RPS 1012b - Advanced imaging in paediatric cancers

RPS 1012b-K
Keynote lecture
RPS 1012b-1
The effect of time-of-flight on reducing the injected [18F] FDG activity for whole-body PET/CT imaging of paediatric oncology patients
Purpose: To assess the effect of time-of-flight (TOF) on image quality at reduced count levels for whole-body [18F] FDG-PET/CT studies of paediatric oncology patients.
Methods: 29 paediatric oncology patients (12F/17M, (12±5)-y/o, 13?BMI?28) who underwent routine whole-body PET/CT examinations on a Siemens Biograph mCT system with TOF capability (500ps) were included. The mean activity concentration was (3.8±08) MBq/kg. Events were randomly removed from the LM data to simulate reduced levels of [18F] FDG activity. All data was reconstructed using the vendor e7-tools with standard iterative image reconstruction OSEM, OSEM plus resolution recovery (PSF), and with and without TOF information. A 5 mm FWHM Gaussian post-filter was applied to all reconstructions. The reconstructed images were evaluated for noise, signal-to-noise-ratio (SNR) (liver), contrast-to-noise-ratio (CNR) (all lesions), and the SNR and CNR gains with TOF were calculated.
Results: When using OSEM, the mean noise level was 11% (6%-18%) and when adding the TOF information, the noise was reduced to 9% (6%-15%). For PSF reconstructions, the lowest noise level was obtained with PSF TOF 8%. At 50% counts, the PSF TOF mean noise level of 10% was close to 11% for the 100% counts OSEM images. A mean gain in the SNR and CNR of 1.2-1.3 was obtained when the TOF information was included. Assuming that image noise levels below 10% are acceptable for clinical work-up, VOI analysis of the liver indicated this level be achieved at 50% counts or more when using PSF TOF.
Conclusion: A 50% dose reduction potential in paediatric PET/CT is possible when using a 10% acceptable threshold on noise levels and iterative reconstruction with PSF TOF. This quantitative analysis will be complemented by the qualitative grading of image quality by clinical experts.
Limitations: n/a
Ethics: Ethics approval 2019/ETH00138.
Funding: Supported by the Austria-FWF Project I3451-N32.
RPS 1012b-2
Oncologic imaging and H2020: the PRIMAGE project helping childhood cancer research with artificial intelligence
Purpose: Digital clinical data (imaging, pathology, genomic analytics, and wearable sensors) and patient electronic records (clinical profiling, treatment, and endpoints) are key enabling factors in clinical practice. This change is promoting clinical innovation models via real-world data-driven inferences. PRIMAGE, a 4-year European Commission-financed project having 16 European partners, is one of the more ambitious medical imaging research projects dealing with artificial intelligence in the paediatric cancer environment: neuroblastoma and diffuse intrinsic pontine glioma (DIPG).
The aim was to construct as an observational in silico study involving high-quality repository with anonymised data (imaging, clinical, molecular, and genetics) for the training of machine learning and multiscale algorithms to finally have a clinical decision support (CDS) tool.
Methods: The use of computational imaging allows the extraction of multiparametric data, multiscale models, visual analytics, and artificial intelligence, leading to a new era in radiomics, characterised by high-throughput extraction, storage, and analysis of a large amount of quantitative imaging features and parameters (imaging biomarkers).
Results: The new platform will able to provide quantitative relevant information (virtual biopsies) for early disease diagnosis, disease phenotyping, disease grading, targeting therapies, and the evaluation of disease response to treatment in children with neuroblastoma and DIPG.
Conclusion: PRIMAGE?s CDS system, offered as an open cloud-based platform, will provide precise clinical assistance for phenotyping, treatment allocation, and patient endpoint estimations based on imaging biomarkers, tumour growth simulation, advanced visualisation, and machine-learning approaches. Final results will be available for the scientific community at the end of the project, ready for translation to other malignant solid tumours.
Limitations: GDPR limitations on patients recruitment will be commented.
Ethics: The project has the ethics committee approval of the coordinator hospital and reserach institute.
