Establish the value of reference cases and text as clinical feedback channels

Establish whether a reference case or short text can reliably communicate domain context or task intent in clinician-in-the-loop few-shot medical image analysis, and determine their value relative to other feedback modalities.

Background

The framework distinguishes several forms of clinician feedback, including points, boundary corrections, reference cases, and text. Points and boundary corrections provide geometric information, whereas reference cases and text may convey acquisition context or task intent. The paper identifies the latter channels as plausible but unvalidated components of clinician-in-the-loop few-shot medical image analysis.

References

Points and short boundary corrections already provide useful geometric information; proposed channels such as a reference case or text may communicate domain context or task intent, but their value remains to be established.

— From Few-Shot Segmentation to Clinician-in-the-Loop Medical Image Analysis  (2609.10001 - Zhu, 9 Sep 2026) in Introduction, Section 1

Reference conditioning may reduce this difficulty, but our design does not identify the mechanism, and other explanations, such as the attention given to long constraint lines, remain open.

— It's Always 10:10: Reference Images Break a Bias That Prompts Only Dent  (2610.11320 - Cazzaniga, 8 Oct 2026) in Discussion, paragraph “Why words did not help more”