SMASH 2026 · Paper #6 · Result card
From Canon to Nikon: Camera Domain Shift and Potential Equity Implications in Retinal AI
With R$1,900,000 for 40 sites, all-Nikon covers every site and misses 955 cases a year, against 2,431 for all-Canon on 20 sites. Diabetic retinopathy AUROC holds across cameras (0.940 vs 0.946).
task Diabetic Retinopathyclassifier @ resolution LR @ 672 pxbudget R$1,900,000sites × patients/yr 40 × 1,000prevalence 10%specificity 0.9re-imaged if inadequate 50%synthetic shift 1.0
Diabetic Retinopathy: Canon → Nikon
0.940 → 0.946
AUROC, LR @ 672 px (paper Table 4); gap -0.006
Missed cases / yr: all-Canon vs all-Nikon
2,431 vs 955
R$1.90M buys 20 Canon or 40 Nikon sites of 40 (Diabetic Retinopathy)
Equity gap at Nikon sites
1.11×
risk that a diseased patient is missed at a Nikon vs a Canon site: 24% vs 22% (sensitivity 0.84 vs 0.82 at spec 0.90; 18.60% vs 8.79% inadequate images)
What drives the gap?
task 67%
share of variance in the internal−external gap across 220 configurations; classifier 1%, resolution 0%
From the abstract
In a Brazilian cohort of 8,524 patients, retinal images acquired with the Nikon NF5050 were more than twice as likely to have inadequate image quality compared with those acquired with the Canon CR-2, despite refurbished Canon systems costing more than twice as much as Nikon systems. We find that camera-associated performance differences vary substantially across tasks, and that neither higher input resolution nor classifier choice consistently mitigates cross-camera variation.
Evidence
Paper vs. your run
| Metric | Paper | Your run | |
|---|---|---|---|
| Diabetic Retinopathy internal → external (Table 2: LR, 672 px) | 0.940 → 0.946 | 0.940 → 0.946 (LR, 672 px) | transcribed, not recomputed |
| Inadequate images, Canon vs Nikon (Table 1) | 8.79% vs 18.60% | 8.79% vs 18.60% | calculator input; discussion rounds to 8.8% vs 18.6% |
| Image-field abnormality, Nikon | 15.84% (Table 1) / 16.5% (text) | 15.84% (used in demo) | paper is internally inconsistent |
| Illumination Cramér's V | 0.028 (Table 1) / 0.023 (text) | — | inconsistent in paper; not used |
| Camera price ratio | "approximately half" (abstract); "1.5 times more" (intro); R$95,000 vs R$35,000 | 2.71× (from the quoted prices) | calculator uses the prices |
| Hemorrhage / hypertensive ΔAUROC (LR, all resolutions) | ∆ AUROC=0.15–0.28 (Fig. 2 caption) | 0.134–0.276 from Table 4 | hemorrhage at 952 px is 0.134 |
| Resolution / classifier fix the gap? | neither ... consistently mitigated cross-camera variation | 952 vs 224 px shrinks the gap for 2/11 tasks, widens it for 6/11 | Table-2 classifier per task |
Try next
- Hemorrhage — a camera-fragile task where Canon wins despite covering half the sites — Hemorrhage has the largest internal−external drop (+0.235 AUROC). At R$1.9M, all-Canon (20 sites, 3,098 missed/yr) is expected to beat all-Nikon (40 sites, 3,455 missed/yr)—the accuracy loss overwhelms the coverage gain, which is the equity case the paper warns about.
{"task": "Hemorrhage", "classifier": "best", "resolution": "best", "budget_brl": 1900000, "n_sites": 40, "patients_per_site": 1000, "prevalence_pct": 10, "specificity": 0.9, "recapture_pct": 50, "shift_strength": 1.0, "seed": 7} - R$3,800,000 budget — the crossover where all-Canon can cover every site — R$3,800,000 = 40 × R$95,000, so all-Canon finally achieves full coverage. This tests whether removing the coverage disadvantage reverses the conclusion for DR; it also shows the budget level at which the best mix converges with all-Canon.
{"task": "Diabetic Retinopathy", "classifier": "best", "resolution": "best", "budget_brl": 3800000, "n_sites": 40, "patients_per_site": 1000, "prevalence_pct": 10, "specificity": 0.9, "recapture_pct": 50, "shift_strength": 1.0, "seed": 7} - No re-imaging (recapture_pct 0) with Hypertensive Retinopathy — worst-case equity gap — Setting recapture to 0 means every inadequate image is a permanently missed patient. Combined with Hypertensive Retinopathy's large accuracy drop (+0.223), this isolates how much of the equity risk at Nikon sites comes from image quality alone vs classifier fragility. Expect a substantially higher equity ratio.
{"task": "Hypertensive Retinopathy", "classifier": "best", "resolution": "best", "budget_brl": 1900000, "n_sites": 40, "patients_per_site": 1000, "prevalence_pct": 10, "specificity": 0.9, "recapture_pct": 0, "shift_strength": 1.0, "seed": 7}
Read this before sharing
- Model performance is transcribed from the paper's Table 4; the screening calculator is an extension.
- Camera type is confounded with case mix and site in the source cohort.