Simultaneous imaging systems combining positron emission tomography (PET) and magnetic resonance imaging (MRI) have been actively investigated. A PET/MR imaging system (GE Healthcare) comprised of a time-of-flight (TOF) PET system utilizing silicon photomultipliers (SiPMs) and 3-tesla (3T) MRI was recently installed at our institution. The small-ring (60 cm diameter) TOF PET subsystem of this PET/MRI system can generate images with higher spatial resolution compared with conventional PET systems. We have examined theoretically and experimentally the effect of uniform magnetic fields on the spatial resolution for high-energy positron emitters. Positron emitters including ¹⁸F, ¹²⁴I, and ⁶⁸Ga were simulated in water using the Geant4 Monte Carlo toolkit in the presence of a uniform magnetic field (0, 3, and 7 Tesla). The positron annihilation position was tracked to determine the 3D spatial distribution of the 511-keV gammy ray emission. The full-width at tenth maximum (FWTM) of the positron point spread function (PSF) was determined. Experimentally, ¹⁸F and ⁶⁸Ga line source phantoms in air and water were imaged with an investigational PET/MRI system and a PET/CT system to investigate the effect of magnetic field on the spatial resolution of PET. The full-width half maximum (FWHM) of the line spread function (LSF) from the line source was determined as the system spatial resolution. Simulations and experimental results show that the in-plane spatial resolution was slightly improved at field strength as low as 3 Tesla, especially when resolving signal from high-energy positron emitters in the air-tissue boundary.

Purpose

Iodine-131-m-iodobenzylguanidine ([(131)I]mIBG)-targeted radionuclide therapy (TRT) is a standard treatment for recurrent or refractory neuroblastoma with response rates of 30-40 %. The aim of this study is to demonstrate patient-specific dosimetry using quantitative [(124)I]mIBG positron emission tomography/X-ray computed tomography (PET/CT) imaging with a GEometry ANd Tracking 4 (Geant4)-based Monte Carlo method for better treatment planning.

Procedures

A Monte Carlo dosimetry method was developed using the Geant4 toolkit with voxelized anatomical geometry and source distribution as input. The presegmented hybrid computational human phantoms developed by the University of Florida and the National Cancer Institute (UF/NCI) were used as a surrogate to characterize the anatomy of a given patient. S values for I-131 were estimated by the phantoms coupled with Geant4 and compared with those estimated by OLINDA|EXM and MCNPX for the newborn model. To obtain patient-specific biodistribution of [(131)I]mIBG, a 10-year-old girl with relapsed neuroblastoma was imaged with [(124)I]mIBG PET/CT at four time points prior to the planned [(131)I]mIBG TRT. The organ- and tumor-absorbed doses of the clinical case were estimated with the Geant4 method using the modified UF/NCI 10-year-old phantom with tumors and the patient-specific residence time.

Results

For the newborn model, the Geant4 S values were consistent with the MCNPX S values. The S value ratio of the Geant4 method to OLINDA|EXM ranged from 0.08 to 6.5 of all major organs. The [(131)I]mIBG residence time quantified from the pretherapy [(124)I]mIBG PET/CT imaging of the 10-year-old patient was mostly comparable to those previously reported. Organ-absorbed dose for the salivary glands was 98.0 Gy, heart wall 36.5 Gy, and liver 34.3 Gy, while tumor-absorbed dose ranged from 143.9 to 1,641.3 Gy in different sites.

Conclusions

Patient-specific dosimetry for [(131)I]mIBG TRT was accomplished using pretherapy [(124)I]mIBG PET/CT imaging and a Geant4-based Monte Carlo dosimetry method. The Geant4 method with quantitative pretherapy imaging can provide dose estimates to normal organs and tumors with more realistic simulation geometry, and thus may improve treatment planning for [(131)I]mIBG TRT.

