Dynamic analysis of MR-PET data on brain tumors

Ana Morgado, Liliana Caldeira, Nuno Silva, Christian Filß, Nuno Matela, Karl‐Josef Langen, N. Jon ShahEJNMMI 2014 · DOI

TL;DR

This paper explores the potential of combining dynamic MRI and PET scans for brain tumor analysis. By extracting parametric images from both modalities, the researchers found that dynamic MR-PET offers complementary information, suggesting improved biological insights and diagnostic capabilities for brain tumors.

What Problem Does This Paper Address?

Hybrid MR-PET scanners offer a powerful way to integrate anatomical and functional imaging. For brain tumors, dynamic susceptibility contrast (DSC)-MRI is used to assess blood perfusion, while dynamic [18F]-fluoro-ethyl-tyrosine (FET)-PET provides functional information about tumor metabolism. However, static analysis of these data might miss crucial dynamic changes. This paper addresses the challenge of effectively analyzing and comparing dynamic MR-PET data to extract meaningful parametric images that can better characterize brain tumors, moving beyond traditional static assessments.

Methods

The study utilizes data from a 3T-MR-BrainPET scanner. Dynamic FET-PET scans were acquired for 60 minutes post-injection and reconstructed using OSEM3D. Simultaneously acquired MRI sequences included MPRAGE (for anatomical detail) and DSC-MRI EPI (for perfusion). The core of the methodology involves co-registering the MR and PET data and then extracting several key parameters from both using Matlab: peak signal intensity, time to peak (TTP), area under the curve (AUC), and wash-in rates. These extracted parameters are then visualized as parametric images, allowing for a direct comparison of the dynamic information derived from each modality.

Datasets

The study analyzed data from three patients with brain tumors. The specific datasets consist of:

  • Simultaneous MPRAGE MRI, DSC-MRI EPI, and dynamic FET-PET scans.
  • Extracted parametric images (peak, AUC) from both MRI and PET data for each patient.

The paper does not introduce a new benchmark dataset but rather presents results from a small cohort of real patient data.

Key Insights and Ideas

  • Complementary Information: Dynamic MR-PET provides distinct yet complementary information. While anatomical MRI shows structural changes, DSC-MRI highlights perfusion dynamics, and FET-PET reveals metabolic activity. Combining these dynamic analyses offers a richer picture than any single modality alone.
  • Parametric Imaging Power: The extraction of parametric images (peak, TTP, AUC, wash-in) from dynamic data is crucial. These images can highlight specific biological processes and spatial distributions that might be obscured in summed or static images.
  • Spatial-Temporal Agreement: The study observed good spatial and temporal agreement between the parametric images derived from dynamic MR and PET data. This suggests that the underlying biological processes being measured are well-correlated across modalities.
  • Potential for Deeper Biological Understanding: The authors propose that dynamic MR-PET analysis can offer deeper insights into tumor biology, potentially aiding in diagnosis, treatment planning, and monitoring.

Key Results

The paper presents qualitative results from three analyzed datasets. Figure 1 shows the MPRAGE, DSC-MRI EPI, and FET-PET images alongside their corresponding extracted parametric images (peak and AUC). The FET-PET parametric images clearly identified uptake areas within regions of morphological change seen in the anatomical MR. Figure 2 displays normalized contrast time curves for both MRI and PET, demonstrating the temporal behavior of contrast uptake and washout. The authors conclude that the extracted parametric images from dynamic MR-PET showed good spatial and temporal agreement, supporting their hypothesis.

Limitations and Future Work

Limitations: * The study analyzed data from only three patients, which is a small sample size. * The FET-PET data was truncated for the time curve comparison in Figure 2, which might affect the full dynamic profile.

Future Work: * The authors suggest that dynamic MR-PET may hold significant extra information about tumor biology. * They propose that the combination of simultaneous acquisition and analysis is beneficial and warrants further investigation. * Further studies with larger patient cohorts are needed to validate these findings and explore the full clinical utility of dynamic MR-PET for brain tumor characterization.

Summary updated

Abstract

The introduction of hybrid MR-PET scanners offers new perspectives to better correlate MR and PET data with respect to time and space domain. In case of brain tumor patients, dynamic susceptibility contrast (DSC)-MRI is often used to measure perfusion levels of brain, while dynamic [18F]-fluoro-ethyl-tyrosine (FET)-PET provides additional functional information. The dynamic analysis of data is becoming more relevant, rather than static analysis, and with this the extraction of parametric images. In this context, the aim of this work is to compare dynamic MR-PET data based on different features. A FET-PET scan was carried out on the 3T-MR-BrainPET [1] on each patient. Data was acquired for 60 min after injection in list-mode and reconstructed using OSEM3D software (4 subsets, 32 iterations). All MRI sequences were acquired simultaneous to the PET measurement. In this work, MPRAGE and DSC-MRI based on an EPI sequence images are used. For the dynamic analysis, data was co-registered and the same parameters were extracted, with Matlab, from MR-PET data to produce peak, time to peak (TTP), area under the curve (AUC) and wash-in parametric images. Results from 3 of the analyzed datasets from patients with brain tumors are presented on Figure ​Figure1.1. MPRAGE, DSC-MRI EPI sequence and FET-PET images and the extracted parametric images are shown. Figure 1 Analyzed datasets (i, ii and iii). MRI images (a) MPRAGE (post contrast), (b) DSC-MRI EPI, and the corresponding extracted parametric images: (c) peak and (d) AUC. (e) FET-PET image (summed image of 20-40 min p.i., previously filtered with a 2 mm Gaussian ... In FET-PET parametric images an uptake area can be identified in the region of morphological changes in anatomical MR. The parametric images provide extra information in addition to the summed FET-PET images. The time curves obtained from a gray-matter region defined on MPRAGE are presented on Figure ​Figure22. Figure 2 MRI and PET contrast time curves. The FET-PET data was truncated (the total acquisition time was 3600 s) in order to merge the two curves in one graph. The MRI and PET contrast time curves were normalized to the maximum to be able to compare the shape ... The extracted parametric images from dynamic MR-PET showed good spatial and temporal agreement (Figure ​(Figure1).1). These results suggest that dynamic MR-PET may have extra information about biology and that the combination of simultaneous acquisition and analysis may be beneficial.
BibTeX
@inproceedings{morgado2014dynamic,
  author = {Ana Morgado and Liliana Caldeira and Nuno Silva and Christian Filß and Nuno Matela and Karl‐Josef Langen and N. Jon Shah},
  title = {{Dynamic analysis of MR-PET data on brain tumors}},
  booktitle = {EJNMMI},
  year = {2014},
  doi = {10.1186/2197-7364-1-s1-a56},
}
Figure