MRI-guided attenuation correction in torso PET/MRI: Assessment of segmentation-, atlas-, and deep learning-based approaches in the presence of outliers

Purpose

We compare the performance of three commonly used MRI‐guided attenuation correction approaches in torso PET/MRI, namely segmentation‐, atlas‐, and deep learning‐based algorithms.

Methods

Twenty‐five co‐registered torso ¹⁸F‐FDG PET/CT and PET/MR images were enrolled. PET attenuation maps were generated from in‐phase Dixon MRI using a three‐tissue class segmentation‐based approach (soft‐tissue, lung, and background air), voxel‐wise weighting atlas‐based approach, and a residual convolutional neural network. The bias in standardized uptake value (SUV) was calculated for each approach considering CT‐based attenuation corrected PET images as reference. In addition to the overall performance assessment of these approaches, the primary focus of this work was on recognizing the origins of potential outliers, notably body truncation, metal‐artifacts, abnormal anatomy, and small malignant lesions in the lungs.

Results

The deep learning approach outperformed both atlas‐ and segmentation‐based methods resulting in less than 4% SUV bias across 25 patients compared to the segmentation‐based method with up to 20% SUV bias in bony structures and the atlas‐based method with 9% bias in the lung. The deep learning‐based method exhibited superior performance. Yet, in case of sever truncation and metallic‐artifacts in the input MRI, this approach was outperformed by the atlas‐based method, exhibiting suboptimal performance in the affected regions. Conversely, for abnormal anatomies, such as a patient presenting with one lung or small malignant lesion in the lung, the deep learning algorithm exhibited promising performance compared to other methods.

Conclusion

The deep learning‐based method provides promising outcome for synthetic CT generation from MRI. However, metal‐artifact and body truncation should be specifically addressed.

Impact of the Noise Penalty Factor on Quantification in Bayesian Penalized Likelihood (Q.Clear) Reconstructions of 68Ga-PSMA PET/CT Scans

Functional imaging with ⁶⁸Ga prostate-specific membrane antigen (PSMA) and positron emission tomography (PET) can fulfill an important role in treatment selection and adjustment in prostate cancer. This article focusses on quantitative assessment of ⁶⁸Ga-PSMA-PET. The effect of various parameters on standardized uptake values (SUVs) is explored, and an optimal Bayesian penalized likelihood (BPL) reconstruction is suggested. PET acquisitions of two phantoms consisting of a background compartment and spheres with diameter 4 mm to 37 mm, both filled with solutions of ⁶⁸Ga in water, were performed with a GE Discovery 710 PET/CT scanner. Recovery coefficients (RCs) in multiple reconstructions with varying noise penalty factors and acquisition times were determined and analyzed. Apparent recovery coefficients of spheres with a diameter smaller than 17 mm were significantly lower than those of spheres with a diameter of 17 mm and bigger (p < 0.001) for a tumor-to-background (T/B) ratio of 10:1 and a scan time of 10 min per bed position. With a T/B ratio of 10:1, the four largest spheres exhibit significantly higher RCs than those with a T/B ratio of 20:1 (p < 0.0001). For spheres with a diameter of 8 mm and less, alignment with the voxel grid potentially affects the RC. Evaluation of PET/CT scans using (semi-)quantitative measures such as SUVs should be performed with great caution, as SUVs are influenced by scanning and reconstruction parameters. Based on the evaluation of multiple reconstructions with different β of phantom scans, an intermediate β (600) is suggested as the optimal value for the reconstruction of clinical ⁶⁸Ga-PSMA PET/CT scans, considering that both detectability and reproducibility are relevant.

Quantitative Comparison of Prone and Supine PERCIST Measurements in Breast Cancer

Positron emission tomography (PET) is typically performed in the supine position. However, breast magnetic resonance imaging (MRI) is performed in prone, as this improves visibility of deep breast tissues. With the emergence of hybrid scanners that integrate molecular information from PET and functional information from MRI, it is of great interest to determine if the prognostic utility of prone PET is equivalent to supine. We compared PERCIST (PET Response Criteria in Solid Tumors) measurements between prone and supine FDG-PET in patients with breast cancer and the effect of orientation on predicting pathologic complete response (pCR). In total, 47 patients were enrolled and received up to 6 cycles of neoadjuvant therapy. Prone and supine FDG-PET were performed at baseline (t₀; n = 46), after cycle 1 (t₁; n = 1) or 2 (t₂; n = 10), or after all neoadjuvant therapy (t₃; n = 19). FDG uptake was quantified by maximum and peak standardized uptake value (SUV) with and without normalization to lean body mass; that is, SUV_max, SUV_peak, SUL_max, and SUL_peak. PERCIST measurements were performed for each paired baseline and post-treatment scan. Receiver operating characteristic analysis for the prediction of pCR was performed using logistic regression that included age and tumor size as covariates. SUV and SUL metrics were significantly different between orientation (P < .001), but were highly correlated (P > .98). Importantly, no differences were observed with the PERCIST measurements (P > .6). Overlapping 95% confidence intervals for the receiver operating characteristic analysis suggested no difference at predicting pCR. Therefore, prone and supine PERCIST in this data set were not statistically different.

