[18F]DCFPyL PET/CT versus [18F]fluoromethylcholine PET/CT in Biochemical Recurrence of Prostate Cancer (PYTHON): a prospective, open label, cross-over, comparative study.

PURPOSE: Primary objective was to compare the per-patient detection rates (DR) of [18F]DCFPyL versus [18F]fluoromethylcholine positron emission tomography/computed tomography (PET/CT), in patients with first prostate cancer (PCa) biochemical recurrence (BCR). Secondary endpoints included safety and impact on patient management (PM). METHODS: This was a prospective, open label, cross-over, comparative study with randomized treatment administration of [18F]DCFPyL (investigational medicinal product) or [18F]fluoromethylcholine (comparator). Men with rising prostate-specific antigen (PSA) after initial curative therapy were enrolled. [18F]DCFPyL and [18F]fluoromethylcholine PET/CTs were performed within a maximum time interval of 12 days. DR was defined as the percentage of positive PET/CT scans identified by 3 central imaging readers. PM was assessed by comparing the proposed pre-PET/CT treatment with the local treatment", defined after considering both PET/CTs. RESULTS: A total of 205 patients with first BCR after radical prostatectomy (73%; median PSA = 0.46 ng/ml [CI 0.16;27.0]) or radiation therapy (27%; median PSA = 4.23 ng/ml [CI 1.4;98.6]) underwent [18F]DCFPyL- and/or [18F]fluoromethylcholine -PET/CTs, between July and December 2020, at 22 European sites. 201 patients completed the study. The per-patient DR was significantly higher for [18F]DCFPyL- compared to [18F]fluoromethylcholine -PET/CTs (58% (117/201 patients) vs. 40% (81/201 patients), p < 0.0001). DR increased with higher PSA values for both tracers (PSA ≤ 0.5 ng/ml: 26/74 (35%) vs. 22/74 (30%); PSA 0.5 to ≤ 1.0 ng/ml: 17/31 (55%) vs. 10/31 (32%); PSA 1.01 to < 2.0 ng/ml: 13/19 (68%) vs. 6/19 (32%);PSA > 2.0: 50/57 (88%) vs. 39/57 (68%) for [18F]DCFPyL- and [18F]fluoromethylcholine -PET/CT, respectively). [18F]DCFPyL PET/CT had an impact on PM in 44% (90/204) of patients versus 29% (58/202) for [18F]fluoromethylcholine. Overall, no drug-related nor serious adverse events were observed. CONCLUSIONS: The primary endpoint of this study was achieved, confirming a significantly higher detection rate for [18F]DCFPyL compared to [18F]fluoromethylcholine, in men with first BCR of PCa, across a wide PSA range. [18F]DCFPyL was safe and well tolerated.

The Sentinel Lymph Node in Treatment Planning: A Narrative Review of Lymph-Flow-Guided Radiotherapy.

The sentinel lymph node technique is minimally invasive and used routinely by surgeons, reducing the need for morbid extensive lymph node dissections, which is a significant advantage for cancer staging and treatment decisions. The sentinel lymph node could also help radiation oncologists to identify tumor drainage for each of their patients, leading to a more personalized radiotherapy, instead of a probabilistic irradiation based on delineation atlases. The aim is both to avoid recurrence in unexpected areas and to limit the volume of irradiated healthy tissues. The aim of our study is to evaluate the impact of sentinel lymph node mapping for radiation oncologists. This concept, relying on sentinel lymph node mapping for treatment planning, is known as lymph-flow-guided radiotherapy. We present an up-to-date narrative literature review showing the potential applications of the sentinel lymph node technique for radiotherapy, as well as the limits that need to be addressed before its routine usage.

Comparison of Hormone-Sensitive Oligorecurrent Prostate Cancer Patients Based on Routine Use of Choline and/or PSMA PET/CT to Guide Metastasis-Directed Therapy.

