Using Imaris to rigorously track PET-defined sites of lung inflammation in Mycobacterium tuberculosis-exposed non-human primates

Abstract

Aerosol exposure of non-human primates (NHPs) to Mycobacterium tuberculosis (Mtb) typically results in discrete sites of inflammation of the lung that is detectable by 2-deoxy-2-[fluorine-18]fluoro-D-glucose (¹⁸F-FDG)-based PET/CT scans. Such scans are often analyzed using software such as Invicro VivoQuant or OsiriX as 3D images by manual labeling sites of PET signal using 2D slices and by reporting maximal SUV either of the whole lung or of individual lesions. Here we propose a pipeline for analysis of the same PET/CT scans using Imaris, a proprietary software typically used for analysis of data from fluorescent microscopy experiments. We show that by using locations of spine vertebra (denoted as "landmarks") we can align serials scans of the same animal, and by using automated (with some manual corrections) image segmentation in 3D as "surfaces", we can accurately define location of all sites of inflammation in the lung and lung-associated thoracic lymph nodes (LNs). We show that there is an excellent correlation between individual lesion's maximum SUV determined by Invicro VivoQuant and maximum intensity determined by Imaris suggesting utility of this approach. Imaris also provides wealth of additional information for each of the identified lesions such as volume, location, shape, surface area, and others, and each lesion can be exported in Virtual Reality file format (.wrl) allowing for detailed and rigorous analyses of how features of these PET-defined lesions evolve over time and correlate with the outcome of infection and/or treatment.

Keywords: Imaging data; Imaris; PET tracers; PET/CT; SUV; radioactivity.

Figure 1:. The pipeline of processing PET/CT imaging data with Invicro VivoQuant and Imaris.

We illustrate how PET/CT scans are typically analyzed with Invicro VivoQuant, the standard proprietary packages for PET/CT data analysis (A), and with Imaris, proprietary packages typically used for fluorescent microscopy data analysis (B). A: With Invicro VivoQuant (i-iv), the operator analyzes reconstructed 3D images by using 2D slices and manually labeling contours of the lesions, defined by PET signal. Invicro VivoQuant then provides various SUV characteristics, such as max SUV, for ether individual lesions or the whole lung. B: With Imaris (i-vi), PET/CT DICOM files are imported as 2-channel images, and scans done at different times can be incorporated into the same Imaris file by adding new time frames. Then scans for different time points are aligned using “Correct for drift” routine by using landmarks (6 locations of spine vertebra). Then region of interest (ROI) is then identified around each suspected lesion and the data in ROI is processed using 3D routine “Surfaces”. All lesions then can be grouped into the specific group depending on the location of the lesion (e.g., lung vs. lymph nodes).

Figure 2:. Lesions identified with the standard PET/CT software (Invicro VivoQuant) can be also traced using semi-automated tools in Imaris.

Lesions identified for animal 40884 using Invicro VivoQuant (Ai and B) or Imaris (Aii and C) at different time points and different views. A: 3D view of lesions at 18 weeks post-infection in Invicro VivoQuant (i) and the same lesions in Imaris (ii). B-C: Slice views of lesions identified with Invicro VivoQuant (B) or Imaris (C) at 6 (i, iv), 14 (ii, v), and 18 (iii, vi) weeks post-infection in both sagittal (i-iii) and coronal (iv-vi) slice views (see also Supplemental Figures S1 and S2). PET signal associated with the heart (shown by an arrow in Ai-ii) was excluded from our analyses.

Figure 3:. Normalized maximum PET channel intensity, determined by Imaris, matches well standard max SUV provided by Invicro VivoQuant.

We analyzed PET/CT data from four Mtb-infected RMs (each with three PET/CT scans) either with Invicro VivoQuant, a standard software for PET/CT data analysis, or Imaris (Bitplane), software typically used for analysis of fluorescent microscopy data. For each lesion we plot maximum SUV as provided by Invicro VivoQuant (x axes) or maximum intensity as provided by Imaris (y axes), normalized using weight of the animal and 18F-FDG injection dose and multiplied by $f={10}^4$ (see eqn. (1)). We show data for animal 41883 (A), 44104 (B), 41634 (C), and 40884 (D) with $n$ indicating the number of lesions detected in all scans and $m$ being a regression slope between SUV values of Invicro VivoQuant vs. Imaris for all lesions or lesions found in individual scans; we show the regression lines in individual panels for all lesions. Scans done at different times are shown by different color intensity. The dashed line shows the slope of 1.

Figure 4:. Tracking volume, area, and sphericity of PET lesions identified by Imaris.

For every lesion identified for animals 41883 (A), 44104 (B), 41634 (C), and 40884 (D) we show lesion’s total volume (i), surface area (ii), and sphericity (iii). Markers denote individual lesions and lines connect average values per time point; color of the markers denotes whether the animal was classified as being asymptomatic (gray) or having TB (dark red) at specific scan times.

Figure 5:. 3D rendering of individual PET-defined lesions in Mtb-infected monkeys using metrics provided by Imaris: rear view.

By using XYZ coordinates, provided by Imaris we plotted individual lesions detected in animals 41883 (A), 44104 (B), 41634 (C), and 40884 (D) at different time points after infection (times are denoted on individual panels). Note that the plots are from rear view of the animals (see also Supplemental Figure S4). We also show the total number of lesions detected at each time point as $n$. Lesions are plotted as spheres with the radius $R=\sqrt[3]{3V/(4\pi )}$ where $V$ is the PET volume determined by Imaris (Figure 4). Intensity of the color denotes scaled maximum PET signal intensity normalized to $I_{\text{max}}$ values for the largest maximum intensity for all lesions in all scans of a given animal; this also takes into account the actual injected FDG dose and weight of the animal (Figure 3). Black colors indicate animals that were asymptomatic and red colors indicate that the animal was diagnosed with active disease (TB).