MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Cerebrovascular Reactivity Mapping Using Resting-State Functional MRI in Patient with Gliomas

Cerebrovascular Reactivity Mapping Using Resting-State Functional MRI in Patient with Gliomas

M Yeh12*, E Gates1, P Hou1, V Kumar3, J Johnson3, K Noll4, S Prabhu5, S Ferguson5, G Rao5, D Schomer3, H Liu1, 1UT MD Anderson Cancer Center, Houston, TX, 2Department of Biomedical Engineering and Environmental Sciences, Hsinchu National Tsing Hua University, 3Departments of Diagnostic Radiology, 4Department of Neuro-oncology, 5Department of Neurosurgery, The University of Texas MD Anderson Cancer Center, Houston, TX

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose:
Cerebrovascular reactivity (CVR) is a key indicator of neurovascular uncoupling in presurgical functional MRI (fMRI). This study aimed to investigate the frequency range selection of resting-state (rs) fMRI signal fluctuations for the CVR mapping in gliomas patients, and to compare the results with breath-hold (BH) MRI.

Methods:
Twenty-four patients with gliomas underwent the MRI study on a 3T scanner. MRI scans included pre- and post-contrast T1-weighted, FLAIR, rs-fMRI (180 dynamics, scan time = 6 mins) and BH MRI (3 cycles of 15-s BH and 45-s rest). The rs-fMRI data were used to estimate CVR maps with rs fluctuation of amplitude (RSFA, > 0.009 Hz) and amplitude of low frequency fluctuation (ALFF, 0.01-0.08 Hz). In addition, an RSFA map was generated after an automatic denoise process (RSFA_fix). The denoise procedure was conducted by using ICA and automatic classification in FIX (v1.06.12, Oxford, UK). CVR mapping with BH-MRI was performed by general linear model based on a respiratory response function. Tumor ROIs were determined by segmentation using a deep learning model and then adjusted by a neuroradiologist. Dice index was calculated to evaluate the similarity between each of the rs-fMRI CVR maps and the CVR map obtained with BH-MRI.

Results:
A subset of our patient data was used to train a denoise model which led to sensitivity of 82.5% and specificity of 81.2%. The CVR from ALFF had significant higher dice than from the RSFA maps in the normal tissues (0.74?0.10 vs. 0.77?0.09, p<0.001). However, the opposite results, i.e. significantly higher dice with RSFA than ALFF (0.55?0.16 vs. 0.49?0.17, p<0.001) were obtained in the lesions.

Conclusion:
The ALFF performed well for CVR mapping in normal tissues, whereas a wider frequency range needed to be incorporated in diseased tissues. The overall lower dice in lesions than in normal tissues requires further studies.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email

Additional Links