BACKGROUND: In hormone-sensitive oligorecurrent prostate cancer (PC), the literature showed [68Ga]Ga-PSMA (PSMA) and [18F]F-choline (FCH) PET/CT can successfully guide metastasis-directed therapies (MDT). This observational retrospective study aimed to explore, in routine use, the impact of FCH or PSMA PET/CT in guiding MDT for hormone-sensitive oligometastatic PC at different recurrences. METHODS: In 2017-2020, patients initially treated with radical prostatectomy but, in biochemical recurrence (with PSA ≤ 2 ng/mL), diagnosed as oligometastatic based on FCH or PSMA PET/CT, were identified. MDT was stereotactic body radiotherapy (SBRT), elective nodal or prostate bed radiotherapy ± boost and ± androgen deprivation therapy (ADT). The primary endpoint was biochemical relapse-free survival (BR-FS), defined as a PSA increase ≥ 0.2 ng/mL above the nadir and increasing over two successive samples and the secondaries were ADT-free survival (ADT-FS). RESULTS: 123 patients (70 PSMA and 53 FCH) were included. The median follow-up was 42.2 months. The median BR-FS was 24.7 months in the PSMA group versus 13.0 months in the FCH group (p = 0.008). Similarly, ADT-FS (p = 0.001) was longer in patients in the PSMA group. In multivariate analysis, a short PSA doubling time before imaging (p = 0.005) and MDT with SBRT (p = 0.001) were poor prognostic factors for BR-FS. CONCLUSIONS: Routine use of FCH or PSMA PET/CT in hormone-sensitive PC showed an advantage for using PSMA PET/CT to guide MDT in terms of BR-FS and ADT-FS in patients with low PSA value. Prospective studies are needed to confirm these hypotheses.

Multicentric development and evaluation of 18F-FDG PET/CT and MRI radiomics models to predict para-aortic lymph node involvement in locally advanced cervical cancer.

PURPOSE: To develop machine learning models to predict para-aortic lymph node (PALN) involvement in patients with locally advanced cervical cancer (LACC) before chemoradiotherapy (CRT) using 18F-FDG PET/CT and MRI radiomics combined with clinical parameters. METHODS: We retrospectively collected 178 patients (60% for training and 40% for testing) in 2 centers and 61 patients corresponding to 2 further external testing cohorts with LACC between 2010 to 2022 and who had undergone pretreatment analog or digital 18F-FDG PET/CT, pelvic MRI and surgical PALN staging. Only primary tumor volumes were delineated. Radiomics features were extracted using the Radiomics toolbox®. The ComBat harmonization method was applied to reduce the batch effect between centers. Different prediction models were trained using a neural network approach with either clinical, radiomics or combined models. They were then evaluated on the testing and external validation sets and compared. RESULTS: In the training set (n = 102), the clinical model achieved a good prediction of the risk of PALN involvement with a C-statistic of 0.80 (95% CI 0.71, 0.87). However, it performed in the testing (n = 76) and external testing sets (n = 30 and n = 31) with C-statistics of only 0.57 to 0.67 (95% CI 0.36, 0.83). The ComBat-radiomic (GLDZM_HISDE_PET_FBN64 and Shape_maxDiameter2D3_PET_FBW0.25) and ComBat-combined (FIGO 2018 and same radiomics features) models achieved very high predictive ability in the training set and both models kept the same performance in the testing sets, with C-statistics from 0.88 to 0.96 (95% CI 0.76, 1.00) and 0.85 to 0.92 (95% CI 0.75, 0.99), respectively. CONCLUSIONS: Radiomic features extracted from pre-CRT analog and digital 18F-FDG PET/CT outperform clinical parameters in the decision to perform a para-aortic node staging or an extended field irradiation to PALN. Prospective validation of our models should now be carried out.

Radioisotope-guided Lymphadenectomy for Pelvic Lymph Node Staging in Patients With Intermediate- and High-risk Prostate Cancer (The Prospective SENTINELLE Study).

