echemi logo
Product
  • Product
  • Supplier
  • Inquiry
    Home > Active Ingredient News > Study of Nervous System > European Radiology: IVIM using convolutional neural networks to predict HCC microvascular invasion

    European Radiology: IVIM using convolutional neural networks to predict HCC microvascular invasion

    • Last Update: 2022-10-25
    • Source: Internet
    • Author: User
    Search more information of high quality chemicals, good prices and reliable suppliers, visit www.echemi.com

    Hepatocellular carcinoma (HCC) is one of the most common malignancies and is ranked as the fourth leading cause
    of cancer-related death worldwide.
    Microvascular invasion (MVI), i.
    e.
    , HCC emboli with micrometastases within hepatic vessels, is considered a key determinant of
    surgical resection and recurrence after liver transplantation.
    Therefore, an accurate preoperative assessment of MVI has important clinical implications
    for selecting appropriate surgical interventions.

    At this stage, medical images, including CT and MR images, have been extracted to predict the MVI of HCC based on morphological characteristics; However, qualitative analysis is limited
    by subjective interpretation differences.
    As an MRI to assess tissue dynamics, diffusion-weighted imaging (DWI) is a promising method
    for accurately predicting MVI.
    Its main advantages are that there is no need to use a contrast agent and no ionizing radiation
    is involved.
    Multib-valued DWIs with incoherent motion within voxels (IVIM) can distinguish true molecular diffusion and microcapillary perfusion using a biexponential model, which may determine tissue perfusion and diffusion more accurately than apparent diffusion
    coefficients (ADCs) calculated using a single-index model.
    At present, the accuracy of IVIM-DWI in clinical application is mainly based on the quantitative description
    of IVIM parameters.
    However, IVIM parameters derived from the biexponential model are very sensitive to motion and artifacts in DWI[, which greatly reduces the performance
    of IVIM-DWI in tumor characterization.

    Developments in machine learning and artificial intelligence have added a new range
    to CT and MR imaging analysis.
    A previous study used a deep convolutional network to clarify the relationship between multiple b-values in DWI, and deep learning feature representations learned from multiple b-value images were used to describe cancer lesions
    .
    In addition, a recently published study that uses a deep convolutional network to extract the depth features of multiple b-value images in DWI can better predict the MVI
    of HCC.
    To our knowledge, previous studies have proposed an attention-based deep learning model to process IVIM data to improve the performance of tumor characterization, but did not take into account any clinical information and statistical analysis
    .

    Recently, a study published in the journal European Radiology developed and evaluated a IVIM-DWI-based preoperative MVI deep learning model to achieve accurate prediction of HCC, better characterize lesions in clinical practice to achieve prognostic prediction
    .

    This retrospective study included 114 patients
    with pathologically confirmed HCC from December 2014 to August 2021.
    All patients underwent preoperative MRI of IVIM sequences with 9 b-values
    .
    First, 9 b-value images are superimposed on the channel dimension to obtain a b-value volume with a shape of 32×32×9 dimensions
    .
    Second, the image resampling method is used for data amplification to generate more samples for training
    .
    Finally, the depth features
    of predicting MVI in HCC are directly derived from the CNN-based b-value body.
    In addition, a deep learning model based on parameter graphs and a fusion model
    combining IVIM, clinical features, and deep features of IVIM parameters were constructed.
    Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance
    of MVI prediction in HCC.

    The depth features extracted directly from IVIM-DWI using CNNs (0.
    810 (range 0.
    760, 0.
    829)) yielded better performance
    for predicting MVI than IVIM parametric plots (0.
    590 (range 0.
    555, 0.
    643)).
    In addition, the fusion model combines the depth profile, clinical features (α-fetal protein (AFP) level and tumor size) and apparent diffusion coefficient (ADC) (0.
    829 (range 0.
    776, 0.
    848)) with slightly improved
    performance.


    The Figure
    class activates the heat map drawn by the attention map
    .
    Dark red indicates a larger contribution to MVI predictions, while dark blue indicates a smaller
    contribution.
    a A 26-year-old male with pathologically confirmed HCC but no MVI; b A 52-year-old man with pathologically confirmed HCC for MVI

    This study showed that the IVIM-based CNN model achieved high diagnostic performance
    in the prediction of MVI before HCC surgery.
    In addition, deep feature models that incorporate IVIM, clinical features (AFP level and tumor size), and ADC achieve better predictive performance
    than IVIM-based models alone.
    Therefore, the model proposed in this study can be used to predict MVI and assist clinicians in formulating precise treatment strategies
    .

     

    Original source:

    Baoer Liu,Qingyuan Zeng,Jianbin Huang,et al.
    IVIM using convolutional neural networks predicts microvascular invasion in HCC.
    DOI:10.
    1007/s00330-022-08927-9.

    This article is an English version of an article which is originally in the Chinese language on echemi.com and is provided for information purposes only. This website makes no representation or warranty of any kind, either expressed or implied, as to the accuracy, completeness ownership or reliability of the article or any translations thereof. If you have any concerns or complaints relating to the article, please send an email, providing a detailed description of the concern or complaint, to service@echemi.com. A staff member will contact you within 5 working days. Once verified, infringing content will be removed immediately.

    Contact Us

    The source of this page with content of products and services is from Internet, which doesn't represent ECHEMI's opinion. If you have any queries, please write to service@echemi.com. It will be replied within 5 days.

    Moreover, if you find any instances of plagiarism from the page, please send email to service@echemi.com with relevant evidence.