Funding: H2020-SC1-DTH-2018-1 GA 826494.
RPS 1012b-3
Computer-aided detection (CAD) of pulmonary nodules in paediatric ultra-low-dose chest CT: a performance analysis
Purpose: To evaluate the dose-dependency of a computer-aided detection (CAD) system for identifying pulmonary nodules in paediatric ultra-low-dose chest CTs (ULDCT).
Methods: Two scan protocols for 100kV ULDCT using tin filtration differing in the reference mAs (ref.mAs) setup for automatic tube current modulation were compared.
146 consecutive scans of paediatric patients who underwent ULDCT were included and divided into 2 study groups: 48 patients (11.1±5.5 years) underwent ULDCT with 30 ref.mAs while 98 patients (13.5±5.5 years) were scanned using 96 ref.mAs.
Patients with inflammatory consolidations were excluded. Each scan was assessed for pulmonary nodules by two radiologists and each lesion was categorised by size (2-3 mm vs ?3 mm). Each nodule marked by the CAD system was validated in consensus. The radiation dose was estimated in size-specific dose estimates (SSDE).
Results: Radiation dose in the 30 ref.mAs group was significantly lower (0.16±0.09 mGy) than in the 96 ref.mAs group (0.56±0.22 mGy, p<0.001).
There was no significant difference between the groups regarding total CAD errors (mean per scan 1.44±3.50 vs 1.37±2.24 for NS ?2 mm and 0.85±2.24 vs 0.71±1.59 for NS ?3 mm) resulting in resembling CAD sensitivities (0.43 vs 0.45 in nodule size (NS) ? 2mm; 0.65 vs 0.74 in NS ?3 mm).
There was no significant correlation between SSDE and CAD errors for NS ?2 mm (r=-0.054) and NS ? 3mm (r=-0.067).
Conclusion: Lowering the parameters for automatic tube current modulation in paediatric ULDCT results in significant dose reduction without compromising the evaluated CAD system.
Limitations: CAD sensitivity and specificity in this oncology-dominant study group was not satisfying in both scan protocols assessed.
Ethics: n/a
Funding: The project underlying this report was funded by the Deutsche Forschungsgemeinschaft (DFG), project number VE1008/1-1 and KO 2938/5-1.
RPS 1012b-5
The value of diffusion-weighted MRI in the response assessment of nephroblastoma
Purpose: To assess the value of diffusion-weighted MRI (DW-MRI) in the prediction of viable blastemal remnant after neoadjuvant chemotherapy in patients with nephroblastoma, which is considered a poor prognostic marker and may result in additional adjuvant treatment.
Methods: This IRB- approved study included 37 paediatric patients who underwent DW-MRI prior and after completion of neoadjuvant chemotherapy for nephroblastoma and subsequent surgical resection. Two blinded radiologists volumetrically assessed each tumour and the parameters mean ADC, median ADC, 12.5th/25th/75th percentile, skewness, and kurtosis were calculated. The associations between the imaging features and blastemal remnant fraction were examined using over-dispersed Poisson regression.
Results: Inter-reader agreement was high for mean ADC, volume, skewness, and kurtosis (ICC: 0.74-0.96). In univariable analysis, pre-therapeutic ADC mean, median, 12.5th percentile, and skewness were significantly associated with the presence of blastemal remnant for reader 1 (p=0.009-0.044) and pre-therapeutic ADC mean, median, 12.5th/25th/75th percentile, and skewness for reader 2 (p=0.009-0.026). In a multivariable analysis, the ADC mean was significantly associated with blastemal remnant for both readers (reader 1: fraction ratio 1.33 (1.07, 1.65), p=0.014; reader 2: fraction ratio 1.1 (1.01, 1.2), p=0.034). Post-therapeutic MRI parameters were not significantly associated with blastemal remnant.
Conclusion: A higher pre-therapeutic ADC was significantly associated with a larger fraction of a blastemal remnant after neoadjuvant chemotherapy. This could allow for a more personalised chemotherapeutic regime in these patients and offer predictive information at the time of the initial diagnosis.
Limitations: The retrospective study design.
