Jove
Visualize
Contact Us

Related Concept Videos

You might also read

Related Articles

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

Sort by
Same author

Towards a Reliable and Rapid Automated Grading System in Facial Palsy Patients: Facial Palsy Surgery Meets Computer Science.

Journal of clinical medicine·2022
Same author

Correction to: New aspects in digital breast assessment: further refinement of a method for automated digital anthropometry.

Archives of gynecology and obstetrics·2021
Same author

New aspects in digital breast assessment: further refinement of a method for automated digital anthropometry.

Archives of gynecology and obstetrics·2020
See all related articles
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 Experiment Video

Updated: Jun 12, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.7K

Enhancing generalization in zero-shot multi-label endoscopic instrument classification.

Raphaela Maerkl1, Tobias Rueckert2,3, David Rauber2

  • 1Regensburg Medical Image Computing (ReMIC), OTH Regensburg, 93053, Regensburg, Germany. raphaela.maerkl@st.oth-regensburg.de.

International Journal of Computer Assisted Radiology and Surgery
|June 11, 2025
PubMed
Summary
This summary is machine-generated.

Improving zero-shot learning for medical AI, this study uses sentence embeddings and z-score normalization to enhance recognition of unseen surgical instruments, boosting accuracy significantly.

Keywords:
Generalized zero-shot learningMulti-label classificationSentence embeddingsSurgical instrumentsZ-score normalization

More Related Videos

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
05:56

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

2.4K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

483

Related Experiment Videos

Last Updated: Jun 12, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.7K
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
05:56

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

2.4K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

483

Area of Science:

  • Computer Vision
  • Machine Learning
  • Medical Imaging

Background:

  • Neural networks struggle with generalizing to unseen classes, a critical issue in safety-critical medical applications.
  • Zero-shot learning (ZSL) offers a solution by leveraging semantic data, but performance hinges on embedding quality.

Purpose of the Study:

  • To investigate the efficacy of full descriptive sentence embeddings versus simpler word embeddings for ZSL in medical image recognition.
  • To evaluate the impact of z-score normalization on embedding performance for unseen classes.

Main Methods:

  • Utilized Sentence-BERT for generating descriptive sentence embeddings as class representations.
  • Compared sentence embeddings with BERT-derived word embeddings.
  • Applied z-score normalization as a post-processing step.
  • Evaluated on a multi-label generalized zero-shot learning task for surgical instrument recognition in endoscopic images.

Main Results:

  • Combining sentence embeddings with z-score normalization significantly improved performance on unseen classes.
  • Area Under the Receiver Operating Characteristic Curve (AUROC) for unseen classes increased from 43.9% to 64.9%.
  • Multi-label accuracy for unseen classes rose from 26.1% to 79.5%.

Conclusions:

  • Sentence embeddings and z-score normalization substantially enhance the generalization capabilities of zero-shot learning models.
  • The proposed method shows promise for improving reliability in medical AI applications.
  • Further validation across diverse datasets and domains is recommended to confirm robustness.