)

• monai/losses/image_dissimilarity.py

---

📝 Walkthrough

Walkthrough

Modified make_gaussian_kernel in the image dissimilarity module to compute the truncated parameter correctly. The function now normalizes the pixel radius by dividing by sigma, converting from absolute pixel units to standard deviation units before passing to gaussian_1d. The scaling and slicing logic remains unchanged.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title clearly and concisely summarizes the main change: fixing the truncated parameter calculation in make_gaussian_kernel for larger kernel sizes.
Description check	✅ Passed	Description includes issue reference, clear explanation of the bug, and the fix with concrete examples; however, it does not address the PR template sections (Types of changes, testing).
Linked Issues check	✅ Passed	Changes directly address issue #8780: correcting truncated calculation from kernel_size // 2 to (kernel_size // 2) / sigma produces properly sized Gaussian kernels.
Out of Scope Changes check	✅ Passed	Changes are tightly scoped: only the truncated parameter computation in make_gaussian_kernel is modified; no extraneous alterations present.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[ericspod]

Hi @atharvajoshi01 thanks for this fix. We had PR #8781 already open for it but it initially didn't mention the original issue so you wouldn't have seen it. I think it's equivalent to your solution, if you wouldn't mind please check that you agree with that. I think we would merge the other PR as it also has tests included. Sorry you spent time on this, we should have gotten to the other PR sooner.

[atharvajoshi01]

Checked #8781 — yes, both fixes address the same root cause. Their approach is actually cleaner since it inlines the Gaussian computation directly instead of going through gaussian_1d, which avoids the truncated parameter issue entirely. Happy to close this one in favor of #8781.