Ethics: IRB approved.
Funding: No funding was received for this work.
RPS 1012b-6
Radiomic features as a marker of metastatic spread in Wilms? tumours
Purpose: To assess the role of radiomic analyses for characterising metastatic and localised paediatric renal Wilms? tumours.
Methods: Paediatric patients affected by Wilms? tumours referred to our tertiary centre for staging (i.e. treatment naïve) from 2012-2018 and examined by contrast-enhanced computed tomography were included in this retrospective study. Patients matching the inclusion criteria were then subdivided according to the presence of metastases (i.e. mW and lW). One radiologist expert in paediatric imaging drew regions of interest along the margins of all primary tumours covering the entire volume using open-source software (3D Slicer). The same software was applied to extract 33 radiomic features from each patient belonging to three categories: first-order statistics, grey-level co-occurrence matrix (GLCM), and the grey-level run-length matrix (GLRLM). The Student?s t-test was used to compare mW and lW patients for each radiomic feature (p<0.05). The accuracy of the variables showing a statistically significant difference was assessed by receiver operating characteristic curves.
Results: 32 patients (16 females, mean age±SD, 4±2.27 yrs) met the inclusion criteria and were analysed in this study. 10 patients were affected by metastatic disease (i.e. mW) and 22 by localised tumours (i.e. lW). Two features (one FOS, one GLRLM) showed a statistically significant difference between mW and lW: variance (p=0.043) and grey-level non-uniformity normalised (GLNUN; p=0.008). The area under the curve was good for GLNUN (AUC=0.805) and fair for the variance (AUC=0.655)
Conclusion: GLNUN can be considered as a robust radiomic marker of metastatic spread in Wilms tumours.
Limitations: Future studies addressing the clinical implications and considering the impact on the therapeutic treatment of our findings should be performed.
Ethics: Ethics committee approval obtained.
Funding: No funding was received for this work.
RPS 1012b-7
The feasibility and value of quantitative semi-automated diffusion-weighted imaging volumetry of neuroblastic tumours
Purpose: To assess the feasibility and value of semi-automated diffusion-weighted imaging (DWI) volumetry of whole neuroblastic tumours with an apparent diffusion coefficient (ADC) map evaluation after neoadjuvant chemotherapy.
Methods: Paediatric patients who underwent surgical resection of neuroblastic tumours at our institution from 2013-2019, and who received a preoperative MRI scan with DWI after chemotherapy, were included. Tumour volumetry was assessed with a semi-automated approach in DWI using a dedicated software prototype. Quantitative ADC values were calculated automatically of the total tumour volume after manual exclusion of necrosis. Manual segmentation in T1-weighted and T2-weighted sequences was used as a reference standard for tumour volume comparison.
Results: 27 patients with 28 lesions (neuroblastoma (NB): n=19, ganglioneuroblastoma (GNB): n=7, and ganglioneuroma (GN): n= 2) could be evaluated. The mean patient age was 4.5±3.2 years. The median volume of standard volumetry (T1w or T2w) was 50.2 ml (interquartile range (IQR): 91.9 ml) vs 45.1 ml (IQR: 98.4 ml) of DWI (p=0.145). Mean ADC values (x10-6 mm2/s) of the total tumour volume (without necrosis) were 1,187±301 in NB vs 1,552±114 in GNB/GN (p=0.037). The 5th percentile of ADC values of NB (614±275) and GNB/GN (1,053±362) provided the most significant difference (p=0.007) with an area under the curve of 0.848 (p<0.001).
Conclusion: Quantitative semi-automated DWI volumetry is feasible in neuroblastic tumours with an integrated analysis of tissue characteristics by providing automatically calculated ADC values of the whole tumour as well as an ADC heatmap.
Limitations: Although the number of subjects is small, it represents one of the largest published cohorts with ADC analysis of neuroblastic tumours due to the low incidence of this disease.
Ethics: IRB approval obtained. Written informed consent waived.
Funding: No funding was received for this work.
