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 Experiment Videos

CROSS-MODAL FINE-TUNING OF 3D CONVOLUTIONAL FOUNDATION MODELS FOR ADHD CLASSIFICATION WITH LOW-RANK ADAPTATION.

Jyun-Ping Kao1,2, Shinyeong Rho1,3, Shahar Lazarev4

  • 1Massachusetts General Brigham and Harvard Medical School, Boston, MA, USA.

Proceedings. IEEE International Symposium on Biomedical Imaging
|May 25, 2026
PubMed
Summary

Related Concept Videos

Attention-Deficit/Hyperactivity Disorder01:30

Attention-Deficit/Hyperactivity Disorder

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder characterized by persistent inattention, hyperactivity, and impulsivity. It affects approximately 5-8% of children globally, with around 60-70% of cases persisting into adulthood. ADHD has significant implications for educational attainment, social interactions, and occupational success.
Diagnostic Criteria and Symptoms
To diagnose ADHD, symptoms must manifest before age 12 and be evident across multiple settings.

You might also read

Related Articles

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

Sort by
Same author

GUSL: A novel and efficient machine learning model for prostate segmentation on MRI.

Computers in biology and medicine·2026
Same author

Microglial clonal dynamics and the impact of clonal hematopoiesis in autologously transplanted rhesus macaques.

Cell reports·2026
Same author

A scoping review of traditional and artificial intelligence methods in malaria diagnostics.

NPJ digital medicine·2026
Same author

Synthesizing vocal tract magnetic resonance imaging sequences with phoneme-aware diffusion models.

Journal of medical imaging (Bellingham, Wash.)·2026
Same author

DSHARP: Deep Incompressible Motion Estimation With Sinusoidal-Transformed Harmonic Phase for Tagged MRI.

IEEE transactions on medical imaging·2026
Same author

Individualized Treatment Effect Inference of Head and Neck Cancer with Multimodal Data.

APSIPA Transactions on Signal and Information Processing·2026

This study introduces an efficient AI method for diagnosing attention-deficit/hyperactivity disorder (ADHD) in children using MRI scans. The novel approach significantly improves diagnostic accuracy with fewer trainable parameters.

Area of Science:

  • Neuroimaging
  • Artificial Intelligence
  • Medical Diagnostics

Background:

  • Early diagnosis of attention-deficit/hyperactivity disorder (ADHD) is vital for children's educational and mental health outcomes.
  • Diagnosing ADHD via neuroimaging is difficult due to symptom overlap and varied presentations.
  • Current methods lack efficiency and struggle with cross-modal data adaptation.

Purpose of the Study:

  • To develop a parameter-efficient transfer learning method for ADHD classification using MRI data.
  • To adapt a large-scale 3D convolutional foundation model, pre-trained on CT images, for MRI-based ADHD diagnosis.
  • To establish a new benchmark for ADHD classification efficiency and accuracy.

Main Methods:

  • Proposed a novel parameter-efficient transfer learning approach using Low-Rank Adaptation (LoRA) in 3D.
Keywords:
ADHDCross-modalFine-tuningFoundation ModelLow-Rank AdaptationMRI

Related Experiment Videos

  • Factorized 3D convolutional kernels into 2D low-rank updates to reduce trainable parameters.
  • Fine-tuned a foundation model pre-trained on CT images for an MRI-based ADHD classification task.

    • Main Results:

    • Achieved state-of-the-art results on a public diffusion MRI database using five-fold cross-validation.
    • One model variant reached 71.9% accuracy, another attained an AUC of 0.716.
    • Used only 1.64 million trainable parameters, over 113 times fewer than full fine-tuning.

    Conclusions:

    • Demonstrated successful cross-modal (CT-to-MRI) adaptation of a foundation model in neuroimaging.
    • Established a new benchmark for ADHD classification, significantly improving efficiency.
    • The 3D LoRA approach offers a highly efficient and accurate method for neuroimaging-based ADHD diagnosis.