| # Investigate a volume-to-surface projection anomaly |
|
|
| The supplied source is a real snapshot of a published neuroimaging analysis |
| library. In one volume-to-surface workflow, a volumetric product distributed |
| with an anatomical lookup table is mapped onto a cortical surface. The command |
| completes, but some finite surface values cannot be associated with any row of |
| that table. |
|
|
| Treat this as a scientific reproduction investigation rather than a textual |
| patch task. Inspect the source and the public reproduction, |
| reproduce the anomaly, and determine why the projected result is incompatible |
| with the accompanying scientific metadata. Repair the public |
| volume-to-surface workflow so that the result can be used in the downstream |
| analysis while other documented uses remain compatible. |
|
|
| The repair should generalize beyond the supplied example and should not be |
| specialized to one volume, surface, or lookup table. Preserve the library's |
| documented behavior for other supported inputs and geometries. |
|
|
| Do not hard-code the public volume, table entries, mesh, sample count, affine, or |
| expected output. Do not modify `reproduce.py`, the fixtures, or the evaluation |
| harness. Work inside the supplied source snapshot and keep the task runnable |
| without network access. |
|
|
| Run the public workflow with: |
|
|
| ```bash |
| python reproduce.py |
| ``` |
|
|
| The initial snapshot is expected to report the observed scientific anomaly. |
| After your repair it should report `post_fix_success`. Generated files belong |
| under `outputs/`. |
|
|
