Error Detection for PET/CT Radiology Reports: Domain-Specific vs Large Language Models

arXiv:2608.30021v1 Announce Type: cross Abstract: Errors in radiology reports can adversely affect patient treatment, yet automated report quality assurance remains challenging because errors are often subtle and require domain expertise to detect. Although large language models (LLMs) have recently been proposed for radiology report verification, their ability to detect clinically meaningful errors beyond chest X-ray datasets remains under-explored. To this end, we present the first systematic evaluation of language models for PET/CT report error detection, comparing compact domain-specific models with SOTA open-weight LLMs. We collected 30,633 oncology FDG PET/CT reports from 23 radiologists over 10 years. We trained domain-specific BERT models to detect clinically motivated synthetic reporting errors and evaluated alongside zero-/few-shot Qwen3-32B, Gemma-3-27B and Llama-3.3-70B on a held-out benchmark of 11,500 reports. A 15M-parameter model achieved 94.4% balanced accuracy with a 5.8% false-positive rate, compared with 84.0% for the strongest prompted LLM. Task-specific adaptation of Llama-3.3-70B closed this performance gap (94.4%) but retained substantially greater computational requirements. Our results suggest that domain-specific training matters more than model scale for PET/CT report error detection, supporting compact models as an accurate and computationally efficient approach to automated radiology report quality assurance.

Abstract:Errors in radiology reports can adversely affect patient treatment, yet automated report quality assurance remains challenging because errors are often subtle and require domain expertise to detect. Although large language models (LLMs) have recently been proposed for radiology report verification, their ability to detect clinically meaningful errors beyond chest X-ray datasets remains under-explored. To this end, we present the first systematic evaluation of language models for PET/CT report error detection, comparing compact domain-specific models with SOTA open-weight LLMs. We collected 30,633 oncology FDG PET/CT reports from 23 radiologists over 10 years. We trained domain-specific BERT models to detect clinically motivated synthetic reporting errors and evaluated alongside zero-/few-shot Qwen3-32B, Gemma-3-27B and Llama-3.3-70B on a held-out benchmark of 11,500 reports. A 15M-parameter model achieved 94.4% balanced accuracy with a 5.8% false-positive rate, compared with 84.0% for the strongest prompted LLM. Task-specific adaptation of Llama-3.3-70B closed this performance gap (94.4%) but retained substantially greater computational requirements. Our results suggest that domain-specific training matters more than model scale for PET/CT report error detection, supporting compact models as an accurate and computationally efficient approach to automated radiology report quality assurance.

Machine Learning (cs.LG); Artificial Intelligence (cs.AI)

arXiv:2608.30021 [cs.LG]

(or arXiv:2608.30021v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2608.30021 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

Error Detection for PET/CT Radiology Reports: Domain-Specific vs Large Language Models

Related coverage