Search papers, labs, and topics across Lattice.
This study investigates the impact of conditioning-availability bias in echocardiographic segmentation models, revealing that models trained with cleaner auxiliary signals may perform poorly when deployed with less reliable signals. By measuring the performance gaps in oracle-estimated versus oracle-random pathways, the authors demonstrate that while some models can maintain usability under certain conditions, they exhibit significant sensitivity to incorrect phase inputs. The proposed deployment-aware checkpoint selection and phase perturbation methods effectively reduce performance gaps without compromising segmentation quality, highlighting the critical need for robust evaluation protocols in clinical applications.
Oracle-conditioned models can fail dramatically in real-world settings, revealing a hidden sensitivity to phase discrepancies that could jeopardize patient outcomes.
Conditional segmentation models may be trained and evaluated with auxiliary signals cleaner than those available at deployment. We study this protocol-level manifestation of shortcut learning and auxiliary-variable shift in phase-conditioned echocardiographic segmentation. The complementary gap pair measures loss on the deployable oracle-estimated pathway and probes sensitivity on the oracle-random pathway. On held-out CAMUS data, one strong-cyclic, oracle-selected run fails severely with estimated phase, while sensitivity to incorrect phase persists across three runs. On EchoNet-Dynamic, the current estimator remains usable, but random-phase testing reveals strong latent sensitivity. Deployment-aware checkpoint selection and phase perturbation reduce both gaps with little change in mean Dice. Exploratory subgroup analyses quantify variation across measured strata, and a downstream ejection fraction (EF) audit shows that recovering segmentation does not necessarily recover EF error or signed bias. Together, the gaps test whether oracle-conditioned performance survives the inference pathway actually available at deployment.
