Umeå University's logo

BACKGROUND: The delineation of intraprostatic lesions is vital for correct delivery of focal radiotherapy boost in patients with prostate cancer (PC). Errors in the delineation could translate into reduced tumour control and potentially increase the side effects. The purpose of this study is to compare PET-based delineation methods with histopathology.

MATERIALS AND METHODS: The study population consisted of 15 patients with confirmed high-risk PC intended for prostatectomy. [68Ga]-PSMA-PET/MR was performed prior to surgery. Prostate lesions identified in histopathology were transferred to the in vivo [68Ga]-PSMA-PET/MR coordinate system. Four radiation oncologists manually delineated intraprostatic lesions based on PET data. Various semi-automatic segmentation methods were employed, including absolute and relative thresholds, adaptive threshold, and multi-level Otsu threshold.

RESULTS: The gross tumour volumes (GTVs) delineated by the oncologists showed a moderate level of interobserver agreement with Dice similarity coefficient (DSC) of 0.68. In comparison with histopathology, manual delineations exhibited the highest median DSC and the lowest false discovery rate (FDR) among all approaches. Among semi-automatic approaches, GTVs generated using standardized uptake value (SUV) thresholds above 4 (SUV > 4) demonstrated the highest median DSC (0.41), with 0.51 median lesion coverage ratio, FDR of 0.66 and the 95th percentile of the Hausdorff distance (HD95%) of 8.22 mm.

INTERPRETATION: Manual delineations showed a moderate level of interobserver agreement. Compared to histopathology, manual delineations and SUV > 4 exhibited the highest DSC and the lowest HD95% values. The methods that resulted in a high lesion coverage were associated with a large overestimation of the size of the lesions.

Positron Emission Tomography (PET) is an important tool for detection, staging and follow-up in a wide range of diseases, including cancer and neurological disorders. As a functional imaging tool, PET can visualize biological processes, where positron emitting radioactive isotopes are connected to molecules with different functions in the body. While PET-images can be visually interpreted, they can also be used for quantitative measurements, where functions such as glucose metabolism, dopamine receptor function, and blood-flow can be quantified. Measurements can be performed in static imaging, or in dynamic imaging where graphical methods can be used for analysis.

PET images benefit from fusion with anatomical images which facilitates the interpretation. The combination of PET with computed tomography (CT) as in PET/CT hybrid equipment is a well-established imaging method. Magnetic Resonance Imaging (MRI) has some advantages over CT such as the high soft tissue contrast, but the combination with PET in a fully integrated system is far more technically challenging. Most of the technical concerns have been solved, and PET/MRI modalities are now commercially available.

Among the remaining challenges, the attenuation correction is still not yet completely solved, where the attenuation maps on the PET/MRI modalities are approximate and bone is not accounted for in all parts of the body. There are also challenges with quantitative PET in general, where for example low spatial resolution and presence of noise can lead to quantitative errors. The purpose of this thesis was to investigate and develop strategies to reduce quantitative errors in PET imaging with special focus on PET/MRI.

In study I, we studied the limits for quantification of size and uptake in small lesions in PET images reconstructed with a resolution modelling algorithm. We constructed a phantom of small balloons and reconstructed images with three different algorithms and measured volume and activity concentration in the images. The measured activity concentration in the lesions was corrected for the low resolution that yields partial-volume effects (PVE). We found that resolution modelling improved quantification of all lesions, and that in combination with correction factors, lesions larger than ~9 mm diameter could be correctly quantified.

Study II is focused on the effect of frame time length on the graphical Logan-analysis for dynamic studies with 11C-raclopride. Logan analysis is reported to be sensitive to noise, and image noise is heavily dependent on the frame time length. Noise can also generate bias when using iterative reconstruction methods. Weivconcluded that with region-based analyses, a bias of approximately 10% in the non-displaceable binding potential was found when using the shortest time frames, and that the bias was mainly caused by the reconstruction algorithm. Long time frames generated stable parameters.