PURPOSE: Our aim was to prospectively evaluate the diagnostic accuracy of sentinel lymph node biopsy-guided lymph node dissection compared to extended pelvic lymph node dissection in patients with intermediate- or high-risk prostate cancer. MATERIALS AND METHODS: We conducted a prospective, single-arm, multicenter study at 3 tertiary centers in France between February 2012 and May 2019. Eligible patients had clinically localized intermediate- or high-risk prostate cancer. After intraprostatic injection of (99m)Tc-nanocolloid, the locations of the sentinel lymph nodes were defined by preoperative lymphoscintigraphy. Surgical excision of the sentinel lymph nodes was performed using intraoperative gamma probe guidance. After resection of the sentinel lymph nodes, extended pelvic lymph node dissection was performed in all patients. We assessed the diagnostic accuracy of the sentinel lymph node biopsy method using extended pelvic lymph node dissection as the reference standard. This trial was registered in ClinicalTrials.gov (NCT02732392). RESULTS: A total of 162 men cN0M0 (CT scan and bone scan) were enrolled: 106 (65.4%) and 56 (34.6%) patients had intermediate- and high-risk prostate cancer, respectively. The median number of nodes retrieved was 14 (mean 16, IQR 10-21) per patient. At final pathological analysis, 22 patients (13.6%) were pN+. Sensitivity, specificity, negative predictive value, and positive predictive value of sentinel lymph node biopsy method in detecting patients with at least 1 lymph node metastasis were 95.4% (95% CI, 75.1-99.7), 100% (95% CI, 96.6-100), 99.2% (95% CI, 95.5-99.9), and 100% (95% CI, 80.7-100), respectively. CONCLUSIONS: Our multicenter prospective study supports that sentinel lymph node biopsy is a very effective and sensitive method for pelvic lymph node staging in patients with intermediate- or high-risk localized prostate cancer.

Immuno-PET: Design options and clinical proof-of-concept.

Radioimmunoconjugates have been used for over 30 years in nuclear medicine applications. In the last few years, advances in cancer biology knowledge have led to the identification of new molecular targets specific to certain patient subgroups. The use of these targets in targeted therapies approaches has allowed the developments of specifically tailored therapeutics for patients. As consequence of the PET-imaging progresses, nuclear medicine has developed powerful imaging tools, based on monoclonal antibodies, to in vivo characterization of these tumor biomarkers. This imaging modality known as immuno-positron emission tomography (immuno-PET) is currently in fastest-growing and its medical value lies in its ability to give a non-invasive method to assess the in vivo target expression and distribution and provide key-information on the tumor targeting. Currently, immuno-PET presents promising probes for different nuclear medicine topics as staging/stratification tool, theranostic approaches or predictive/prognostic biomarkers. To develop a radiopharmaceutical drug that can be used in immuno-PET approach, it is necessary to find the best compromise between the isotope choice and the immunologic structure (full monoclonal antibody or derivatives). Through some clinical applications, this paper review aims to discuss the most important aspects of the isotope choice and the usable proteic structure that can be used to meet the clinical needs.

Influence of inputs for bone lesion segmentation in longitudinal 18F-FDG PET/CT imaging studies.

In metastatic breast cancer, bone metastases are prevalent and associated with multiple complications. Assessing their response to treatment is therefore crucial. Most deep learning methods segment or detect lesions on a single acquisition while only a few focus on longitudinal studies. In this work, 45 patients with baseline (BL) and follow-up (FU) images recruited in the context of the EPICUREseinmeta study were analyzed. The aim was to determine if a network trained for a particular timepoint can generalize well to another one, and to explore different improvement strategies. Four networks based on the same 3D U-Net framework to segment bone lesions on BL and FU images were trained with different strategies and compared. These four networks were trained 1) only with BL images 2) only with FU images 3) with both BL and FU images 4) only with FU images but with BL images and bone lesion segmentations registered as input channels. With the obtained segmentations, we computed the PET Bone Index (PBI) which assesses the bone metastases burden of patients and we analyzed its potential for treatment response evaluation. Dice scores of 0.53, 0.55, 0.59 and 0.62 were respectively obtained on FU acquisitions. The under-performance of the first and third networks may be explained by the lower SUV uptake due to treatment response in FU images compared to BL images. The fourth network gives better results than the second network showing that the addition of BL PET images and bone lesion segmentations as prior knowledge has its importance. With an AUC of 0.86, the difference of PBI between two acquisitions could be used to assess treatment response. Clinical relevance- To assess the response to treatment of bone metastases, it is crucial to detect and segment them on several acquisitions from a same patient. We proposed a completely automatic method to detect and segment these metastases on longitudinal 18F-FDG PET/CT images in the context of metastatic breast cancer. We also proposed an automatic PBI to quantitatively assess the evolution of the bone metastases burden of patient and to automatically evaluate their response to treatment.

