Precise knowledge of the functional brain areas in the individual subject is crucial for the accurate quantitative analysis of brain PET images and their meaningful interpretation. Often however, functional regions are manually outlined in a casual manner, with subjective results as the outcome. This fundamental flaw can be overcome with PMOD’s PNEURO tool which offers both objective region outlining as well as pattern analysis methods via easy step-by-step procedures. These methods are provided as three modules on separate pages of the user interface.
Automatic Brain Regions by Human Maximum Probability Atlas and other Atlases
A well established way of obtaining brain volumes of interest (VOIs) automatically is by leveraging the most likely localization of brain areas as encoded in the maximum probability atlas N30R83 constructed by Hammers et al. [1]. The corresponding PNEURO component allows the user to adjust the atlas to the subject brain anatomy with a spatial normalization procedure, preferably obtained from the T1-MR image, and the resulting VOI set that can be masked by the individual subject grey matter. Alternatively, the normalization can be directly derived from the PET image, making this solution also applicable for PET-only studies.
The spatial normalization and segmentation of human brain data with pronounced atrophy is a well known challenge. Normalization according to tissue probability and subsequent masking of atlas segments according to grey matter probability is widely successful in cortical regions, but can be insufficient for deep nuclei around enlarged ventricular spaces. The ANTS methodology can help to deal with moderate cases, but processing time can be substantial. Novel AI-based approaches yield an interesting solution for analysis in the subject space. A new hybrid-AI solution has been added to our popular PNEURO tool. This solution allows the user to replace the deep nuclei segments with the result of prediction with a trained neural network.
Automatic Human Brain Regions by T1-MRI Parcellation
It is desirable that every brain PET study be complemented with a T1-weighted MRI which precisely represents the subject’s brain anatomy. The PNEURO tool includes a sophisticated component which applies knowledge-base technology to accurately segment the cortex and the basal ganglia from the T1-MRIs. These segments are then converted to VOIs which can be projected to the PET images and used for regional statistics.
Parametric Mapping
If the pixel-wise modeling tool PXMOD has been licensed, PNEURO supports a Parametric Mapping workflow after VOI outlining has been completed. The integration has several advantages: (1) The VOIs generated by PNEURO can be used during the PXMOD model configuration. (2) The resulting parametric maps can immediately be regionally analyzed. (3) The parametric maps can be calculated in an image space or the atlas space.
Normal Brain PET Databases
PET studies with diseased subjects can rarely be done in a fully quantitative manner. Rather, a static image of the PET tracer concentration is acquired after an appropriate equilibration time. In this situation, an image analysis is based on the fact that with consistent study protocols, the normal brain uptake exhibits a characteristic uptake pattern. The database component in the PNEURO tool readily allows for pooling the uptake across a set of normal volunteers, and thus establishing the normal pattern together with its variability. Once such a normal database has been created, the anomalies of the tracer uptake in a subject’s brain can be easily localized without any prior assumptions or operator variability and presented as a z-score map.
System Requirements
For productively working with the parcellation tool, the following workstation system requirements should be met:
▪64-Bit operating system (Windows, Mac OS X, Linux)
▪≥ 32 GB RAM
▪≥ 8 processing cores (hyper-threading is also viable)
16 GB RAM and 4 cores is at the edge of practicability.
Starting the PNEURO Tool
The PNEURO tool is started with the Neuro button:
or by directly dragging image files onto the above button.
Organization of the PNEURO User Interface
The user interface of PNEURO consists of five pages which can be selected by tabs:
1.Maximum Probability Atlas page: Creation of brain VOIs by adjusting the N30R83 maximum probability atlas or other atlases to the subject anatomy.
2.Brain Parcellation page: Creation of brain VOIs by knowledge-based segmentation of anatomical T1-MR images.
3.Mapping page: When the PXMOD Parametric Mapping Tool has been additionally licensed, parametric mapping may be directly accessed for dynamic PET data processed in the Maximum Probability Atlas or Brain Parcellation workflows.
4.Brain Norm page: Comparison of a static PET image to a Brain Norm. Brain Norms can be created in PNEURO from a set of images acquired with normal controls using the Edit Norm tool from the menu.
5.Results page: The VOIs created by the two methods above can directly be applied to the (dynamic) PET images for calculating the regional uptake, optionally with a partial volume correction. The resulting statistics or curves are shown on the Results page and from there can be saved or submitted to the kinetic modeling tool.
The user selects the appropriate module for data processing by the corresponding tab in the user interface of PNEURO. Each of the pages is described in a separate section of this guide.