RPS 1012b-9
Dynamic contrast-enhanced perfusion MRI and diffusion-weighted imaging as an imaging biomarker for paediatric cancer
Purpose: Neuroblastomas are the most frequent solid extracranial cancer in childhood. Its diagnosis, prognosis, and monitoring are based on the information provided by multiparametric magnetic resonance images. Our focus is to explore the utility of diffusion and perfusion changes in neuroblastomas as an early biomarker of diagnosis, using diffusion-weighted and dynamic contrast-enhanced MRIs.
Methods: Multiple MR imaging real-word sequences were used from 30 patients. Volumes-of-interest were calculated and transferred to DCE perfusion and apparent diffusion coefficient (ADC) maps. Histogram analysis and clustering unsupervised ML algorithms were used to determine the values of the mean and standard deviation of the initial area under the curve at 60 seconds (IAUC60) and ADC for automatic differentiation of neuroblastic tumours. The resolution was estimated and the data was smeared accordingly to identify and remove the noise and low-quality voxels.
Results: Significant differences in the mean ADC were found for neuroblastic tumours: 1.0 for ganglioneuroma, 0.82 for ganglioneuroblastoma, and 0.52 for neuroblastoma, with an uncertainty of 0.11%, 42%, and 16%. This result improves tumour differentiation with respect to state-of-the-art voxel-by-voxel methodologies, which were found to be: 1.6 for ganglioneuroma, 1.7 for ganglioneuroblastoma, and 1.3 for neuroblastoma, with an uncertainty of 3.6%, 12%, and 17%. THe mean IAUC60 was found to have a value of 43 (and 14% uncertainty) for neuroblastoma, as opposed to a value of 17 (and 17% uncertainty) with state-of-the-art voxel-by-voxel methodologies.
Conclusion: The proposed novel technique to determine IAUC60 and ADC parameters holds promise for differentiating benign and malignant neuroblastic tumours.
Limitations: The dataset size and harmonisation among different machines and data collection techniques.
Ethics: La Fe Hospital received approval from the hospital's ethics committee.
Funding: Horizon2020. Topic:SC1-DTH-07-2018-RIA. GA:826494.
RPS 1012b-10
The differentiation of low- and high-grade paediatric brain tumours by using intravoxel incoherent motion imaging and diffusion kurtosis imaging (recorded)
Purpose: To demonstrate that a new set of parameters from IVIM and DKI can be used to improve the accuracy of MR imaging for differentiating among low- and high-grade paediatric brain tumours.
Methods: 49 paediatric patients with histologically-proved brain tumours who underwent IVIM and DKI were recruited in this study. The mean, minimum, maximum IVIM [pure diffusion coefficient (D), pseudo-diffusion coefficient (D*), and perfusion fraction (f)], and DKI [diffusion kurtosis (K) and diffusion coefficient (Dk)] parameters were measured. The IVIM and DKI values were measured in solid tumour regions and in normal-appearing grey matter as a control. These values were compared between the low-and high-grade paediatric brain tumours by using the Mann-Whitney U test. Receiver-operating characteristic (ROC) analysis and logistic regression analysis were performed to evaluate the diagnostic performance of single-parametric and multiparametric models.
Results: None of the IVIM and DKI parameters exhibited significant differences in normal-appearing grey matter (P>0.05). The Dk and D values were lower, whereas the K and fmin values were higher in high-grade paediatric brain tumours than those in low-grade paediatric brain tumours (all p<0.05). The combination of DKmin and Kmax provided the largest area under the ROC curve (0.955) in the ROC analysis compared with individual parameters (Dmin 0.891, DKmin 0.933, Kmax 0.923, and fmin 0.734), indicating an improved diagnostic performance for tumour grading.
Conclusion: The parameters derived from IVIM and DKI can be used to distinguish low-grade paediatric brain tumours from high-grade paediatric brain tumours. The combination of DKI parameters may serve as non-invasive and quantitative imaging parameters for grading paediatric brain tumours in vivo.
Limitations: A retrospective study covering a relatively small sample of subjects.
Ethics: n/a
Funding: No funding was received for this work.
RPS 1012b-11
Post-transplant thymus restoration is a really good predictor of transplant outcomes?