The last two studies focused on the attenuation correction in PET/MRI hybrid equipment. In study III, a method for attenuation correction in PET/MRI was implemented and evaluated. The method is developed for the pelvic region and is based on statistical decomposition of T2-weighted images. We found that the new method improved quantification, especially in regions in vicinity of bone. In study IV, we proposed a concept for patient-specific quality assurance of attenuation maps, based on measurements of the MRI B0-field. The method shows potential to find errors in the attenuation map related to metallic implants, air, and patient contour.

The work in this thesis has contributed to increased knowledge about the effect of resolution and noise for quantification in PET images. It has also introduced a new method for attenuation correction in PET/MRI, and a concept for quality assurance of PET/MRI attenuation maps.

Background: Attenuation correction of PET/MRI is a remaining problem for whole-body PET/MRI. The statistical decomposition algorithm (SDA) is a probabilistic atlas-based method that calculates synthetic CTs from T2-weighted MRI scans. In this study, we evaluated the application of SDA for attenuation correction of PET images in the pelvic region.

Materials and method: Twelve patients were retrospectively selected from an ongoing prostate cancer research study. The patients had same-day scans of [11C]acetate PET/MRI and CT. The CT images were non-rigidly registered to the PET/MRI geometry, and PET images were reconstructed with attenuation correction employing CT, SDA-generated CT, and the built-in Dixon sequence-based method of the scanner. The PET images reconstructed using CT-based attenuation correction were used as ground truth.

Results: The mean whole-image PET uptake error was reduced from - 5.4% for Dixon-PET to - 0.9% for SDA-PET. The prostate standardized uptake value (SUV) quantification error was significantly reduced from - 5.6% for Dixon-PET to - 2.3% for SDA-PET.

Conclusion: Attenuation correction with SDA improves quantification of PET/MR images in the pelvic region compared to the Dixon-based method.

Aim/Introduction: Quantification in PET/MRI is of importance, and its accuracy is currently limited by the MR based attenuation correction estimate. A common method for attenuation correction of the pelvic region is based on a 2-echo Dixon MRI sequence for segmentation of fat and water and does not account for bone. In this work, we evaluate a new method for attenuation correction using an algorithm based on statistical decomposition of a T2 weighted MRI scan.

Materials and Methods: Substitute CT images (sCTs) were calculated from T2 weighted MRI scans with a statistical decomposition algorithm, originally developed for MRI-based radiotherapy dose-planning [1]. These sCTs benefits from having bone density information included, in addition to fat and water information. Prostate cancer patients from the PARAPLY study [2] were retrospectivelyselected, scanned with PET/MRI 11C-Acatate and CT the same day. The stand-alone CT images were transformed to the same geometry as the PET and MR images, using a non-rigid registration. CT images, generated sCT images, and the Dixonbased attenuation maps (MRAC), all in the same geometry, were together with the PET raw data used to reconstruct attenuation-corrected PET images using the PETrecon toolbox [GE Healthcare]. The two MR-based attenuation corrections were compared to the CT-based attenuation correction with root mean squared error (RMSE). Lesion analysis will also be reported. PET/MRI images were acquired on a Signa PET/MRI (GE Healthcare), and the CT images on a Brilliance Big Bore (Phillips Healthcare). The study will include 12 patients and a subset of 6 patients has been analyzed so far and is presented here.

Results: Soft tissue in-between pelvic bone structures were overestimated with 13% in MRAC-PET, and the error was reduced to 5% with sCT attenuation corrected PET (sCT-PET). For the whole patient volume, an average underestimation of 6% was found in the MRAC-PET, compared to 1% for sCTPET. RMSE within the body was reduced with a factor 2.5 with sCT-PET (RMSE=3.6%), compared to MRAC-PET (RMSE=8.8%).

Conclusion: Applying sCT from statistical decomposition as a base for calculation of attenuation maps reduces quantification errors in PET-images of the pelvic region compared to the common Dixon based method.

