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

Redefining Cytotechnologist Practice in Sweden: National Survey on Workload, Expanded Roles, and Future Competency Needs.

Acta cytologica·2026
Same author

Incremental prognostic value of immune cell densities beyond clinical parameters in non-small cell lung cancer.

Lung cancer (Amsterdam, Netherlands)·2026
Same author

UpStory: the uppsala storytelling dataset.

Frontiers in robotics and AI·2025
Same author

Let it shine: Autofluorescence of Papanicolaou-stain improves AI-based cytological oral cancer detection.

Computers in biology and medicine·2024
Same author

Cross-modality sub-image retrieval using contrastive multimodal image representations.

Scientific reports·2024
Same author

Early Detection of Oral Potentially Malignant Disorders: A Review on Prospective Screening Methods with Regard to Global Challenges.

Journal of maxillofacial and oral surgery·2024

Related Experiment Video

Updated: Jun 27, 2025

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

2.5K

Deep multiple instance learning versus conventional deep single instance learning for interpretable oral cancer

Nadezhda Koriakina1, Nataša Sladoje1, Vladimir Bašić2,3

  • 1Centre for Image Analysis, Department of Information Technology, Uppsala University, Uppsala, Sweden.

Plos One
|April 30, 2024
PubMed
Summary

This study compared single instance learning (SIL) and multiple instance learning (MIL) for oral cancer (OC) detection. Surprisingly, SIL performed better or equal to MIL on both synthetic and real datasets, identifying abnormal and malignant cells.

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

2.7K
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
05:49

Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization

Published on: February 23, 2024

832

Related Experiment Videos

Last Updated: Jun 27, 2025

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

2.5K
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
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
05:49

Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization

Published on: February 23, 2024

832

Area of Science:

  • Medical Imaging
  • Computational Pathology
  • Artificial Intelligence in Oncology

Background:

  • Histological examination is the standard for oral cancer (OC) diagnosis but is invasive and time-consuming.
  • Cytological analysis of brush samples is less invasive but faces challenges in clinical implementation due to resource and expert limitations.
  • Deep learning methods are explored to assist cytotechnologists in OC detection, requiring reliable cancer detection with interpretable insights into cell relevance.

Purpose of the Study:

  • To compare the performance of single instance learning (SIL) and multiple instance learning (MIL) for oral cancer detection and interpretation.
  • To evaluate deep learning approaches using both real oral cancer data and a novel synthetic dataset (PAP-QMNIST) for systematic assessment.
  • To assess the ability of these methods to identify diagnostically relevant cells for improved understanding and supervision.

Main Methods:

  • Comparison of conventional SIL and modern MIL approaches for OC detection.
  • Utilized a real oral cancer dataset with patient-level annotations and a synthetic PAP-QMNIST dataset with per-instance ground truth.
  • Evaluated performance across three different neural network architectures on both datasets.

Main Results:

  • The single instance learning (SIL) approach demonstrated performance equal to or better than the multiple instance learning (MIL) approach on both synthetic and real oral cancer datasets.
  • Both methods successfully identified cells deviating from normality, including malignant and dysplasia-suspicious cells, as confirmed by visual examination.
  • The study provides open-source code for the developed methods.

Conclusions:

  • SIL may be a more effective approach than MIL for oral cancer detection and interpretation in the evaluated scenarios.
  • Deep learning models can assist cytotechnologists by highlighting abnormal cells, potentially improving diagnostic accuracy and efficiency.
  • The synthetic PAP-QMNIST dataset serves as a valuable proxy for real-world data, facilitating research in computational pathology.