Purpose

Radiation dose calculated on tumors for radiopharmaceutical therapy varies significantly from tumor to tumor and from patient to patient. Accurate estimation of radiation dose requires multiple time point measurements using radionuclide imaging modalities such as SPECT or PET. In this report, we show our technical development of reducing the number of scans needed for reasonable estimation of tumor and normal organ dose in our pretherapy imaging and dosimetry platform of ¹²⁴ I-metaiodobenzylguanidine (MIBG) positron emission tomography/computed tomography (PET/CT) for ¹³¹ I-MIBG therapy of neuroblastoma.

Methods

We analyzed the simplest kinetic data, areas of two-time point data for five patients with neuroblastoma who underwent 3 or 4 times of ¹²⁴ I-MIBG PET/CT scan prior to ¹³¹ I-MIBG therapy. The data for which we derived areas were percent of injected activity (%IA) and standardized uptake value of tumors. These areas were correlated with time-integrated activity coefficients (TIACs) from full data (3 or 4 time points). TIACs are direct correlates with radiation dose as long as the volume and the radionuclide are known.

Results

The areas of %IAs between data obtained from all the two-time points with time points 1 and 2 (day 0 and day 1), time points 2 and 3 (day 1 and day 2), and time points 1 and 3 (day 0 and day 2) showed reasonable correlation (Pearson's correlation coefficient |r| > 0.5) with not only tumor and organ TIACs but also tumor and organ absorbed doses. The tumor and organ doses calculated using %IA areas of time point 1 and time point 2 were our best fits at about 20% individual percent difference compared to doses calculated using 3 or 4 time points.

Conclusions

We could achieve reasonable accuracy of estimating tumor doses for subsequent radiopharmaceutical therapy using only the two-time point imaging sessions. Images obtained from these time points (within the 48-h after administration of radiopharmaceutical) were also viewed as useful for diagnostic reading. Although our analysis was specific to ¹²⁴ I-MIBG PET/CT pretherapy imaging data for ¹³¹ I-MIBG therapy of neuroblastoma and the number of imaging datasets was not large, this feasible methodology would generally be applicable to other imaging and therapeutic radionuclides with an appropriate data analysis similar to our analysis to other imaging and therapeutic radiopharmaceuticals.

The metaiodobenzylguanidine (MIBG) scan is one of the most sensitive noninvasive lesion detection modalities for neuroblastoma. Unlike ¹²³I-MIBG, ¹²⁴I-MIBG allows high-resolution PET. We evaluated ¹²⁴I-MIBG PET/CT for its diagnostic performance as directly compared with paired ¹²³I-MIBG scans. Methods: Before ¹³¹I-MIBG therapy, standard ¹²³I-MIBG imaging (5.2 MBq/kg) was performed on 7 patients, including whole-body (anterior-posterior) planar imaging, focused-field-of-view SPECT/CT, and whole-body ¹²⁴I-MIBG PET/CT (1.05 MBq/kg). After therapy, 2 of 7 patients also completed ¹²⁴I-MIBG PET/CT as well as paired ¹²³I-MIBG planar imaging and SPECT/CT. One patient underwent ¹²⁴I-MIBG PET/CT only after therapy. We evaluated all 8 patients who showed at least 1 ¹²³I-MIBG-positive lesion with a total of 10 scans. In 8 pairs, ¹²³I-MIBG and ¹²⁴I-MIBG were performed within 1 mo of each other. The locations of identified lesions, the number of total lesions, and the curie scores were recorded for the ¹²³I-MIBG and ¹²⁴I-MIBG scans. Finally, for 5 patients who completed at least 3 PET/CT scans after administration of ¹²⁴I-MIBG, we estimated the effective dose of ¹²⁴I-MIBG. Results:¹²³I-MIBG whole-body planar scans, focused-field-of-view SPECT/CT scans, and whole-body ¹²⁴I-MIBG PET scans found 25, 32, and 87 total lesions, respectively. There was a statistically significant difference in lesion detection for ¹²⁴I-MIBG PET/CT versus ¹²³I-MIBG planar imaging (P < 0.0001) and ¹²³I-MIBG SPECT/CT (P < 0.0001). The curie scores were also higher for ¹²⁴I-MIBG PET/CT than for ¹²³I-MIBG planar imaging and SPECT/CT in 6 of 10 patients. ¹²⁴I-MIBG PET/CT demonstrated better detection of lesions throughout the body, including the chest, spine, head and neck, and extremities. The effective dose estimated for patient-specific ¹²⁴I-MIBG was approximately 10 times that of ¹²³I-MIBG; however, given that we administered a very low activity of ¹²⁴I-MIBG (1.05 MBq/kg), the effective dose was only approximately twice that of ¹²³I-MIBG despite the large difference in half-lives (100 vs. 13.2 h). Conclusion: The first-in-humans use of low-dose ¹²⁴I-MIBG PET for monitoring disease burden demonstrated tumor detection capability superior to that of ¹²³I-MIBG planar imaging and SPECT/CT.