Deformable image registration with deep network priors: a study on longitudinal PET images.

Objective.This paper proposes a novel approach for the longitudinal registration of PET imaging acquired for the monitoring of patients with metastatic breast cancer. Unlike with other image analysis tasks, the use of deep learning (DL) has not significantly improved the performance of image registration. With this work, we propose a new registration approach to bridge the performance gap between conventional and DL-based methods: medical image registration method regularized by architecture (MIRRBA).Approach.MIRRBAis a subject-specific deformable registration method which relies on a deep pyramidal architecture to parametrize the deformation field. Diverging from the usual deep-learning paradigms,MIRRBAdoes not require a learning database, but only a pair of images to be registered that is used to optimize the network's parameters. We appliedMIRRBAon a private dataset of 110 whole-body PET images of patients with metastatic breast cancer. We used different architecture configurations to produce the deformation field and studied the results obtained. We also compared our method to several standard registration approaches: two conventional iterative registration methods (ANTs and Elastix) and two supervised DL-based models (LapIRN and Voxelmorph). Registration accuracy was evaluated using the Dice score, the target registration error, the average Hausdorff distance and the detection rate, while the realism of the registration obtained was evaluated using Jacobian's determinant. The ability of the different methods to shrink disappearing lesions was also computed with the disappearing rate.Main results.MIRRBA significantly improved all metrics when compared to DL-based approaches. The organ and lesion Dice scores of Voxelmorph improved by 6% and 52% respectively, while the ones of LapIRN increased by 5% and 65%. Regarding conventional approaches, MIRRBA presented comparable results showing the feasibility of our method.Significance.In this paper, we also demonstrate the regularizing power of deep architectures and present new elements to understand the role of the architecture in DL methods used for registration.

Vaccine-Related Lymph Nodes: The Emerging Pitfalls of 18F-Fluorocholine and 68Ga-PSMA-11 PET/CT in the Era of COVID-19 Vaccination.

