Sagittal H-F R=2 anomaly diagnosis

Generated: local Esche diagnostic run

Question: Why does BI_7 look best for sagittal_hf_r2, while Mammaria_18Ch is expected to be stronger from channel count and geometry?

Status: Not final. This is a diagnostic page. Do not use the current heat table as final product story until ROI/model/formula causes are resolved.

Focus table: sagittal_hf_r2

Setup	g_p95	cond_p95	smin_p05	ROI pixels	ROI Y range	ROI Z range	corr(g,cond)	corr(g,1/smin)
BI_7	2.565	4.998	0.001655	13300	-125.8..1.595	-97.47..97.47	0.9935	0.9515
Mammaria_8Ch	6.622	14.05	0.0006813	9450	34.92..134	-90.4..91.41	0.9966	0.9853
Mammaria_12Ch	6.033	12.51	0.0009403	9576	34.65..133.8	-90.4..91.41	0.996	0.9835
Mammaria_18Ch	5.795	11.77	0.0012	8118	35.7..123	-88.95..88.51	0.988	0.9762

Diagnostic plots

Four-panel diagnostic

Source heat table

Source hotspot summary

Source sagittal H-F R=2 map

Source sagittal H-F R=3 map

Interpretation checklist

If BI_7 also has much better cond_p95 and smin_p05, the simplified model is genuinely favoring BI_7 in this ROI/slice and we must inspect the sensitivity model and slice/ROI definition.
If BI_7 has lower g_p95 but not better condition/smin, the g-factor formula or normalization is suspect.
If ROI pixels/ranges are very different, the heat table is not comparing homologous breast regions.
If Mammaria_18Ch improves at R=3 but not R=2, the issue may be thresholding/ROI p95 behavior rather than pure channel-count benefit.