Radiomics is an emerging technology for imaging biomarker discovery and disease-specific personalized treatment management. This paper aims to determine the benefit of using multi-modality radiomics data from PET and MR images in the characterization breast cancer phenotype and prognosis. Eighty-four features were extracted from PET and MR images of 113 breast cancer patients. Unsupervised clustering based on PET and MRI radiomic features created three subgroups. These derived subgroups were statistically significantly associated with tumor grade (p = 2.0 × 10^-6), tumor overall stage (p = 0.037), breast cancer subtypes (p = 0.0085), and disease recurrence status (p = 0.0053). The PET-derived first-order statistics and gray level co-occurrence matrix (GLCM) textural features were discriminative of breast cancer tumor grade, which was confirmed by the results of L2-regularization logistic regression (with repeated nested cross-validation) with an estimated area under the receiver operating characteristic curve (AUC) of 0.76 (95% confidence interval (CI) = [0.62, 0.83]). The results of ElasticNet logistic regression indicated that PET and MR radiomics distinguished recurrence-free survival, with a mean AUC of 0.75 (95% CI = [0.62, 0.88]) and 0.68 (95% CI = [0.58, 0.81]) for 1 and 2 years, respectively. The MRI-derived GLCM inverse difference moment normalized (IDMN) and the PET-derived GLCM cluster prominence were among the key features in the predictive models for recurrence-free survival. In conclusion, radiomic features from PET and MR images could be helpful in deciphering breast cancer phenotypes and may have potential as imaging biomarkers for prediction of breast cancer recurrence-free survival.

Purpose To develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease (AD), mild cognitive impairment, or neither at fluorine 18 (¹⁸F) fluorodeoxyglucose (FDG) PET of the brain and compare its performance to that of radiologic readers. Materials and Methods Prospective ¹⁸F-FDG PET brain images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (2109 imaging studies from 2005 to 2017, 1002 patients) and retrospective independent test set (40 imaging studies from 2006 to 2016, 40 patients) were collected. Final clinical diagnosis at follow-up was recorded. Convolutional neural network of InceptionV3 architecture was trained on 90% of ADNI data set and tested on the remaining 10%, as well as the independent test set, with performance compared to radiologic readers. Model was analyzed with sensitivity, specificity, receiver operating characteristic (ROC), saliency map, and t-distributed stochastic neighbor embedding. Results The algorithm achieved area under the ROC curve of 0.98 (95% confidence interval: 0.94, 1.00) when evaluated on predicting the final clinical diagnosis of AD in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity; P < .05). Saliency map demonstrated attention to known areas of interest but with focus on the entire brain. Conclusion By using fluorine 18 fluorodeoxyglucose PET of the brain, a deep learning algorithm developed for early prediction of Alzheimer disease achieved 82% specificity at 100% sensitivity, an average of 75.8 months prior to the final diagnosis. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Larvie in this issue.