PURPOSE: Vaccination against coronavirus disease 2019 (COVID-19) is currently under worldwide deployment. The consequences of this vaccination can be seen in radiology and nuclear medicine explorations with visualization of axillary lymph nodes (LNs), as observed on ultrasonography, MRI, or 18F-FDG PET/CT.We aimed to evaluate on PET/CT the incidence of vaccine-related LNs and their characteristics after COVID-19 vaccination, using several radiopharmaceuticals different from 18F-FDG. PATIENTS AND METHODS: Between February and July 2021, all consecutive patients undergoing a whole-body PET/CT for any indication using a different radiopharmaceutical from 18F-FDG were eligible for inclusion if they had received at least 1 dose of the COVID-19 vaccine. The radiopharmaceutical administered and vaccine type were recorded for each patient. The incidence of positive vaccine-related axillary and supraclavicular LNs on PET/CT was our primary finding, along with the nodes characteristics. Statistical analyses were performed for patients with prostate cancer (PCa) to determine certain interaction factors that were associated with the detection of vaccine-related LNs. RESULTS: Of the 226 patients in our cohort study, 120 patients underwent an 18F-fluorocholine PET/CT, 79 a 68Ga-PSMA-11 PET/CT, 6 an 18F-FDOPA PET/CT, and 21 a 68Ga-DOTATOC PET/CT. A total of 67.3% of patients (152/226) received BNT162b2mRNA (Pfizer-BioNTech), 26.5% (60/226) ChAdOx1-S (AstraZeneca), 4.9% (11/226) mRNA-1273 (Moderna), and 1.3% (3/226) Ad26.COV2.S (Janssen). The incidence of positive vaccine-related axillary and supraclavicular LNs was 42.5% (51/120 patients) on PET/CT using 18F-fluorocholine and 12.7% (10/79 patients) with 68Ga-PSMA-11. None of our patients undergoing 18F-FDOPA or 68Ga-DOTATOC PET/CT presented any vaccine-related lymphadenopathy. Vaccine-related LNs were statistically associated with the nature of the radiopharmaceutical (P < 10-4), with the number of vaccine doses received (P = 0.041), with a short delay between vaccination and PET/CT realization (P < 10-5), and with a higher prostate-specific antigen level for patients with PCa (P = 0.032), but not with age or vaccine type. The vaccine-related nodes appeared in 85% of the cases, in the 30 days after vaccine injection, were limited in size and uptake, and were most often limited to the axilla level 1 area. CONCLUSIONS: Detecting positive LNs after COVID-19 vaccination is not an exclusive 18F-FDG PET/CT pattern but is common on 18F-fluorocholine and possible on 68Ga-PSMA-11 PET/CT. Confronting PET/CT findings with clinical data (such as date and site of injection) seems essential in the current pandemic context, just as it does for the radiopharmaceuticals used in PCa to avoid PET/CT misinterpretation and incorrect patient treatment. For 18F-FDOPA or 68Ga-DOTATOC PET/CT, this seems to have a lesser impact.

Bone Uptake in Prostate Cancer Patients: Diagnostic Performances of PSMA-RADS v1.0, Clinical, Biological, and 68 Ga-PSMA-11 PET Features to Predict Metastasis After Biochemical Recurrence.

PURPOSE: 68 Gallium-labeled prostate-specific membrane antigen-11 (PSMA) PET/CT is the new reference to identify relapse during biochemical recurrence of prostate cancer (PCa). However, this method lacks specificity for bone foci. This study aimed to report the prevalence of PCa bone metastases and to assess the diagnostic performances of PSMA reporting and data systems (RADS), clinical, biological, and imaging features for identification. PATIENTS AND METHODS: A multicentric retrospective cohort of consecutive patients with biochemical recurrence after local treatment was analyzed. Clinical and biological features at initial staging and during recurrence were retrieved from medical reports. The metastatic status of each bone uptake on PSMA PET/CT was determined according to histopathology, comparisons with concomitant and previous conventional imaging, prostate-specific antigen kinetic, and follow-up. Two nuclear medicine physicians assessed PSMA-RADS, anatomic location, radiological patterns, SUV max , and the presence of other molecular lesions. Univariate and multivariate analyses were conducted to identify independent predictors of PCa metastases. RESULTS: In the eligible population, 98/298 patients (32.9%) showed bone uptake on PSMA PET/CT. In patients with a final diagnosis, 28/81 lesions (34.6%) were metastases. PSMA-RADS-4 or 5 showed sensitivity of 79%, specificity of 94%, and accuracy of 89%. PSMA-RADS had a significantly higher area under the receiver operating characteristic curve than the initial reading in clinical practice (0.91 vs 0.83, P = 0.0074). Initial Gleason score ≥8, age ≤71 years at recurrence, and SUV max >6.21 were independent predictors of PCa metastases in multivariate logistic regression ( P = 0.0314, 0.0179, and 0.0003, respectively). CONCLUSIONS: Most bone uptakes at PSMA PET/CT were benign lesions. PSMA-RADS, patients and tumor characteristics, and SUV max could help identify PCa bone metastases.

Is Glucagon Administration Compatible With FDG PET/MRI?

