Abstract
The claim that artificial intelligence will replace radiologists, we argue, targets the wrong mechanism and the wrong timeline. The threat is routinely attributed to autonomous, end-to-end models that interpret the whole imaging study; we contend these remain a long-horizon prospect of uncertain timing, and that the more realistic near-term prospect is human–AI collaboration in which a physician selects and verbalizes the salient imaging findings and a large language model (LLM) reasons over them together with the clinical context, assisting differential diagnosis (DDx) and improving the reporting workflow itself: drafting the conclusion and assigning structured-reporting categories. A plausible nearterm shift is that non-radiologist clinicians may increasingly use LLMs to reason over imaging findings that they or others have verbalized; this makes the fidelity of finding selection and verbalization a central bottleneck rather than a peripheral implementation detail. The key design variables are who verbalizes the findings, where verification occurs, when the model should defer, and how responsibility for the final clinical communication is preserved.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/HAIC26_010.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=cb1Wi4WokC
BibTex
@InProceedings{HanTae_Verbalized_MICCAISAT2026,
author = { Han, Taewon AND Shin, Jaeseung},
title = { { Verbalized Findings as the Interface for Near-Term Human–AI Collaboration in Radiology: A Clinical Perspective on LLM-Assisted Diagnosis } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17279},
month = {pending},
page = {pending}
}