Purpose: To find a correlation between post-transplant thymic volume and thymopoiesis, to associate thymic renewal with early and long-term transplant outcomes, and to expand the role of the radiology team in transplant clinical decisions and the post-transplant management of paediatric patients.
Methods: Thymic volume assessment and thymocytes analysis were performed before HSCT (baseline volume), at 3, 6, 12 months, and long-term after HSCT.
74 paediatric patients who underwent an allo-HSCT from 2002-2018 were included. The control group consisted of 311 paediatric patients undergoing chest MRI for orthopaedic reasons.
The analysis was performed with HOROS software. Manual thymus tracing was performed for each cut, with the generation of a 3D reconstruction and a volume calculation.
Statistical analyses were performed using WinStat and MedCalc.
Results: There is a statistically significant correlation between thymus volume and thymopoiesis with p<0.0001 and r=0.5720.
The average thymic volume at all evaluations is significantly greater in the subgroup without GVHD: 16.1 (±10.0) cm3 vs 11.2 (±8.7) cm3; p<0.05.
There is a significant difference in the average thymic volume between patients who survived after the HSCT (68.8 ± 48.46) and patients who did not (2.9 ± 2.11); p<0.0001.
Conclusion: The restoration of thymic volume is a good indicator of positive outcomes in allogeneic HSCT, while a restoration failure is related to transplant-related mortality.
Thymic MRI assessment is now part of the current clinical management of transplanted patients and improves their integrated care pathways.
The next aim will be to create specific sequences that allow us to analyse the microenvironment of thymic tissue (?virtual biopsy?).
Limitations: A monocentric and retrospective study with a limited sample of subjects.
Currently, our MRI protocol allows us to study only the thymic morphology but not the microenvironment of the thymic tissue.
Ethics: n/a
Funding: No funding was received for this work.
RPS 1012b-12
Ultrasound-guided core biopsy with smear cytology: the fastest route for diagnosis in paediatric oncology
Purpose: In our hospital, ultrasound-guided fine-needle biopsy (US-guided FNB) is the first choice for tissue diagnosis in the paediatric population. In 2018, we added smear cytology in some of our biopsies, allowing for an immediate primary diagnosis. We retrospectively reviewed our experience with FNB assessing the accuracy rate, safety, and availability of the procedure, as well as the accuracy of the smear cytology and its added value for patient management.
Methods: Paediatric ultrasound-guided biopsies done in our hospital from 01/2018-09/2019 were studied. Data collection included demographics, clinical and procedural data, and follow-up.
Results: 106 biopsies were performed on 91 patients, 15 of them on known oncologic patients. 96.2% of biopsies were performed within 36 hours. 79 tumours were correctly diagnosed and 1 malignancy was misdiagnosed as a benign lesion. 23 biopsies were correctly diagnosed as non-tumour. Smear cytology was performed in 30 cases; 25 tumours and 5 reactive lymph nodes. The cytologist correctly differentiated the tumour from inflammation in all cases and diagnosed the tumour in 24 of the cases. The sensitivity of ultrasound-guided FNB is 98.7%, specificity 100%, and accuracy 99%. The accuracy of smear cytology for differentiating tumours from non-tumours is 100% and for final diagnosis 97%.
Conclusion: We find ultrasound-guided FNB for suspected malignancy in the paediatric population highly available, safe, and accurate. In our small cohort, smear cytology was an excellent, real-time tool for differentiating tumour and non-tumour tissue, and in most cases allowed for early correct tumour diagnosis. This procedure may accelerate patient management and improve patient care.
Limitations: A retrospective study with a small cohort.
Ethics: All biopsies were done under informed consent. The local Helsinki committee approved the study.
Funding: No funding was received for this work.


Maria Carla Calcagno (Italy)

Annemieke Simone Littooij (Netherlands)

European Society of Radiology

Watch ECR 2020 live

This session is part of ECR 2020 Live. Please register for ECR 2020 Live in order to get access.

    WITHOUT CME €39.00
    FROM HOME €350.00