ABSTRACT: Digestive peristalsis generates many artifacts that limit the abdominal and pelvic MRI interpretation. Apart from the hypoglycemia treatment in patients with diabetes, glucagon analog is also indicated for the digestive peristalsis reduction to reduce MRI artifacts. However, its use in PET/MRI is not described, given the risk of interaction with the metabolism of FDG. To assess the importance of this interaction on the FDG PET images, we report FDG PET/MRI images obtained with and without glucagon analog injection.

A transfer learning approach to facilitate ComBat-based harmonization of multicentre radiomic features in new datasets.

PURPOSE: To facilitate the demonstration of the prognostic value of radiomics, multicenter radiomics studies are needed. Pooling radiomic features of such data in a statistical analysis is however challenging, as they are sensitive to the variability in scanner models, acquisition protocols and reconstruction settings, which is often unavoidable in a multicentre retrospective analysis. A statistical harmonization strategy called ComBat was utilized in radiomics studies to deal with the "center-effect". The goal of the present work was to integrate a transfer learning (TL) technique within ComBat-and recently developed alternate versions of ComBat with improved flexibility (M-ComBat) and robustness (B-ComBat)-to allow the use of a previously determined harmonization transform to the radiomic feature values of new patients from an already known center. MATERIAL AND METHODS: The proposed TL approach were incorporated in the four versions of ComBat (standard, B, M, and B-M ComBat). The proposed approach was evaluated using a dataset of 189 locally advanced cervical cancer patients from 3 centers, with magnetic resonance imaging (MRI) and positron emission tomography (PET) images, with the clinical endpoint of predicting local failure. The impact performance of the TL approach was evaluated by comparing the harmonization achieved using only parts of the data to the reference (harmonization achieved using all the available data). It was performed through three different machine learning pipelines. RESULTS: The proposed TL technique was successful in harmonizing features of new patients from a known center in all versions of ComBat, leading to predictive models reaching similar performance as the ones developed using the features harmonized with all the data available. CONCLUSION: The proposed TL approach enables applying a previously determined ComBat transform to new, previously unseen data.

COVID-19 vaccination may cause FDG uptake beyond axillary area.

BACKGROUND: The vaccination immune response may induce false-positive 18F-FDG PET/CT uptake. CASE PRESENTATION: An extended supraclavicular lymph nodal activation after coronavirus disease 2019 (COVID-19) vaccination revealed on 18F-FDG PET/CT mimics a Virchow nodule in a patient with medical history of well-differentiated appendicular adenocarcinoma. CONCLUSION: This case highlights a nodal activation beyond axillary area and the importance of documenting vaccination history at the time of scanning to avoid false-positive results.

Reasons for delay in timely administration of adjuvant chemotherapy for patients with stage III colon cancer: a multicentre cohort study from the McGill University Department of Oncology.

PURPOSE: Adjuvant chemotherapy within 56 or 84 days following curative resection is globally accepted as the standard of care for stage III colon cancer as it has been associated with improved overall survival. Initiation of adjuvant chemotherapy within this time frame is therefore recommended by clinical practice guidelines, including the European Society for Medical Oncology. The objective of this study was to evaluate adherence to these clinical practice guidelines for patients with stage III colon cancer across the Rossy Cancer Network (RCN); a partnership of McGill University's Faculty of Medicine, McGill University Health Centre, Jewish General Hospital and St Mary's Hospital Center. PATIENTS AND METHODS: 187 patients who had been diagnosed with stage III colon cancer and received adjuvant chemotherapy within the RCN partner hospitals from 2012 to 2015 were included. Patient and treatment information was retrospectively determined by chart review. Χ2 and Wilcoxon rank-sum tests were used to measure associations and a multivariate Cox regression model was used to determine risk factors contributing to delays in administration of adjuvant chemotherapy. RESULTS: The median turnaround time between surgery and adjuvant chemotherapy was 69 days. Importantly, only 27% of patients met the 56-day target, and 71% met the 84-day target. Increasing age, having more than one surgical complication and being diagnosed between 2013-2014 and 2014-2015 reduced the likelihood that patients met these targets. Furthermore, delays were observed at most intervals from surgery to first adjuvant chemotherapy treatment. CONCLUSION: Our study found that within these academic hospital settings, 27% of patients met the 56-day target, and 71% met the 84-day target. Delays were associated with hospital, surgeon and patient-related factors. Initiatives in quality improvement are needed in order to improve adherence to recommended treatment guidelines for prompt administration of adjuvant chemotherapy for stage III colon cancer.

