Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

Difference from Background: Limit of Detection01:05

Difference from Background: Limit of Detection

The limit of detection (LOD) is the smallest amount of analyte that can be distinguished from the background noise. The LOD value corresponds to the concentration at which the analyte signal is three times larger than the standard deviation of the blank signal. Below this value, the analyte signal cannot be differentiated from the background noise. It is calculated by dividing the calibration slope by 3 times the standard deviation of the blank signals.
The LOD indicates the presence or absence...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Cortico-cortical and subcortico-cortical evoked potentials during general anesthesia.

Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology·2026
Same author

Clinical outcomes of conservative, endovascular and microsurgical management of unruptured giant intracranial aneurysms: 3-year follow-up of the prospective, multinational Giant Intracranial Aneurysm Registry.

Journal of neurology, neurosurgery, and psychiatry·2026
Same author

Application of Language Models for the Analysis of Adverse Drug Events in Pharmaceutical Research and Development: Scoping Review.

JMIR AI·2026
Same author

Genome-wide association study links COL6A6 and PIK3R4 to delayed cerebral ischaemia.

Brain : a journal of neurology·2026
Same author

Ethical Implications of Small Language Models (SLMs) in Healthcare Applications.

Studies in health technology and informatics·2026
Same author

CONORM-DEID: Robustness Evaluation of a Multilingual De-Identification System for Clinical Texts.

Studies in health technology and informatics·2026

Related Experiment Video

Updated: May 24, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Impact of LLM Scale and Quantization on Information Extraction from Clinical Text.

Alban Bornet1, Abiram Sandralegar2, Anthony Yazdani1

  • 1Department of Radiology and Medical Informatics, University of Geneva, Switzerland.

Studies in Health Technology and Informatics
|May 23, 2026
PubMed
Summary

Large Language Models (LLMs) can extract clinical data locally. Optimal performance for privacy-preserving healthcare applications is achieved with moderately sized models and 4-5-bit precision.

Keywords:
LLMsclinical NLPmodified Rankin Scalequantizationscaling law

More Related Videos

A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts
07:50

A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts

Published on: September 20, 2018

Related Experiment Videos

Last Updated: May 24, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts
07:50

A Metadata Extraction Approach for Clinical Case Reports to Enable Advanced Understanding of Biomedical Concepts

Published on: September 20, 2018

Area of Science:

  • Artificial Intelligence in Medicine
  • Natural Language Processing for Healthcare
  • Computational Neuroscience

Background:

  • Large Language Models (LLMs) demonstrate significant potential for extracting structured data from unstructured clinical text.
  • Healthcare adoption of LLMs is limited by privacy concerns, necessitating local deployment on resource-constrained systems.
  • Evaluating the impact of model size and quantization on LLM performance in clinical settings is crucial.

Purpose of the Study:

  • To systematically investigate the trade-offs between Large Language Model size and numerical precision (quantization) for clinical information extraction.
  • To assess the feasibility of using locally deployed, open-source LLMs for privacy-preserving healthcare applications.
  • To determine optimal model configurations for accurate clinical data extraction under resource limitations.

Main Methods:

  • Evaluated seven open-source Large Language Models (Qwen3, DeepSeek-R1 families, 0.6B-70B parameters) across six quantization levels (2-8 bits).
  • Assessed model performance on classifying the modified Rankin Scale (mRS) score from 105 neurosurgery patient notes.
  • Utilized linear mixed-effects analysis to determine the effects of model size, quantization, and their interaction on extraction error.

Main Results:

  • Both model size (β=-0.84, p<0.001) and quantization (β=-0.12, p<0.001) significantly impacted extraction error, with a notable interaction effect (β=0.08, p<0.001).
  • Larger models and higher numerical precision generally reduced extraction errors.
  • Performance gains from increased precision plateaued around 4-5 bits, indicating diminishing returns beyond this point.

Conclusions:

  • Clinically useful information extraction is achievable using moderately sized, locally deployed Large Language Models.
  • Quantization to 4-5 bits offers a balance between performance and computational efficiency for clinical LLM applications.
  • These findings support the feasibility of LLMs for privacy-preserving clinical data extraction in healthcare settings.