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

Oncometabolite signatures from tumor-stroma crosstalk as potential non-invasive biomarkers.

Cell death discovery·2026
Same author

Sparse Temporal AutoEncoder for ECG Anomaly Detection.

Sensors (Basel, Switzerland)·2026
Same author

Gated Attention-Augmented Double U-Net for White Blood Cell Segmentation.

Journal of imaging·2025
Same author

Diagnostic Accuracy of Magnified Dermoscopy and Reflectance Confocal Microscopy in Assessing Melanocytic Lesions.

Dermatology practical & conceptual·2025
Same author

Magnified dermoscopy terminology for skin tumours: International Dermoscopy Society Delphi consensus.

Journal of the European Academy of Dermatology and Venereology : JEADV·2025
Same author

An Innovative IoT and Edge Intelligence Framework for Monitoring Elderly People Using Anomaly Detection on Data from Non-Wearable Sensors.

Sensors (Basel, Switzerland)·2025

Related Experiment Video

Updated: Jun 26, 2026

Dermoscopy Aids in the Diagnosis of Discoid Lupus Erythematosus
05:39

Dermoscopy Aids in the Diagnosis of Discoid Lupus Erythematosus

Published on: May 16, 2025

When AI and Experts Agree on Error: Intrinsic Ambiguity in Dermatoscopic Images.

Loris Cino1, Pier Luigi Mazzeo2, Alessandro Martella3

  • 1Dipartimento di Ingegneria Informatica, Automatica e Gestionale (DIAG), Sapienza Università di Roma, Via Ariosto, 25, 00185 Rome, Italy.

Journal of Imaging
|June 25, 2026
PubMed
Summary

Artificial intelligence (AI) in dermatology shows promise, but some skin images are difficult for even experts to diagnose. Poor image quality causes both AI and human diagnostic errors.

Keywords:
artificial intelligence (AI)convolutional neural networks (CNNs)deep learningdermatologyimage qualitymachine learning (ML)medical image analysisstatistical analysis

More Related Videos

Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
09:37

Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition

Published on: August 18, 2022

Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
05:49

Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images

Published on: February 23, 2024

Related Experiment Videos

Last Updated: Jun 26, 2026

Dermoscopy Aids in the Diagnosis of Discoid Lupus Erythematosus
05:39

Dermoscopy Aids in the Diagnosis of Discoid Lupus Erythematosus

Published on: May 16, 2025

Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
09:37

Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition

Published on: August 18, 2022

Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
05:49

Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images

Published on: February 23, 2024

Area of Science:

  • Dermatology
  • Artificial Intelligence
  • Medical Imaging Analysis

Background:

  • Artificial intelligence (AI), especially convolutional neural networks (CNNs), holds significant potential for improving dermatological diagnosis.
  • Current research often compares AI performance to human experts, but the intrinsic complexity of dermatoscopic images remains underexplored.

Purpose of the Study:

  • To investigate the reasons behind systematic misclassifications of dermatoscopic images by multiple CNN architectures.
  • To determine if AI failures are due to algorithmic bias or inherent visual ambiguity in the images.

Main Methods:

  • Multiple CNN architectures were tested on a dataset of dermatoscopic images.
  • A subset of images consistently misclassified by all CNNs was identified.
  • Expert dermatologists independently evaluated these challenging images and a control group, assessing agreement with ground truth and inter-rater reliability.

Main Results:

  • A statistically significant subset of images was systematically misclassified by all tested CNNs.
  • Human diagnostic performance collapsed on these AI-misclassified images, with significantly lower agreement with ground truth (Cohen's kappa = 0.08) and reduced inter-rater reliability (Fleiss' kappa = 0.275) compared to control images.
  • Poor image quality was identified as a primary factor contributing to both AI and human diagnostic failures.

Conclusions:

  • Inherent visual ambiguity and poor image quality in dermatoscopic images pose significant challenges for AI diagnostic tools.
  • These factors also severely impact human expert diagnostic performance, highlighting a critical limitation in current AI benchmarking.
  • Publicly releasing data, code, and models promotes transparency and reproducibility in AI-driven dermatological research.