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

Prospective and external validation of stroke discharge planning machine learning models.

Journal of clinical neuroscience : official journal of the Neurosurgical Society of Australasia·2022
Same author

Assessment of the effect of a comprehensive chest radiograph deep learning model on radiologist reports and patient outcomes: a real-world observational study.

BMJ open·2021
Same author

Do comprehensive deep learning algorithms suffer from hidden stratification? A retrospective study on pneumothorax detection in chest radiography.

BMJ open·2021
Same author

The false hope of current approaches to explainable artificial intelligence in health care.

The Lancet. Digital health·2021
Same author

Medical student knowledge and critical appraisal of machine learning: a multicentre international cross-sectional study.

Internal medicine journal·2021
Same author

Assessing the accuracy of <sup>68</sup> Ga-PSMA PET/CT compared with MRI in the initial diagnosis of prostate malignancy: A cohort analysis of 114 consecutive patients.

Journal of medical imaging and radiation oncology·2021
Same journal

Improving Reliability of MRI Lumbar Spinal Stenosis Assessment Across Radiology and Spine Specialties: Impact of a Structured Education Intervention.

Academic radiology·2026
Same journal

Advances in CT and MRI for Yttrium-90 Radioembolization of Hepatocellular Carcinoma.

Academic radiology·2026
Same journal

Homogeneity of Liver Fat Distribution Serves as a Diagnostic Marker for Metabolic Dysfunction-Associated Steatohepatitis.

Academic radiology·2026
Same journal

MRI-based Predictors and Risk Constellations of Chronic Ankle Instability After Acute Lateral Ankle Sprain: A Multicenter Study.

Academic radiology·2026
Same journal

Early Prediction of Pathological Complete Response to Neoadjuvant Chemotherapy in Breast Cancer using a Longitudinal US-based Stack-model.

Academic radiology·2026
Same journal

Evaluating the Impact of Embolization on Outcomes in Iliopsoas Hematomas: A Multicenter Retrospective Propensity-matched Study.

Academic radiology·2026
See all related articles

Related Experiment Video

Updated: Jan 4, 2026

Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function
02:09

Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function

Published on: April 12, 2024

960

Exploring Large-scale Public Medical Image Datasets.

Luke Oakden-Rayner1

  • 1Australian Institute for Machine Learning, North Terrace, Adelaide, Australia; School of Public Health, University of Adelaide, North Terrace, Adelaide 5000, Australia; Royal Adelaide Hospital, North Terrace, Adelaide, Australia.

Academic Radiology
|November 11, 2019
PubMed
Summary
This summary is machine-generated.

Public medical imaging datasets like ChestXray14 and MURA have labeling inaccuracies. Visual inspection is crucial for quality control to ensure the reliability of artificial intelligence (AI) training data.

Keywords:
Artificial intelligencedatasetdeep learningexploratory analysisquality control

More Related Videos

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

3.3K
Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics
10:17

Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics

Published on: January 8, 2018

13.6K

Related Experiment Videos

Last Updated: Jan 4, 2026

Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function
02:09

Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function

Published on: April 12, 2024

960
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

3.3K
Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics
10:17

Guidelines and Experience Using Imaging Biomarker Explorer IBEX for Radiomics

Published on: January 8, 2018

13.6K

Area of Science:

  • Medical imaging analysis
  • Artificial intelligence in healthcare
  • Radiology informatics

Background:

  • Medical artificial intelligence (AI) relies on large, well-characterized datasets.
  • Publicly available datasets are valuable but present challenges due to the gap between data generation and usage.
  • These challenges can limit the utility and reliability of AI models trained on such data.

Purpose of the Study:

  • To visually explore and assess the accuracy of labels in two large public medical imaging datasets: ChestXray14 and Musculoskeletal Radiology (MURA).
  • To identify potential subtle problems within these datasets beyond label accuracy, such as hidden stratification and label disambiguation failures.

Main Methods:

  • Visual inspection of a subset of approximately 700 images from both the ChestXray14 (112,120 frontal chest films) and MURA (40,561 upper limb radiographs) datasets.
  • Review of image subsets by a board-certified radiologist to determine the quality and accuracy of the original dataset labels.

Main Results:

  • ChestXray14 labels showed significant inaccuracies, with positive predictive values 10-30% lower than reported, alongside issues like hidden stratification and label disambiguation failure.
  • MURA dataset labels were generally more accurate, but normal/abnormal labels for degenerative joint disease cases had a sensitivity of 60% and specificity of 82%.

Conclusions:

  • Visual inspection of medical imaging data is essential for understanding dataset content and ensuring quality.
  • Dataset creators should implement rigorous visual quality control and provide detailed documentation on data generation procedures and labeling rules.
  • Improved dataset quality control is necessary to enhance the reliability of AI systems in medical applications.