How green are large language models for radiology report labelling? Comparing human, rule-based and hybrid workflows
Matthias A. Fink, Arved Bischoff, Edem Atsiatorme, Alexander Kremer, Jonas Kroschke, Martin Moll, Patrick Stein, Veronika Riebl, Timo Leichenich, Hans-Ulrich Kauczor, Kai Schlamp · 2026
Read How green are large language models for radiology report labelling? Comparing human, rule-based and hybrid workflows by Matthias A. Fink, Arved Bischoff, Edem Atsiatorme, Alexander Kremer, Jonas Kroschke, Martin Moll, Patrick Stein, Veronika Riebl, Timo Leichenich, Hans-Ulrich Kauczor, Kai Schlamp (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13216412.txt.
Description
Matthias A. Fink and co-authors present a 2026 companion study to their report-labelling work, this time weighing the environmental and economic cost of each labelling route for CT pulmonary embolism reports. Hospitals want automation, but large language models consume real energy.
The same 2,923 reports were labelled four ways: by hand, by a rule-based extractor, by language models alone, and by a hybrid that escalated only rule failures. Manual labelling took 32.8 hours at 95.0 percent accuracy. Model-only pipelines were faster and cheaper per report in labour terms yet less accurate, at 85.1 percent. The hybrid won on every axis, reaching 98.5 percent accuracy while cutting time to under an hour in aggregate comparisons, with cost and carbon dioxide emissions falling roughly sevenfold against model-only runs.
Catalog record
| CREATOR | Matthias A. Fink, Arved Bischoff, Edem Atsiatorme, Alexander Kremer, Jonas Kroschke, Martin Moll, Patrick Stein, Veronika Riebl, Timo Leichenich, Hans-Ulrich Kauczor, Kai Schlamp |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |