Automated RECIST tumor response classification through prompt-guided large language models
Markus Mergen, Felix Busch, Andreas P. Sauter, Daniela Pfeiffer, Marcus R. Makowski, Daniel Spitzl, Florian T. Gassert · 2026
Read Automated RECIST tumor response classification through prompt-guided large language models by Markus Mergen, Felix Busch, Andreas P. Sauter, Daniela Pfeiffer, Marcus R. Makowski, Daniel Spitzl, Florian T. Gassert (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-13216305.txt.
Description
Markus Mergen and co-authors present a 2026 study testing whether a general-purpose language model can sort cancer scan reports into standard tumour response categories by itself. The model ran entirely offline inside hospital infrastructure, which keeps patient data private.
The team fed 142 real CT reports, with their expert labels hidden, to a locally hosted LLaMA-3.3 model using three prompting styles. Simple zero-shot and example-based prompts already performed strongly. Asking the model to reason step by step worked best, reaching a micro F1 score of 0.81 against expert labels across the five RECIST categories. The authors see prompt-guided offline models as a scalable way to standardize response assessment without sending data outside the hospital.
Catalog record
| CREATOR | Markus Mergen, Felix Busch, Andreas P. Sauter, Daniela Pfeiffer, Marcus R. Makowski, Daniel Spitzl, Florian T. Gassert |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |