Deep Learning Denoising Improves and Homogenizes Patient [18F]FDG PET Image Quality in Digital PET/CT

Given the constant pressure to increase patient throughput while respecting radiation protection, global body PET image quality (IQ) is not satisfactory in all patients. We first studied the association between IQ and other variables, in particular body habitus, on a digital PET/CT. Second, to improve and homogenize IQ, we evaluated a deep learning PET denoising solution (Subtle PET^TM) using convolutional neural networks. We analysed retrospectively in 113 patients visual IQ (by a 5-point Likert score in two readers) and semi-quantitative IQ (by the coefficient of variation in the liver, CV_liv) as well as lesion detection and quantification in native and denoised PET. In native PET, visual and semi-quantitative IQ were lower in patients with larger body habitus (p < 0.0001 for both) and in men vs. women (p ≤ 0.03 for CV_liv). After PET denoising, visual IQ scores increased and became more homogeneous between patients (4.8 ± 0.3 in denoised vs. 3.6 ± 0.6 in native PET; p < 0.0001). CV_liv were lower in denoised PET than in native PET, 6.9 ± 0.9% vs. 12.2 ± 1.6%; p < 0.0001. The slope calculated by linear regression of CV_liv according to weight was significantly lower in denoised than in native PET (p = 0.0002), demonstrating more uniform CV_liv. Lesion concordance rate between both PET series was 369/371 (99.5%), with two lesions exclusively detected in native PET. SUV_max and SUV_peak of up to the five most intense native PET lesions per patient were lower in denoised PET (p < 0.001), with an average relative bias of -7.7% and -2.8%, respectively. DL-based PET denoising by Subtle PET^TM allowed [¹⁸F]FDG PET global image quality to be improved and homogenized, while maintaining satisfactory lesion detection and quantification. DL-based denoising may render body habitus adaptive PET protocols unnecessary, and pave the way for the improvement and homogenization of PET modalities.

Artificial intelligence-based 68Ga-DOTATOC PET denoising for optimizing 68Ge/68Ga generator use throughout its lifetime

Introduction

The yield per elution of a 68Ge/68Ga generator decreases during its lifespan. This affects the number of patients injected per elution or the injected dose per patient, thereby negatively affecting the cost of examinations and the quality of PET images due to increased image noise. We aimed to investigate whether AI-based PET denoising can offset this decrease in image quality parameters.

Methods

All patients addressed to our PET unit for a 68Ga-DOTATOC PET/CT from April 2020 to February 2021 were enrolled. Forty-four patients underwent their PET scans according to Protocol_FixedDose (150 MBq) and 32 according to Protocol_WeightDose (1.5 MBq/kg). Protocol_WeightDose examinations were processed using the Subtle PET software (Protocol_WeightDoseAI). Liver and vascular SUV mean were recorded as well as SUVmax, SUVmean and metabolic tumour volume (MTV) of the most intense tumoural lesion and its background SUVmean. Liver and vascular coefficients of variation (CV), tumour-to-background and tumour-to-liver ratios were calculated.

Results

The mean injected dose of 2.1 (0.4) MBq/kg per patient was significantly higher in the Protocol_FixedDose group as compared to 1.5 (0.1) MBq/kg for the Protocol_WeightDose group. Protocol_WeightDose led to noisier images than Protocol_FixedDose with higher CVs for liver (15.57% ± 4.32 vs. 13.04% ± 3.51, p = 0.018) and blood-pool (28.67% ± 8.65 vs. 22.25% ± 10.37, p = 0.0003). Protocol_WeightDoseAI led to less noisy images than Protocol_WeightDose with lower liver CVs (11.42% ± 3.05 vs. 15.57% ± 4.32, p &lt; 0.0001) and vascular CVs (16.62% ± 6.40 vs. 28.67% ± 8.65, p &lt; 0.0001). Tumour-to-background and tumour-to-liver ratios were lower for protocol_WeightDoseAI: 6.78 ± 3.49 vs. 7.57 ± 4.73 (p = 0.01) and 5.96 ± 5.43 vs. 6.77 ± 6.19 (p &lt; 0.0001), respectively. MTVs were higher after denoising whereas tumour SUVmax were lower: the mean% differences in MTV and SUVmax were + 11.14% (95% CI = 4.84–17.43) and −3.92% (95% CI = −6.25 to −1.59).

Conclusion

The degradation of PET image quality due to a reduction in injected dose at the end of the 68Ge/68Ga generator lifespan can be effectively counterbalanced by using AI-based PET denoising.

Prediction of the Presence of Targetable Molecular Alteration(s) with Clinico-Metabolic 18 F-FDG PET Radiomics in Non-Asian Lung Adenocarcinoma Patients

This study aimed to investigate if combining clinical characteristics with pre-therapeutic 18 F-fluorodeoxyglucose (18 F-FDG) positron emission tomography (PET) radiomics could predict the presence of molecular alteration(s) in key molecular targets in lung adenocarcinoma. This non-interventional monocentric study included patients with newly diagnosed lung adenocarcinoma referred for baseline PET who had tumour molecular analyses. The data were randomly split into training and test datasets. LASSO regression with 100-fold cross-validation was performed, including sex, age, smoking history, AJCC cancer stage and 31 PET variables. In total, 109 patients were analysed, and it was found that 63 (57.8%) patients had at least one molecular alteration. Using the training dataset (n = 87), the model included 10 variables, namely age, sex, smoking history, AJCC stage, excessKustosis_HISTO, sphericity_SHAPE, variance_GLCM, correlation_GLCM, LZE_GLZLM, and GLNU_GLZLM. The ROC analysis for molecular alteration prediction using this model found an AUC equal to 0.866 (p < 0.0001). A cut-off value set to 0.48 led to a sensitivity of 90.6% and a positive likelihood ratio (LR+) value equal to 2.4. After application of this cut-off value in the unseen test dataset of patients (n = 22), the test presented a sensitivity equal to 90.0% and an LR+ value of 1.35. A clinico-metabolic 18 F-FDG PET phenotype allows the detection of key molecular target alterations with high sensitivity and negative predictive value. Hence, it opens the way to the selection of patients for molecular analysis.

The Impact of Artificial Intelligence CNN Based Denoising on FDG PET Radiomics

Background

With a constantly increasing number of diagnostic images performed each year, Artificial Intelligence (AI) denoising methods offer an opportunity to respond to the growing demand. However, it may affect information in the image in an unknown manner. This study quantifies the effect of AI-based denoising on FDG PET textural information in comparison to a convolution with a standard gaussian postfilter (EARL1).

Methods

The study was carried out on 113 patients who underwent a digital FDG PET/CT (VEREOS, Philips Healthcare). 101 FDG avid lesions were segmented semi-automatically by a nuclear medicine physician. VOIs in the liver and lung as reference organs were contoured. PET textural features were extracted with pyradiomics. Texture features from AI denoised and EARL1 versus original PET images were compared with a Concordance Correlation Coefficient (CCC). Features with CCC values ≥ 0.85 threshold were considered concordant. Scatter plots of variable pairs with R2 coefficients of the more relevant features were computed. A Wilcoxon signed rank test to compare the absolute values between AI denoised and original images was performed.

Results

The ratio of concordant features was 90/104 (86.5%) in AI denoised versus 46/104 (44.2%) with EARL1 denoising. In the reference organs, the concordant ratio for AI and EARL1 denoised images was low, respectively 12/104 (11.5%) and 7/104 (6.7%) in the liver, 26/104 (25%) and 24/104 (23.1%) in the lung. SUVpeak was stable after the application of both algorithms in comparison to SUVmax. Scatter plots of variable pairs showed that AI filtering affected more lower versus high intensity regions unlike EARL1 gaussian post filters, affecting both in a similar way. In lesions, the majority of texture features 79/100 (79%) were significantly (p<0.05) different between AI denoised and original PET images.

Conclusions

Applying an AI-based denoising on FDG PET images maintains most of the lesion’s texture information in contrast to EARL1-compatible Gaussian filter. Predictive features of a trained model could be thus the same, however with an adapted threshold. Artificial intelligence based denoising in PET is a very promising approach as it adapts the denoising in function of the tissue type, preserving information where it should.

Radioimmunotherapy for Brain Metastases: The Potential for Inflammation as a Target of Choice

Brain metastases (BM) are frequently detected during the follow-up of patients with malignant tumors, particularly in those with advanced disease. Despite a major progress in systemic anti-cancer treatments, the average overall survival of these patients remains limited (6 months from diagnosis). Also, cognitive decline is regularly reported especially in patients treated with whole brain external beam radiotherapy (WBRT), due to the absorbed radiation dose in healthy brain tissue. New targeted therapies, for an earlier and/or more specific treatment of the tumor and its microenvironment, are needed. Radioimmunotherapy (RIT), a combination of a radionuclide to a specific antibody, appears to be a promising tool. Inflammation, which is involved in multiple steps, including the early phase, of BM development is attractive as a relevant target for RIT. This review will focus on the (1) early biomarkers of inflammation in BM pertinent for RIT, (2) state of the art studies on RIT for BM, and (3) the importance of dosimetry to RIT in BM. These two last points will be addressed in comparison to the conventional EBRT treatment, particularly with respect to the balance between tumor control and healthy tissue complications. Finally, because new diagnostic imaging techniques show a potential for the detection of BM at an early stage of the disease, we focus particularly on this therapeutic window.

Author Gender Inequality in Medical Imaging Journals and the COVID-19 Pandemic

Background Early reports show the unequal effect the COVID-19 pandemic might have on men versus women engaged in medical research. Purpose To investigate whether the COVID-19 pandemic has had an impact on scientific publishing by female physicians in medical imaging. Materials and Methods The authors conducted a descriptive bibliometric analysis of the gender of the first and last authors of manuscripts submitted to the top 50 medical imaging journals from March to May 2020 (n = 2480) compared with the same period of the year in 2018 (n = 2238) and 2019 (n = 2355). Manuscript title, date of submission, first and last names of the first and last authors, journal impact factor, and author country of provenance were recorded. The Gender-API software was used to determine author gender. Statistical analysis comprised χ2 tests and multivariable logistic regression. Results Percentages of women listed as first and last authors were 31.6% (1172 of 3711 articles) and 19.3% (717 of 3711 articles), respectively, in 2018-2019 versus 32.3% (725 of 2248 articles) and 20.7% (465 of 2248 articles) in 2020 (P = .61 and P = .21, respectively). For COVID-19-related articles, 35.2% (89 of 253 articles) of first authors and 20.6% (52 of 253 articles) of last authors were women. No associations were found between first- and last-author gender, year of publication, and region of provenance. First and last authorship of high-ranking articles was not in favor of North American women whatever the year (odds ratio [OR], 0.79 [P = .05] and 0.72 [P = .02], respectively). Higher rates of female last authorship of high-ranking articles were observed in Europe (P = .003) and of female first authorship of low-ranking publications in Asia in 2020 (OR, 1.38; 95% CI: 0.98, 1.92; P = .06). Female first and last authorship of COVID-19-related articles was overrepresented for lowest-rank publications (P = .02 and P = .01, respectively). Conclusion One in three first authors and one in five last authors were women in 2018-2019 and 2020, respectively. Although the first 2020 lockdown did not diminish the quantity of women-authored publications, the impact on the quality was variable. ©RSNA, 2021 See also the editorial by Robbins and Khosa in this issue.

Gated 18F-FDG PET/CT of the Lung Using a Respiratory Spirometric Gating Device: A Feasibility Study

Spirometric gating devices (SGDs) can measure the respiratory signal with high temporal resolution and accuracy. The primary objective of this study was to assess the feasibility and tolerance of a gated lung PET/CT acquisition using an SGD. The secondary objective was to compare the technical quality, accuracy, and interoperability of the SGD with that of a standard respiratory gating device, Real-Time Position Management (RPM), based on measurement of vertical thoracoabdominal displacement. Methods: A prospective phase I monocentric clinical study was performed on patients undergoing ¹⁸F-FDG PET/CT for assessment of a solitary lung nodule, staging of lung malignancy, or planning of radiotherapy. After whole-body PET/CT, a centered gated acquisition of both PET and CT was simultaneously obtained with the SGD and RPM during normal breathing. Results: Of the 46 patients who were included, 6 were prematurely excluded (1 because of hyperglycemia and 5 because of distant metastases revealed by whole-body PET/CT, leading to an unjustified extra gated acquisition). No serious adverse events were observed. Of the 40 remaining patients, the gated acquisition was prematurely stopped in 1 patient because of mask discomfort (2.5%; confidence interval [CI], 0.1%-13.2%). This event was considered patient tolerance failure. The SGD generated accurately gated PET/CT images, with more than 95% of the breathing cycle detected and high temporal resolution, in 34 of the 39 patients (87.2%; 95% CI, 60.0%-100.0%) and failed to generate a biologic tumor volume in 1 of 21 patients with increased ¹⁸F-FDG uptake (4.8%; 95% CI, 0.1%-26.5%). The quality and accuracy of respiratory signal detection and synchronization were significantly better than those obtained with RPM (P < 0.05). Conclusion: This trial supports the use of an SGD for gated lung PET/CT because of its high patient tolerance and accuracy. Although this technique seems to technically outperform RPM for gated PET/CT, further assessment of its superiority and the clinical benefit is warranted. We believe that this technique could be used as a gold standard to develop innovative approaches to eliminate respiration-induced blurring artifacts.