Objective: The reference Logan plot is a tool for determining the non-displaceable binding potential for dynamic PET exams using tracers with reversible bindings. Dynamic frame protocols affect noise in PET images and short frames can lead to quantitative uncertainties and noise-induced reconstruction bias. The aim of this study was to analyze the effect of frame binning on ¹¹C-Raclopride striatal binding potential from reference Logan analysis. Methods: 12 healthy volunteers were scanned in list mode using ¹¹C-raclopride, and the image data were reconstructed into 9 different frame binning schemes whereof 3 clinical schemes. Reconstruction was performed with 3 different algorithms, one based on filtered back projection (FBP) and two based on ordered subset expectation maximization (OSEM); one including resolution recovery. Logan plots were used for calculating the non-displaceable binding potential. Variation in binding potential was evaluated using Students t-tests. Results: It was found that frame lengths of up to 60 s gave significantly different results compared to the reference clinical protocol for OSEM, both with and without resolution recovery (maximum deviation: 10.3 % for the 15 s protocol). For FBP, frame lengths of up to 30 s gave significantly different results with a maximum deviation of 2.8 %. The higher sampling dependence of OSEM compared to FBP is likely due to noise-dependent bias in the OSEM algorithm, most apparent at high noise levels. Conclusions: Bias related to OSEM reconstruction of high-noise data is an important factor for dynamic PET protocols. Time frames of 120 s or more generate the most stable values for the striatum binding potential with the reference Logan plot for ¹¹C-Raclopride brain PET.

In PET, partial-volume effects cause errors in estimation of size and activity for small objects with radiopharmaceutical uptake. Recent methods for image reconstruction, compared with traditional reconstruction techniques, include algorithms for resolution recovery that result in images with higher resolution and enable quantification of size and activity of smaller objects. The purpose of this study was to evaluate a combination of 2 algorithms for volume delineation and partial-volume correction on uptake volumes smaller than 0.7 mL using image reconstruction algorithms with and without resolution recovery.

METHODS: Volumes of interests (VOIs) were delineated using a threshold intensity calculated as a weighted sum of tumor and background intensities. These VOIs were used for calculating correction factors by convolving a tumor mask with the system point-spread function. The methods algorithms were evaluated using a phantom constructed from 5 small different-sized balloons filled with (18)F-FDG in background activity. Six different backgrounds were used. Data were acquired using a PET/CT scanner, and the images were reconstructed using 2 iterative algorithms, one of which used a resolution recovery algorithm.

RESULTS: For the images reconstructed using the resolution recovery algorithm, the method for volume delineation resulted in VOI sizes that were correct within 1 SD for all balloons of a volume of 0.35 mL (equivalent diameter, 8.8 mm) and larger, in all backgrounds. For the images reconstructed without resolution recovery, the VOI sizes were background-dependent and generally less accurate. Correct volume delineations generally led to accurate activity estimates.

CONCLUSION: The algorithms tested on the phantom developed for this study could, for this PET camera and these reconstruction algorithms, be used for accurate volume delineation and activity quantification of lesions 0.35 mL and larger.

The Logan plot is a graphical method for reversible tracer bindings. The bias and uncertainties of this method have previously been analyzed with respect to noise, but little is known about the direct effects from varying the time sampling scheme. This study aims to investigate the effect of time sampling on the binding potential from the reference Logan plot. Image data from seven healthy subjects imaged with [11C]raclopride was reconstructed into six dynamic series of equal length time frames with frame times between 15 s and 480 s. Images were reconstructed using both filtered back projection (FBP) and a resolution enhanced ordered subset expectation maximization (OSEM) algorithm, SharpIR. For each sampling scheme, the nondisplaceable binding potential (BPND) parameter was calculated from the reference Logan plot with cerebellum as a reference region. The variation in BPND was analyzed as percentage deviations from the BPND for the 480 s scheme. R-2 of the linear fit was also analyzed. Comparison between all sampling schemes showed that the largest deviation in BPND was 7.4% between the 15 s sampling scheme and the 480 s sampling scheme reconstructed with SharpIR. The corresponding deviation for FBP images was 1.6%. R-2 was highest for long time frames, but all R-2 values were above 0.997 in this study.