All-4 Breast-ROI Coil Comparison

Executive Summary and Scope Limits

The primary PPA ROI is the geometry-derived breast imaging volume between medial and lateral H-F/near-vertical conductor elements. It excludes outer unused FOV regions from scaling and ranking where the conductor-derived mask is available.

This remains a simplified relative design comparison. It does not claim electromagnetic simulation, clinical validation, coupling behavior, matching networks, electronics performance, real noise covariance, or measured SNR.

Validation Summary

This public page shows a readable validation summary. The machine-readable contract is copied to tables/final_report_contract.json.

Static scope	PASS	4 setups, accepted sagittal H-F and transversal R-L PPA only, R=2/R=3
SNR source	PASS	current_step_overlays
Breast-ROI g-factor/PPA source	PASS	physical-alias v4 breast-ROI maps and summary table
Breast-ROI ranking source	PASS	physical-alias v4 NPZ-mask-derived ROI ranking
Accepted image path policy	PASS	no rejected display image paths selected

Metric Definitions

SNR-like map: Relative receive-sensitivity display from the current STEP trace overlay source. It is a comparative visualization of receive sensitivity behavior and is not measured SNR.
g-Factor: The g-factor is the actual parallel-imaging geometry and noise-amplification penalty. Lower is better. A value near 1 is ideal. High g-factor means that folded positions along the alias direction are locally hard to distinguish from the coil sensitivity profiles. In simplified SENSE-like terms: SNR_PI ~= SNR_full / (sqrt(R) * g). sqrt(R) is the unavoidable acceleration penalty; g is the additional geometry/noise penalty. g-factor is not measured SNR and not absolute clinical image quality. It can only be compared within the same valid acceleration direction, ROI definition, and modeling assumptions.
Condition number: The condition number is a numerical stability diagnostic of the local encoding / unfolding matrix. Lower is better. High condition number means that the local matrix inversion is unstable: small errors, noise, or sensitivity-model changes can cause large changes in the solution. In the 2-norm: cond2(A) = sigma_max / sigma_min. Condition number is not SNR and not the final g-factor. It helps explain why g-factor problems may occur.
Smallest singular value: The smallest singular value is the weakest independent encoding mode of the local sensitivity matrix. Larger is better. Small smin means low encoding margin: there is a combination of folded positions that the channel set can barely distinguish. Small smin often explains high condition number and high g-factor regions. smin is not SNR and not a final quality metric; it is a linear-algebra diagnostic.
Breast ROI: Geometry-derived primary ROI between medial and lateral H-F/near-vertical conductor elements. It is intended to suppress unused outer FOV influence and is not an anatomy mask.
Hotspot: Fraction of breast ROI pixels above an explicit threshold. The v4 threshold is max(2.0, min(3.0, breast-ROI g p95)) for R=2 and max(5.0, min(10.0, breast-ROI g p95)) for R=3.
ROI ranking: Method-review aggregate setup comparison based primarily on mean breast-ROI g p95 across the accepted conditions, with worst-condition and hotspot views retained to expose condition-specific behavior. It is not a clinical ranking.
Slice position QA: Red straight guide lines mark the evaluated transversal Z, coronal Y, and sagittal X slices. These are separate from the red ROI contours used in the condition maps.
Model limitations: This report is a simplified relative design indicator. It does not claim electromagnetic simulation, clinical validation, coupling behavior, matching networks, electronics performance, real noise covariance, or measured SNR.

SNR Maps

SNR-like source policy for this revision requires the current STEP trace overlay source. The old V8-widebox family is review-only fallback and fails the final contract if selected.

SNR-like overview interpretation. Use these three current STEP-overlay planes to compare relative spatial coverage and conductor placement across the four setups. The displays support qualitative comparison only and should not be read as measured SNR.

**SNR-like transversal**Current STEP trace overlay source with all sampled conductor traces drawn.

**SNR-like coronal**Current STEP trace overlay source for superior-inferior and lateral coverage comparison.

**SNR-like sagittal**Current STEP trace overlay source for left-right slice comparison.

Slice Position QA

Red straight lines mark the evaluated slice positions: fixed Z for the transversal slice, fixed Y for the coronal slice, and fixed X for the sagittal slice. These straight guide lines are separate from the red ROI contour in the condition maps.

Slice-position QA interpretation. The coronal Y line is setup-specific. The table reports the final value used, the automatic detector value, and any explicit review override. If an override is used, it is labelled as a review override rather than pure automatic detection.

**Three-plane slice-position QA**X/Y, X/Z, and Z/Y projections with conductor overlays and per-setup sagittal X positions.

Slice Position Table

Setup	final coronal_y0_mm	coronal policy	automatic coronal_y0_mm	sagittal_x0_mm	derivation/source
BI_7	-52.35067933681667	automatic_lateral_main_vertical_midheight	-52.35067933681667	73.7161372342265	automatic_lateral_main_vertical_midheight
Mammaria_8Ch	87.01167404000553	automatic_lateral_main_vertical_midheight	87.01167404000553	103.45608161859383	automatic_lateral_main_vertical_midheight
Mammaria_12Ch	87.01167404000553	automatic_lateral_main_vertical_midheight	87.01167404000553	103.45608161859383	automatic_lateral_main_vertical_midheight
Mammaria_18Ch	105.91628706755631	automatic_lateral_main_vertical_midheight	105.91628706755631	99.65300425675834	automatic_lateral_main_vertical_midheight

Breast-ROI g-factor and PPA Diagnostics

Accepted PPA diagnostics are limited to sagittal H-F and axial/transversal R-L at R=2 and R=3. Red contours show the exact geometry-derived ROI used for scaling and ranking.

