Large Language Models in Radiologic Numerical Tasks: A Thorough Evaluation and Error Analysis.

The purpose of this study was to investigate the performance of LLMs in radiology numerical tasks and perform a comprehensive error analysis. We defined six tasks: extracting (1) minimum T-score from DEXA report, (2) maximum common bile duct (CBD) diameter from ultrasound report, and (3) maximum lun...

Full description

Bibliographic Details
Published in:Journal of Imaging Informatics in Medicine p. 1
Main Authors: Nowroozi, Ali, Bondarenko, Masha, Serapio, Adrian, Schnitzler, Tician, Brar, Sukhmanjit S., Sohn, Jae Ho
Format: Journal Article
Published: Springer Nature Jan2026
Online Access:View this record in EBSCOhost
fields @attributes:
  recordID: 1
pdfLink:
plink: https://search.ebscohost.com/login.aspx?direct=true&db=ccm&AN=191004292&site=ehost-live
header:
  @attributes:
    shortDbName: ccm
    uiTerm: 191004292
    longDbName: CINAHL Complete
    uiTag: AN
  controlInfo:
    bkinfo:
    dissinfo:
    jinfo:
      jid:
        29482925
        NR3A
      jtl: Journal of Imaging Informatics in Medicine
      issn: 29482925
      maglogo: N
    pubinfo:
      dt: Jan2026
      pid: 237
      pub: Springer Nature
      place: New York, New York
    artinfo:
      ui:
        191004292
        10.1007/s10278-025-01824-9
        191004292
      ppf: 1
      formats:
      tig:
        atl: Large Language Models in Radiologic Numerical Tasks: A Thorough Evaluation and Error Analysis.
      aug:
        au:
          Nowroozi, Ali
          Bondarenko, Masha
          Serapio, Adrian
          Schnitzler, Tician
          Brar, Sukhmanjit S.
          Sohn, Jae Ho
        affil: Center for Intelligent Imaging, Department of Radiology and Biomedical Imaging, University of California, San Francisco (UCSF)
      sug:
      ab: The purpose of this study was to investigate the performance of LLMs in radiology numerical tasks and perform a comprehensive error analysis. We defined six tasks: extracting (1) minimum T-score from DEXA report, (2) maximum common bile duct (CBD) diameter from ultrasound report, and (3) maximum lung nodule size from CT report, and judging (1) presence of a highly hypermetabolic region on a PET report, (2) whether a patient is osteoporotic based on a DEXA report, and (3) whether a patient has a dilated CBD based on an ultrasound report. Reports were extracted from the MIMIC III and our institution’s databases, and the ground truths were extracted manually. The models used were Llama 3.1 8b, DeepSeek R1 distilled Llama 8b, OpenAI o1-mini, and OpenAI GPT-5-mini. We manually reviewed all incorrect outputs and performed a comprehensive error analysis. In extraction tasks, while Llama showed relatively variable results (ranging 86%–98.7%) across tasks, other models performed consistently well (accuracies > 95%). In judgment tasks, the lowest accuracies of Llama, DeepSeek distilled Llama, o1-mini, and GPT-5-mini were 62.0%, 91.7%, 91.7%, and 99.0%, respectively, while o1-mini and GPT-5-mini did reach 100% performance in detecting osteoporosis. We found no mathematical errors in the outputs of o1-mini and GPT-5-mini. Answer-only output format significantly reduced performance in Llama and DeepSeek distilled Llama but not in o1-mini or GPT-5-mini. To conclude, reinforcement learning (RL) reasoning models perform consistently well in radiology numerical tasks and show no mathematical errors. Simpler non-RL reasoning models may also achieve acceptable performance depending on the task.
      pubtype: Academic Journal
      doctype: Journal Article
      ougenre: Article
    language: English
    refInfo:
    holdings:
      @attributes:
        islocal: N