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

Piriform cortex gates learned fear through a brain-spleen neuroimmune axis.

Neuron·2026
Same author

Orchestrating the metastatic symphony: the role of extracellular vesicles in the epithelial-mesenchymal transition and pulmonary niche formation of breast cancer.

Biological reviews of the Cambridge Philosophical Society·2026
Same author

Design strategies for hydrogels with paracrine effects and their ability to promote wound healing: status and development.

International journal of pharmaceutics·2026
Same author

Efficacy of BMP-2 in rotator cuff repair: a randomized controlled trial based on radiological and clinical outcomes.

BMC musculoskeletal disorders·2026
Same author

A preliminary study to evaluate efficacy and safety of Lugol's solution following radioiodine for remnant ablation in differentiated thyroid cancer.

Frontiers in oncology·2026
Same author

Development and Preliminary Evaluations of a Novel <sup>18</sup>F-Labeled Tracer for Detection of BRAF<sup>V600E</sup> Mutant Status in Animal Models of Melanoma.

Molecular pharmaceutics·2026

Related Experiment Video

Updated: Apr 10, 2026

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

3.7K

A PET/CT Cross-Modal Wavelet Fusion and Pseudo-Mask Guided Network With Frozen SAM Decoder for Multiple Myeloma

Jingxin Han, Hong Chen, Chengfan Li

    IEEE Transactions on Bio-Medical Engineering
    |April 8, 2026
    PubMed
    Summary

    This study introduces a novel framework for segmenting Multiple Myeloma lesions in PET/CT scans, achieving high accuracy with significantly fewer parameters. The efficient model facilitates clinical deployment for improved patient prognosis.

    More Related Videos

    Deep Learning-Based Segmentation of Cryo-Electron Tomograms
    10:25

    Deep Learning-Based Segmentation of Cryo-Electron Tomograms

    Published on: November 11, 2022

    11.3K
    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
    04:48

    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

    Published on: July 5, 2024

    878

    Related Experiment Videos

    Last Updated: Apr 10, 2026

    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

    3.7K
    Deep Learning-Based Segmentation of Cryo-Electron Tomograms
    10:25

    Deep Learning-Based Segmentation of Cryo-Electron Tomograms

    Published on: November 11, 2022

    11.3K
    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
    04:48

    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

    Published on: July 5, 2024

    878

    Area of Science:

    • Medical Imaging
    • Artificial Intelligence
    • Radiology

    Background:

    • Accurate segmentation of Multiple Myeloma (MM) lesions in PET/CT scans is crucial for patient prognosis.
    • Challenges include lesion heterogeneity and cross-modal frequency discrepancies between PET and CT.
    • Existing methods often struggle with parameter efficiency and robustness.

    Purpose of the Study:

    • To develop a robust and parameter-efficient framework for accurate MM lesion segmentation from PET/CT.
    • To address the complexities of cross-modal frequency discrepancies and lesion heterogeneity.
    • To create a computationally efficient model suitable for clinical workflows.

    Main Methods:

    • Introduced a Cross-Modal Dual-Wavelet Fusion Network utilizing a Frozen SAM Decoder.
    • Replaced the standard SAM image encoder with a custom dual-branch wavelet encoder for explicit cross-modal feature alignment.
    • Employed a "decompose-and-inject" mechanism to fuse high-frequency CT edges with low-frequency PET metabolic cues.
    • Leveraged a pre-trained SAM mask decoder for geometric priors and a Tiny Pseudo Decoder for boundary supervision.

    Main Results:

    • Achieved Dice scores of 0.8323 on an in-house MM dataset and 0.8465 on the HECKTOR 2022 dataset.
    • Significantly outperformed state-of-the-art baselines (nnU-Net, MedSAM) with p < 0.05.
    • Model parameter count is 10.28M, a ~96% reduction compared to standard SAM (271.24M), while maintaining superior accuracy.

    Conclusions:

    • The proposed framework successfully bridges the gap between lightweight deployment and high-performance cross-modal segmentation.
    • Offers a practical, efficient, and accurate alternative for PET/CT segmentation in clinical settings.
    • Demonstrates strong accuracy, cross-dataset generalizability, and improved computational efficiency.