Image derived input function applied in CBF Studies with [15O]water PET in an integrated MR-PET

Jörg Mauler, Nuno Silva, Christian Filß, Heinz H. Coenen, Hans Herzog, N. Jon ShahEJNMMI 2014 · DOI

TL;DR

This paper explores a non-invasive method for measuring cerebral blood flow (CBF) using PET scans. By deriving the "input function" (a measure of tracer concentration over time) directly from PET images of the internal carotid artery, researchers can potentially replace the traditional, invasive arterial blood sampling method. This image-derived input function (IDIF), after accounting for image artifacts, shows promise for accurate CBF quantification in integrated MR-PET systems.

What Problem Does This Paper Address?

Quantifying cerebral blood flow (CBF) using PET, particularly with tracers like [15O]water, requires knowing the concentration of the tracer in the blood over time. This is known as the "input function." The standard method for obtaining this is through arterial blood sampling, which is invasive and can be technically challenging. Researchers are seeking non-invasive alternatives that can provide accurate input functions, making CBF studies more accessible and less burdensome for patients. This paper investigates whether an input function can be derived directly from the PET images themselves, specifically by targeting the internal carotid artery (ICA), which supplies blood to the brain.

Methods

The study utilized a 3T MR-PET scanner to acquire data from five healthy volunteers. Each volunteer received four bolus injections of [15O]water. Simultaneously, arterial blood samples were drawn from the radial artery to measure the conventional input function (TACblood).

Key methodological steps included:

  • PET Data Acquisition and Reconstruction: Dynamic PET scans were acquired for 3 minutes post-injection and reconstructed using an iterative OSEM3D algorithm.
  • MR Image Acquisition: A high-resolution T1-weighted MPRAGE MRI scan was acquired for anatomical reference.
  • Volume of Interest (VOI) Definition: VOIs were defined in the internal carotid arteries (ICAs) on the high-resolution MRI images.
  • Image-Derived Input Function (IDIF) Generation: These ICA VOIs were then transferred to the dynamic PET images to generate an IDIF (TACcar).
  • Input Function Correction: The conventional arterial blood input function (TACblood) was corrected for dispersion and delay using a one-compartment model and the whole-brain TAC.
  • CBF Quantification: Cerebral blood flow (CBF) was calculated using the standard autoradiographic [15O]water model, employing both the corrected arterial input function (TACbloodcorr) and the image-derived input function (TACcar) as inputs.
  • Artifact Correction: The IDIF (TACcar) was further corrected for partial volume effects (PVE) and spillover (SP) from surrounding tissues.

Datasets

The study used data from five healthy young volunteers. No new benchmarks were introduced; the focus was on evaluating a novel method against a standard one using existing data.

Key Insights and Ideas

  • Non-invasive Input Function is Feasible: The core idea is that by carefully selecting a region of interest (ROI) in the internal carotid artery on the PET images, one can extract a time-activity curve that closely mimics the arterial blood input function.
  • MR-PET Synergy is Crucial: The integrated MR-PET system is highlighted as a key enabler. The high-resolution MR images provide excellent anatomical detail for accurately placing the ICA ROIs, which are then applied to the PET data.
  • Image Artifacts are Significant: The derived input function (TACcar) is inherently affected by partial volume effects (PVE) and spillover (SP) from adjacent brain regions. These artifacts can distort the measured tracer concentration.
  • Correction is Essential for Accuracy: Crucially, the paper demonstrates that correcting the IDIF for PVE and SP is necessary to achieve CBF values comparable to those obtained with the invasive arterial blood sampling method.

Key Results

  • The PVE factor for the ICA VOIs was found to be 0.21 ± 0.05.
  • The spillover factor from the temporal lobe was 0.24 ± 0.10.
  • When the corrected IDIF (TACcar after PVE and SP correction) was used, the global CBF values differed by only 8.4% ± 9.4% compared to the results obtained using the corrected arterial blood input function (TACbloodcorr).

Limitations and Future Work

  • Limited Sample Size: The study was conducted on only five healthy volunteers, which is a small sample size.
  • Healthy Volunteers Only: The findings are based on healthy individuals, and it's unclear how well this method would perform in patients with neurological conditions that might affect blood flow or vascular anatomy.
  • Specific Tracer: The study focused on [15O]water. Further research is needed to validate this IDIF approach for other PET tracers used in CBF quantification.
  • Future Work: The authors suggest that after appropriate correction for PVE and SP, the image-derived input function may indeed replace invasive arterial blood sampling for [15O]water PET studies, paving the way for more routine and less invasive CBF measurements.

Summary updated

Abstract

To quantify CBF studies using [15O]water PET an input function must be provided, e.g. by means of arterial blood sampling. An image derived input function (IDIF) gained from the dynamic PET images may be a non-invasive alternative. The time-activity curve (TAC) of the blood can be obtained from a volume of interest (VOI) capturing the internal carotid artery (ICA). Five healthy young volunteers were administered 555 MBq [15O]water four times each, by bolus intravenous injection. A 3 min PET listmode acquisition was started at the time of the injection (3TMR-BrainPET). Blood was automatically withdrawn from the radial artery to measure its TAC (TACblood, coincidence detector). A T1-weighted MPRAGE image (192x256x256 voxels, 1 mm3) was acquired. Applying corrections, the PET data were iteratively (OSEM3D) reconstructed (frame length 4s, 153x256x256 voxels, 1.25x1.25x1.25 mm3) and filtered (4mm 3D Gaussian). VOIs defining the ICAs in the MPRAGE image were transferred to the dynamic PET images to obtain an IDIF (TACcar). The dispersion and delay of TACblood were corrected (TACbloodcorr) using a one-compartment model and the whole-brain TAC. Cerebral blood flow (CBF) was quantified applying the autoradiographic 15O-water model using both TACbloodcorr and TACcar as input data. Compared to TACbloodcorr, TACcar is affected by the partial volume effect (PVE) and the spillover (SP) from the temporal lobe. The PVE factor and the SP factor were 0.21±0.05 and 0.24±0.10, respectively. Global CBF differed by 8.4% ± 9.4%, when the TACcar after correction for PVE and SP was used as input data instead of TACbloodcorr. The inherently coregistered, high-resolution 3TMR-BrainPET data are a suitable prerequisite for measuring the IDIF. After correction for PVE and SP, the IDIF may replace invasive measurement of 15O-activity.
BibTeX
@inproceedings{mauler2014image,
  author = {Jörg Mauler and Nuno Silva and Christian Filß and Heinz H. Coenen and Hans Herzog and N. Jon Shah},
  title = {{Image derived input function applied in CBF Studies with [15O]water PET in an integrated MR-PET}},
  booktitle = {EJNMMI},
  year = {2014},
  doi = {10.1186/2197-7364-1-s1-a30},
}
Figure