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

You might also read

Related Articles

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

Sort by
Same author

RNF2 mediates H2A ubiquitination to promote colitis via suppressing monocyte-macrophage transition in mice.

Nature communications·2026
Same author

Mesh-represented and learning-empowered hologram synthesis for full 3D holographic displays.

Nature communications·2026
Same author

Macrophage USP9X attenuates colitis by restricting oncostatin M production via K27-linked deubiquitination of STAT1.

Cellular & molecular immunology·2026
Same author

Ranking of potential hazards from engineered nanoparticles and micro/nano-plastics in food systems.

Environment international·2026
Same author

Multifunctional Manganese-Based Nanocomposites for Enhanced Immunotherapy by Regulating Lactate Metabolism and Activating the cGAS-STING Pathway.

Small (Weinheim an der Bergstrasse, Germany)·2026
Same author

Clinical outcomes and spatial transcriptomic profiles of CD19/20 CAR-T therapy in relapsed or refractory B-cell non-Hodgkin's lymphoma.

Journal for immunotherapy of cancer·2026
Same journal

Real-world pharmacovigilance of drug-related bone metabolism disorders: integrating FAERS and VigiAccess with a Bradford Hill-based causal plausibility assessment.

Frontiers in pharmacology·2026
Same journal

Voriconazole therapeutic drug monitoring and safety in HIV-infected patients with invasive fungal disease.

Frontiers in pharmacology·2026
Same journal

Hepatoprotective effects of astaxanthin and kokum on lipid peroxidation and fatty acid profile in mice with diet-induced NAFLD.

Frontiers in pharmacology·2026
Same journal

Modified sini powder for the management of postoperative depression in non-small cell lung cancer patients: a multicenter, randomized, double-blind, placebo-controlled trial protocol.

Frontiers in pharmacology·2026
Same journal

Cepharanthine improves blood glucose levels and mitigates renal injury in streptozotocin-induced diabetic rats via restoring pancreatic β-cell integrity and reducing inflammation.

Frontiers in pharmacology·2026
Same journal

Integrative management of gastric cardia cancer and precancerous lesions: prospects and mechanistic rationale for Banxia Xiexin Tang.

Frontiers in pharmacology·2026
See all related articles

Related Experiment Video

Updated: Jul 19, 2025

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

1.9K

A universal lesion detection method based on partially supervised learning.

Xun Wang1,2, Xin Shi1, Xiangyu Meng1

  • 1Department of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, China.

Frontiers in Pharmacology
|August 18, 2023
PubMed
Summary
This summary is machine-generated.

Partially supervised learning (PSL) improves universal lesion detection (ULD) segmentation models by introducing a novel loss function. This method effectively reduces negative anchor misclassification, enhancing ULD detector performance on CT images.

Keywords:
3D modelPSLULDmedical image learningneural network learning

More Related Videos

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
07:15

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Published on: August 16, 2020

6.9K
Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis
09:16

Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis

Published on: June 18, 2020

6.9K

Related Experiment Videos

Last Updated: Jul 19, 2025

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

1.9K
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
07:15

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Published on: August 16, 2020

6.9K
Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis
09:16

Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis

Published on: June 18, 2020

6.9K

Area of Science:

  • Medical Imaging
  • Machine Learning
  • Computer-Aided Diagnosis

Background:

  • Partially supervised learning (PSL) is essential for developing efficient universal lesion detection (ULD) segmentation models.
  • Acquiring fully annotated datasets for ULD is challenging due to the large volume of Computed Tomography (CT) images and a shortage of expert annotators for computer-aided detection/diagnosis (CADe/CADx).
  • Existing methods struggle with performance degradation caused by the misclassification of lesion areas (positive samples) as negative bounding boxes.

Purpose of the Study:

  • To propose a novel loss function for ULD segmentation models that mitigates the negative impact of misclassified anchors.
  • To enhance the accuracy and efficiency of ULD segmentation using partially annotated datasets.

Main Methods:

  • A novel loss function is introduced that generates a mask to selectively reduce the number of negative anchors considered during loss calculation.
  • A parameter is employed to control the proportion of negative samples, minimizing the adverse effects of misclassification on the ULD model.
  • Experiments were conducted using a 3D framework on the DeepLesion dataset, a large-scale public dataset for ULD from CT images.

Main Results:

  • The proposed loss function significantly improved the performance of the universal lesion detection detector.
  • Extensive experiments were performed to optimize the parameter for the loss function, identifying the most suitable value for enhanced performance.
  • The method demonstrated a notable reduction in the adverse effects of misclassification on the ULD model's accuracy.

Conclusions:

  • The novel loss function effectively addresses the challenge of negative anchor misclassification in ULD segmentation.
  • The proposed approach offers a promising solution for developing efficient ULD models with partially annotated datasets.
  • The method shows significant performance improvements for ULD detectors, contributing to advancements in medical image analysis and CADe/CADx.