Accepted PPA scope. The accepted breast PPA scope is sagittal H-F and axial/transversal R-L only, each at R=2 and R=3. Sagittal R-L, transversal H-F, A-P phase direction, and coronal PPA are excluded from accepted interpretation.

Transversal PPA interpretation. In transversal maps, the valid accepted phase/alias direction is R-L. Review the breast-ROI-scaled g-factor, condition number, and smallest singular value panels for local instability inside the red ROI boundary.

Sagittal PPA interpretation. In sagittal maps, the per-setup X slice is retained from the v3 medial/lateral slice logic and the valid accepted phase/alias direction is H-F. Review condition-specific behavior rather than total channel count alone.

**transversal RL R=2**Physical-alias v4 diagnostic panel with breast-ROI robust scaling and red ROI boundary.

**transversal RL R=3**Physical-alias v4 diagnostic panel with breast-ROI robust scaling and red ROI boundary.

**sagittal HF R=2**Physical-alias v4 diagnostic panel with breast-ROI robust scaling and red ROI boundary.

**sagittal HF R=3**Physical-alias v4 diagnostic panel with breast-ROI robust scaling and red ROI boundary.

Breast ROI Ranking, Heat Tables, and Hotspots

Ranking is method-review only. It is based on accepted PPA conditions and on the v4 NPZ mask when present. Fallback is allowed only for missing masks and is labeled explicitly.

**Breast ROI ranking by mean g p95**Lower aggregate g p95 is better under the geometry-derived breast ROI rule.

**PPA/g-factor performance bars**Mean g p95, worst g p95, clip-risk count, and R=3-only mean g p95 by setup.

**Breast-ROI g p95 heat table**Each cell is one setup and one accepted valid condition; data fields carry risk flags.

**Breast ROI hotspot summary**Hotspot fraction of ROI by setup and accepted condition.

ROI ranking interpretation. The aggregate rank is a compact method-review summary, not a replacement for map review and not a clinical ranking. A setup can rank close to another overall while still improving selected directions or accelerations.

Hotspot interpretation. Hotspot fraction quantifies how much of the breast ROI exceeds the explicit R-dependent threshold. Small concentrated hotspots can matter even when aggregate mean g p95 is close.

Ranking Preview

rank_by_mean_g_p95	coil	mean_g_p95	worst_g_p95	clip_risk_conditions
1	Mammaria_18Ch	5.966164039571545	13.581063010920813	2
2	BI_7	6.5874477024734945	18.292166820148054	1
3	Mammaria_12Ch	7.556774627698841	19.580636855997728	2
4	Mammaria_8Ch	12.483730974935753	37.58584736826873	2

Hotspot Preview

coil	condition	hotspot_threshold	hotspot_fraction_roi	hotspot_pixels
BI_7	transversal_rl_r2	2.0	0.0	0
BI_7	transversal_rl_r3	5.0	0.01979405034324943	173
BI_7	sagittal_hf_r2	2.565481378420473	0.05	665
BI_7	sagittal_hf_r3	10.0	0.38533834586466165	5125
Mammaria_8Ch	transversal_rl_r2	2.0	0.0	0
Mammaria_8Ch	transversal_rl_r3	5.0	0.0	0
Mammaria_8Ch	sagittal_hf_r2	3.0	0.6197883597883598	5857
Mammaria_8Ch	sagittal_hf_r3	10.0	0.596931216931217	5641

Expert Comparison

Read the method-review ranking together with the red ROI boundaries and the condition-specific diagnostics. The ROI is geometry-derived from conductor placement and does not establish anatomy or clinical performance.

Setup	Rank	Mean g p95	Worst g p95	High-risk conditions	Clip-risk conditions
BI_7	2	6.5874477024734945	18.292166820148054	3	1
Mammaria_8Ch	4	12.483730974935753	37.58584736826873	2	2
Mammaria_12Ch	3	7.556774627698841	19.580636855997728	2	2
Mammaria_18Ch	1	5.966164039571545	13.581063010920813	2	2

Current Conclusion and Method Limitations

The current accepted-condition method-review ranking favors Mammaria_18Ch over BI_7 by mean breast-ROI g p95. Mammaria_18Ch should still be reviewed condition by condition; total channel count helps only when added elements provide independent sensitivity variation along the accepted alias direction inside the ROI. Its copied mean g p95 is 5.966164039571545.

This revision reduces outer-FOV influence by using the geometry-derived breast ROI for scaling, ranking, and hotspot summaries. Manual review must still confirm that the red ROI boundaries plausibly bracket the breast imaging volume for every setup and condition.

The conclusion remains conservative: no clinical validation, electromagnetic simulation claim, measured SNR claim, or hardware electronics claim is made here.

SNR visual source	current_step_overlays
g-factor/PPA source	physical-alias v4 breast ROI producer
ROI source	physical-alias v4 NPZ-mask-derived ROI ranking
Accepted setups	BI_7, Mammaria_8Ch, Mammaria_12Ch, Mammaria_18Ch
Accepted conditions	transversal_rl_r2, transversal_rl_r3, sagittal_hf_r2, sagittal_hf_r3
Excluded conditions	transversal_hf_r2, transversal_hf_r3, sagittal_rl_r2, sagittal_rl_r3
Contract status	pass