[18F]FDG PET radiomics to predict disease-free survival in cervical cancer: a multi-scanner/center study with external validation.

PURPOSE: To test the performances of native and tumour to liver ratio (TLR) radiomic features extracted from pre-treatment 2-[18F] fluoro-2-deoxy-D-glucose ([18F]FDG) PET/CT and combined with machine learning (ML) for predicting cancer recurrence in patients with locally advanced cervical cancer (LACC). METHODS: One hundred fifty-eight patients with LACC from multiple centers were retrospectively included in the study. Tumours were segmented using the Fuzzy Local Adaptive Bayesian (FLAB) algorithm. Radiomic features were extracted from the tumours and from regions drawn over the normal liver. Cox proportional hazard model was used to test statistical significance of clinical and radiomic features. Fivefold cross validation was used to tune the number of features. Seven different feature selection methods and four classifiers were tested. The models with the selected features were trained using bootstrapping and tested in data from each scanner independently. Reproducibility of radiomics features, clinical data added value and effect of ComBat-based harmonisation were evaluated across scanners. RESULTS: After a median follow-up of 23 months, 29% of the patients recurred. No individual radiomic or clinical features were significantly associated with cancer recurrence. The best model was obtained using 10 TLR features combined with clinical information. The area under the curve (AUC), F1-score, precision and recall were respectively 0.78 (0.67-0.88), 0.49 (0.25-0.67), 0.42 (0.25-0.60) and 0.63 (0.20-0.80). ComBat did not improve the predictive performance of the best models. Both the TLR and the native models performance varied across scanners used in the test set. CONCLUSION: [18F]FDG PET radiomic features combined with ML add relevant information to the standard clinical parameters in terms of LACC patient's outcome but remain subject to variability across PET/CT devices.

Convolutional neural networks for PET functional volume fully automatic segmentation: development and validation in a multi-center setting.

PURPOSE: In this work, we addressed fully automatic determination of tumor functional uptake from positron emission tomography (PET) images without relying on other image modalities or additional prior constraints, in the context of multicenter images with heterogeneous characteristics. METHODS: In cervical cancer, an additional challenge is the location of the tumor uptake near or even stuck to the bladder. PET datasets of 232 patients from five institutions were exploited. To avoid unreliable manual delineations, the ground truth was generated with a semi-automated approach: a volume containing the tumor and excluding the bladder was first manually determined, then a well-validated, semi-automated approach relying on the Fuzzy locally Adaptive Bayesian (FLAB) algorithm was applied to generate the ground truth. Our model built on the U-Net architecture incorporates residual blocks with concurrent spatial squeeze and excitation modules, as well as learnable non-linear downsampling and upsampling blocks. Experiments relied on cross-validation (four institutions for training and validation, and the fifth for testing). RESULTS: The model achieved good Dice similarity coefficient (DSC) with little variability across institutions (0.80 ± 0.03), with higher recall (0.90 ± 0.05) than precision (0.75 ± 0.05) and improved results over the standard U-Net (DSC 0.77 ± 0.05, recall 0.87 ± 0.02, precision 0.74 ± 0.08). Both vastly outperformed a fixed threshold at 40% of SUVmax (DSC 0.33 ± 0.15, recall 0.52 ± 0.17, precision 0.30 ± 0.16). In all cases, the model could determine the tumor uptake without including the bladder. Neither shape priors nor anatomical information was required to achieve efficient training. CONCLUSION: The proposed method could facilitate the deployment of a fully automated radiomics pipeline in such a challenging multicenter context.