Generated: external_pi_reference_audit_20260505_205248
Purpose: independent reference audit using synthetic complex coil sensitivity maps. This does not replace the Coil-Compare pipeline and is not clinical validation.
Main question: Does a Mammaria_18-like synthetic complex sensitivity set outperform a BI_7-like set for H-F/y acceleration when the sensitivity variation is modeled explicitly?
| Array | Condition | Channels | Local g p95 | PyGRAPPA g p95 | RMS diff | PyGRAPPA error |
|---|---|---|---|---|---|---|
| BI_7_like_broad_7ch | HF_y_R2 | 7 | 1.248 | 1.352 | 0.1329 | |
| BI_7_like_broad_7ch | HF_y_R3 | 7 | 2.885 | 4.025 | 1.451 | |
| BI_7_like_broad_7ch | RL_x_R2 | 7 | 1.582 | 1.218 | 0.2612 | |
| BI_7_like_broad_7ch | RL_x_R3 | 7 | 1e+06 | 2.3 | 1e+06 | |
| Mammaria_18_like_vertical_18ch | HF_y_R2 | 18 | 1.003 | 1.046 | 0.01982 | |
| Mammaria_18_like_vertical_18ch | HF_y_R3 | 18 | 1.118 | 1.548 | 0.3576 | |
| Mammaria_18_like_vertical_18ch | RL_x_R2 | 18 | 1.304 | 1.001 | 0.1297 | |
| Mammaria_18_like_vertical_18ch | RL_x_R3 | 18 | 1e+06 | 1.041 | 1e+06 | |
| Mammaria_18_like_magnitude_only | HF_y_R2 | 18 | 1.004 | 1.313 | 0.1334 | |
| Mammaria_18_like_magnitude_only | HF_y_R3 | 18 | 1.147 | 1.004 | 0.4403 | |
| Mammaria_18_like_magnitude_only | RL_x_R2 | 18 | 1.313 | 1.004 | 0.1334 | |
| Mammaria_18_like_magnitude_only | RL_x_R3 | 18 | 1e+06 | 1.147 | 1e+06 |
| Case | Expected | Observed | Pass |
|---|---|---|---|
| orthogonal_two_channel_R2 | g near 1 | 1.0 | True |
| identical_two_channel_R2 | rank deficient high capped g | 1000000.0 | True |
| global_scale_invariance | global sensitivity scale should not change g | 7.310578359934718e-17 | True |
{
"assertions": [
{
"case": "orthogonal_two_channel_R2",
"expected": "g near 1",
"local_g_p95": 1.0,
"pass": true
},
{
"case": "identical_two_channel_R2",
"expected": "rank deficient high capped g",
"local_g_p95": 1000000.0,
"pass": true
},
{
"case": "global_scale_invariance",
"expected": "global sensitivity scale should not change g",
"pass": true,
"rmsdiff": 7.310578359934718e-17
}
],
"bart_status": {
"available": true,
"returncode": 0,
"stderr": "",
"stdout": "v0.9.00"
},
"interpretation": [
"If Mammaria_18_like beats BI_7_like under synthetic complex H-F maps, external tools agree with the intuition that richer H-F sensitivity can improve g-factor.",
"If the live STEP model still favors BI_7, the issue is the STEP-derived sensitivity surrogate, not the SENSE/g-factor arithmetic alone.",
"PyGRAPPA and local formula need not match pixel-perfectly, but trends should be interpretable."
],
"purpose": "External PI reference audit using synthetic complex sensitivity maps.",
"sigpy_status": {
"available": true,
"note": "Imported SigPy SenseRecon/EspiritCalib. Full reference recon requires synthetic or measured k-space plus sensitivity maps."
}
}