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

Feasibility of in-vivo 4D prompt-gamma treatment verification in proton therapy for pancreatic cancer.

Physics and imaging in radiation oncology·2026
Same author

LANTERN 2: Association Between Gene Molecular Profile and STAS in Lung Adenocarcinoma: A Comparative Analysis in a Prospective Real-World Population.

Genes·2026
Same author

Integrated NMR/MD investigation reveals differences after reweighting in conformational ensembles of GAAG and GCAA tetraloops.

RNA (New York, N.Y.)·2026
Same author

Dual-energy CT for proton therapy: Impact of advanced slice-wise patient-thickness estimation methods for improved stopping-power prediction.

Journal of applied clinical medical physics·2026
Same author

RNA•DNA:DNA Triplex Formation Modulates Individual Base Pair Stabilities in the DNA Target Duplex.

RNA (New York, N.Y.)·2026
Same author

Error detection sensitivity and operational efficiency of phantom-based and phantom-less patient-specific quality assurance in proton therapy.

Zeitschrift fur medizinische Physik·2026

Related Experiment Video

Updated: Jul 13, 2025

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

439

Multitask Learning with Convolutional Neural Networks and Vision Transformers Can Improve Outcome Prediction for Head

Sebastian Starke1,2,3, Alex Zwanenburg2,3,4, Karoline Leger2,3,4,5

  • 1Helmholtz-Zentrum Dresden-Rossendorf, Department of Information Services and Computing, 01328 Dresden, Germany.

Cancers
|October 14, 2023
PubMed
Summary

Multitask learning with convolutional neural networks (CNNs) improves head and neck cancer outcome prediction. This approach enhances treatment personalization by effectively stratifying patients by disease progression risk using CT and PET/CT imaging.

Keywords:
Cox proportional hazardsconvolutional neural networkdiscrete-time survival modelshead and neck cancerloco-regional controlmultitask learningprogression-free survivalsurvival analysistumor segmentationvision transformer

More Related Videos

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
Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
04:09

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

8.3K

Related Experiment Videos

Last Updated: Jul 13, 2025

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

439
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
Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
04:09

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

8.3K

Area of Science:

  • Artificial Intelligence in Oncology
  • Medical Imaging Analysis
  • Machine Learning for Cancer Prediction

Background:

  • Personalized treatment for head and neck cancer can be advanced by neural-network-based outcome predictions.
  • Developing effective neural networks is challenging with limited patient data.
  • Multitask learning (MTL) offers a potential solution to improve model performance in data-scarce scenarios.

Purpose of the Study:

  • To investigate if multitask learning strategies can enhance the performance of CNNs and Vision Transformers (ViTs).
  • To evaluate MTL by simultaneously optimizing two outcome objectives (multi-outcome) and incorporating a tumor segmentation task.
  • To assess the models' ability to predict loco-regional control (LRC) and progression-free survival (PFS) in head and neck cancer.

Main Methods:

  • Trained CNN and ViT models on two multicenter datasets: pre-treatment CT scans (290 patients) and combined PET/CT scans (224 patients).
  • Employed multitask learning, combining multi-outcome prediction with tumor segmentation loss.
  • Assessed model performance using concordance index (C-index) for discrimination and log-rank tests for risk stratification.

Main Results:

  • Multitask approaches generally showed improved performance across both datasets and model types (CNNs and ViTs).
  • Multi-outcome CNNs trained with segmentation loss emerged as the optimal strategy.
  • On PET/CT data, the best model achieved a C-index of 0.29 and significant risk stratification (p=0.003); on CT data, models achieved C-indices of 0.26 with significant LRC risk stratification (p=0.002 and p=0.011).

Conclusions:

  • Multitask learning, particularly with multi-outcome CNNs and segmentation loss, significantly enhances predictive performance for head and neck cancer.
  • These models demonstrate robust discrimination and risk stratification capabilities, paving the way for improved treatment personalization.
  • Further validation in prospective studies is planned to confirm the clinical utility of these advanced AI models.