Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
This paper quantifies the Bayesian uncertainty in the agreement of a fixed breast-cancer pathology AI with a future qualified reader, accounting for reader- and annotation-condition-dependent reference labels. By analyzing linked Evaluation and Unbiased NuCLS multirater data using a fixed external HoVer-Net decision vector, the study reveals that while absolute agreement is structurally sensitive and varies significantly, the contrast between AI agreement states under different conditions remains comparatively stable. The findings underscore the critical need to report detailed metadata in AI evaluations using variable human references to ensure robust and interpretable results.
Human reference labels for medical AI are more variable than previously assumed, with nearly 20% of reader-anchor pairs flipping AI-agreement state across annotation conditions.
Human reference labels used to evaluate medical AI can vary across readers and annotation workflows. We quantified agreement of a fixed pathology AI system with a future qualified reader without treating inferred consensus as error-free biological truth. Linked Evaluation and Unbiased NuCLS multirater data were analyzed using a fixed external HoVer-Net decision vector. The selected frame contained 1,144 anchors, 52 fields of view, five slides/patients, and five primary readers. The target was fixed-AI agreement with a future reader, with descriptive contrast $\Delta=\theta_E-\theta_U$. Analyses included raw agreement and iMRMC benchmarks, reader-only and shared-detectability Bayesian models, structural sensitivities, adversarial simulations, and predictive assessment. Among 5,188 reader-anchor pairs observed in both conditions, 19.62% changed AI-agreement state, while raw agreement was 0.5466 in Unbiased and 0.5473 in Evaluation. The reader-only model gave $\Delta=-0.0008$ (95% CrI, $-0.0033$ to $0.0008$), and the shared-detectability model gave $\Delta\approx0$ ($-0.0140$ to $0.0127$). Across structural perturbations, absolute agreement ranged from 0.420 to 0.646 in Unbiased and 0.409 to 0.640 in Evaluation. Frozen adversarial simulations met the comparative-stability criterion (median $R=0.295$; 14/16 model-scenario cells had $R<1$). Held-out-AI prediction supported shared predictive information, but varied by slide composition. The Evaluation-minus-Unbiased contrast was comparatively stable, whereas absolute agreement was structurally sensitive. These findings do not establish equivalence, causal assistance effects, biological accuracy, or new-patient transportability. Evaluations using variable human references should report reader population, selected frame, observation mechanism, and structural sensitivity.