Quantitative applications in positron emission tomography achieved through signal modelling

Positron emission tomography (PET) is a valuable tool in medical imaging, but it provides limited quantitative utility in a number of important applications, such as mapping of tracer accumulation in small tissues and quantitative assessment of factors affecting tracer uptake. We aimed to develop a quantification approach based on signal modelling, to address the above limitations. Our signal modelling approach allows for a comprehensive description of target and background signals. We used in silico simulations to exemplify the quantitative utility of signal modelling in a number of applications and conducted scans of standardized PET phantoms to validate our computer simulation algorithms. The simulations showed that the modelling approach allows applications not supported by current techniques, such as estimation of activity fractions of sub-resolution small tissues and accurate quantification of the effect of biological factors, such as hypoxia, on tracer accumulation. There was strong agreement between the simulation data and actual scans of phantoms, providing support for the validity of the simulation algorithms. We conclude that the presented signal modelling approach may provide a framework for image analysis that can improve and expand the quantitative capacity of PET imaging.

Comparison of 3 Devices for Automated Infusion of Positron-Emitting Radiotracers

The administration accuracy and precision of an automated infusion device for positron-emitting radiotracers are directly associated with bias and variance in the SUVs of ¹⁸F-FDG PET/CT. Therefore, the accuracy of such devices must be confirmed and calibrated at locations in which they are used. The present study aimed to validate the administration accuracy of 3 automated infusion devices for quantitative PET assessment. Methods: Temporal variations as well as variations in radioactive concentrations and dispensed volumes of ¹⁸F-FDG were determined for the M-130, AI-300, and UG-05 automated infusion devices. The total-test dispensed volumes were 25, 20, and 18.5 mL, respectively. A reference value was generated by measuring amounts of radioactivity using a standard dose calibrator. Administration accuracy was validated according to the criteria of the Japanese Society of Nuclear Medicine. Results: The temporal variation in the M-130 and UG-05 for a specified 185 MBq was relatively stable, in the range of -1.60%-0.92% and 1.16%-5.35%, respectively, whereas that in the AI-300 was -0.55%-8.68%. For the M-130 and UG-05 devices, the difference between measured and reference value was in the range of -5%-5%. The values measured by the AI-300 deviated from the reference values by a maximum of 30%, which depends on radioactive concentration and dispensed volume of ¹⁸F-FDG. Conclusion: The administration accuracy of the AI-300 varied considerably under different conditions, but a software update might somewhat improve this. Our findings indicate that dispensed volumes of ¹⁸F-FDG should be carefully considered when the radioactive concentration is high. Administration accuracy should be regularly confirmed at each location to maintain the quality of quantitative PET assessment. The present study provides useful information about how to confirm the administration accuracy of automated infusion devices.

Quantitative FDG PET/CT in the community: experience from interpretation of outside oncologic PET/CT exams in referred cancer patients

PURPOSE

Tertiary care institutions often deal with patients who have had a baseline positron emission tomography (PET)/computed tomography (CT) scan performed elsewhere. Little data exist regarding the quality of these PET/CT scans and whether they are fully suitable for qualitative or quantitative interpretation. We evaluated outside PET/CT scans from cancer patients referred to our institution and compared them with PET/CT scans acquired locally.

METHODS

This Health Insurance Portability and Accountability Act-compliant retrospective study was approved by our institutional review board. Informed consent requirements were waived. One hundred seventy recent whole-body outside PET/CT exams from many sites were digitally imported into our radiology imaging system and reviewed for key quality metrics including time from injection until imaging, availability of patient height and weight information, serum glucose level and [(18) F]fluoro-2-deoxy-D-glucose (FDG) dose. The standardised uptake value (SUV) and SUV based on lean body mass (SUL) in the liver were measured whenever possible. These were compared with 170 internal studies performed at our centre during the same period.

RESULTS

Missing data were common in outside scans with height in 62%, weight 35%, uptake time 25%, FDG dose 28% and glucose levels in 64% of cases. In quantitatively evaluable cases, mean liver SUL, SUV, FDG dose and uptake time were much more variable in outside than in internal studies.

CONCLUSION

Approximately one-third of the outside PET/CT studies submitted digitally for analysis lacked key information required to secure any quantitative imaging data. Only about a third of these studies had all necessary information available for accurate SUL determination and had acceptable quality that was comparable with locally acquired scans. This suggests that many of PET studies performed in the community cannot be relied upon to provide quantitative image data that can be applied in a different centre. Greater standardisation of oncologic PET/CT studies among different centres must still be pursued.