Masked volume wise Principal Component Analysis of small adrenocortical tumours in dynamic [11C]-metomidate Positron Emission Tomography

Abstract

Background: In previous clinical Positron Emission Tomography (PET) studies novel approaches for application of Principal Component Analysis (PCA) on dynamic PET images such as Masked Volume Wise PCA (MVW-PCA) have been introduced. MVW-PCA was shown to be a feasible multivariate analysis technique, which, without modeling assumptions, could extract and separate organs and tissues with different kinetic behaviors into different principal components (MVW-PCs) and improve the image quality.

Methods: In this study, MVW-PCA was applied to 14 dynamic 11C-metomidate-PET (MTO-PET) examinations of 7 patients with small adrenocortical tumours. MTO-PET was performed before and 3 days after starting per oral cortisone treatment. The whole dataset, reconstructed by filtered back projection (FBP) 0-45 minutes after the tracer injection, was used to study the tracer pharmacokinetics.

Results: Early, intermediate and late pharmacokinetic phases could be isolated in this manner. The MVW-PC1 images correlated well to the conventionally summed image data (15-45 minutes) but the image noise in the former was considerably lower. PET measurements performed by defining "hot spot" regions of interest (ROIs) comprising 4 contiguous pixels with the highest radioactivity concentration showed a trend towards higher SUVs when the ROIs were outlined in the MVW-PC1 component than in the summed images. Time activity curves derived from "50% cut-off" ROIs based on an isocontour function whereby the pixels with SUVs between 50 to 100% of the highest radioactivity concentration were delineated, showed a significant decrease of the SUVs in normal adrenal glands and in adrenocortical adenomas after cortisone treatment.

Conclusion: In addition to the clear decrease in image noise and the improved contrast between different structures with MVW-PCA, the results indicate that the definition of ROIs may be more accurate and precise in MVW-PC1 images than in conventional summed images. This might improve the precision of PET measurements, for instance in therapy monitoring as well as for delineation of the tumour in radiation therapy planning.

Figure 1

First Column shows the summed images generated by summing images through frame 9–14 (15–45 min) and second column shows the corresponding MVW-PC1 images generated by application of MVW-PCA on the whole imaging sequence of 0–45 minutes. These are shown in the transaxial (first row) sagittal (second row), and coronal planes (third row).

Figure 2

Visual comparison between the summed image (2a), corresponding MVW-PC1 image (2b), arbitrary chosen image of MVW-PC2 (2c) and MVW-PC3 (2d). Notably, the MVW-PC2 and MVW-PC3 images represent tissues with two different and uncorrelated tracer pharmacokinetics compared to MVW-PC1 images. In MVW-PC2, the early pharmacokinetic events are displayed showing the parenchymal tracer accumulation in the kidneys. In MVW-PC3, the pharmacokinetic events are displayed and clearly demonstrate the [¹¹C]-MTO characteristic accumulation of radioactivity in the gastric juice.

Figure 3

Bar chart showing the SUV of a normal adrenal gland (upper chart) and an adrenocortical adenoma (lower chart) for each patient as obtained by using the "hot spot" method defined in both the summed and MVW-PC1 image in the MTO-PET examination in the untreated and treated state.

Figure 4

Regression analysis of the SUVs obtained using the "hot spot" method defined in the MVW-PC1 images when plotted against the SUVs obtained using ROIs defined in the summed images shown for the adrenocortical adenomas (dots) and the normal adrenal glands (triangles).

Figure 5

Mean TACs of all patients (error bars have been omitted for clarity), demonstrating the pharmacokinetic behavior of [¹¹C]-MTO in the adrenocortical adenomas and normal adrenal glands before and after cortisone treatment, represented by defined ROIs using the 50% cut-off method in the MVW-PCA images and summed images.

Figure 6

Bar chart shows the mean SUVs of 7 patients calculated in ROIs representing the adrenocortical adenomas and normal adrenal glands. The means are defined in both the summed and MVW-PC1 images, using the 50% cut-off method before and after cortisone treatment.