Research Presentation Session: Physics in Medical Imaging

RPS 1713 - Optimisation and its tools

March 2, 08:00 - 09:00 CET

7 min
CT acquisition parameter selection in the real world: impacts on paediatric radiation dose and variation amongst 155 institutions
Yifei Wang, Sunnyvale / United States
    Author Block: Y. Wang, P. Chu, C. Stewart, R. Smith-Bindman; San Francisco, CA/USPurpose: Computed tomography (CT) exams have expanded rapidly since introduction, such that theoretical understandings of best dose optimisation practices and real-world applications are incongruent. This remains the case even in children, where the limitation of dose is even more important. We seek to quantify how often parameters are adjusted in real-world practice and their degree of contribution to real-world dose distribution. We identify discrepancies between parameters that are impactful in theory and impactful in practice.Methods or Background: This study analyses 25,000 consecutive children routine abdomen exams performed between November 2015 and Jan 2021 in the UCSF International CT Dose Registry of 155 institutions. We calculated geometric standard deviation (gSD) for five parameters (kV, mAs, spiral pitch, number of phases, scan length) to assess variation in practice. A Gaussian mixed regression model was performed to predict the radiation dose-length product (DLP) using the parameters. To reflect the theoretical impact, we predict the increase in DLP per 10% increase in the parameter. To reflect the real-world practical impact, we predict the increase in DLP per gSD increase in the parameter.Results or Findings: Among studied examinations, mAs, number of phases, and scan length were frequently manipulated (gSD
  1. 47–1.79); kV was rarely manipulated (gSD 1.12). Theoretically, kV is the most impactful parameter (31% increase in DLP per 10% increase in kV, 8–11% for other parameters). In real-world practice, kV is less impactful; for each gSD increase in kV, the DLP increases by 40%. This value is highly surpassed by the analogous number in mAs (68%), while being comparable to other technical parameters.
  2. Conclusion: Despite the potential impact of kV on radiation dose, this parameter is rarely manipulated in common practice and its potential remains untapped.Limitations: This study uses mainly US Data.Funding for this study: The registry was funded by NIH and PCORI.Has your study been approved by an ethics committee? YesEthics committee - additional information: The institutional review board of the primary responsible university approved this research, and collaborating institutions obtained local approval or relied on the primary site's approval.
7 min
Establishing indication-based diagnostic reference levels for paediatric computed tomography and international comparison: data from an international registry
Denise Bos, Essen / Germany
    Author Block: D. Bos1, Y. Wang2, C. Stewart2, S. Zensen1, J. Luong2, P. Chu2, R. Smith-Bindman2; 1Essen/DE, 2San Francisco, CA/USPurpose: Computed tomography (CT) scans are essential for diagnosing paediatric patients, especially in emergencies. However, concerns arise about the potential carcinogenic risk associated with radiation exposure. CT radiation doses can vary widely depending on medical indications, protocols, and local practice. Our aim is to establish diagnostic reference levels (DRLs) based on clinical indications in paediatric patients.Methods or Background: A comprehensive analysis was performed using CT data from an international dose registry that includes CT scans performed in children under 18 years at more than 150 sites between January 2016 and January
  1. DRLs, defined as the 75th percentile, were calculated for volume-weighted CT dose index (CTDIvol) and dose-length product (DLP) across 15 broad CT categories, which reflect both the anatomic areas and radiation doses (low-dose [LD], routine-dose [RD], high-dose [HD]) required by the underlying imaging indications. We compared RD categories for head, chest, and abdomen/pelvis scans between facilities in the United States (US) and Europe.
  2. Results or Findings: A total of 95,047 CT scans (54% male, 45% female) from 41 different indications were included in the analysis. The DRLs of DLP and CTDIvol mainly increased significantly with increasing age group (p<
  3. 05). For head scans in 10- to 14-year-old children, the DRLs for DLP varied from 362 mGy·cm (LD) to 734 mGy·cm (RD) to 2,058 mGy·cm (HD). US DRLs were significantly higher than European DRLs for RD chest and abdomen/pelvis. Differences were inconsistent for RD head.
  4. Conclusion: Optimising radiation dose for paediatric patients presents significant challenges due to variations based on age, size, clinical indication and protocol. Therefore, we have established indication-based DRLs for different CT categories and age groups to minimise excessive radiation doses and standardise practice.Limitations: An identified limitation is that the registry contains mainly US data.Funding for this study: The registry was funded by NIH and PCORI.Has your study been approved by an ethics committee? YesEthics committee - additional information: The institutional review board of the primary responsible university approved this research, and collaborating institutions obtained local approval or relied on the primary site’s approval.
7 min
Automated dose and noise assessment as a guide for harmonisation of abdominal CT protocols across multiple scanners
Janne Vignero, Leuven / Belgium
    Author Block: J. Vignero1, N. Fitousi1, B. Miseur1, J. Binst1, K. Lemmens1, V. Nuttens2, H. Bosmans1; 1Leuven/BE, 2Aalst/BEPurpose: In this study a systematic approach is proposed for performance evaluation and, if necessary, harmonisation of abdominal computed tomography (CT) protocols across vendors using dose and noise metrics.Methods or Background: CTDIvol, water equivalent diameter (WED), global noise level (GNL) and metadata of 4,933 abdominal CT exams from six scanners were calculated by DOSE (Qaelum). Paediatric patients (<15 years), hard kernel reconstructions and scans with substantial image truncation were excluded. Patients were categorised into five size groups based on WED. Subgroups were defined based on protocol, scanner, kVp and kernel. Local working points (LWP) for CTDIvol and GNL were calculated from the median values for every subgroup. Reference working points (RWPs) were determined for each size group based on the median CTDIvol and GNL across all subgroups. The performance of each subgroup was assessed by comparing LWP to RWP.Results or Findings: Patient size groups ranged between 200 mm and 422 mm, with corresponding increases in dose and noise RWP from 5 to 16 mGy and
  1. 9 to 12.5 HU, respectively. Variations in LWP among subgroups were large. The GNL with iterative reconstruction was significantly (p=0.0) lower than using filtered back projection. Some subgroups appeared to be overdosing or requiring higher image quality, while others might benefit from higher radiation doses. One scanner with substantially higher dose and noise LWP than RWP was identified as underperforming. Suggested actions were confirmed in a team comprising of radiologists, a radiographer and a physicist.
  2. Conclusion: Combined evaluation of automated image noise evaluation, radiation dose and patient size points to clinically relevant adjustments of the scan protocols and was successfully applied in a multi-vendor CT scanner scenario.Limitations: No limitations were identified.Funding for this study: No funding was provided for this study.Has your study been approved by an ethics committee? Not applicableEthics committee - additional information: No information provided by the submitter.
7 min
Clinical CT protocol comparison: an Italian multicentre study
Raffaele Villa, Monza / Italy
Author Block: R. Villa1, M. Daniotti1, C. Ingraito1, A. Italian Digital Radiology Working Group2, N. Paruccini1; 1Monza/IT, 2Milan/IT
Purpose: In CT imaging, differences in manufacturers, models, scanning parameters, reconstruction algorithms, clinical tasks, and user’s experience, introduce variability in the trade-off between diagnostic information and patient’s radiation exposure. The extent to which each of these sources of variabilities impacts the overall procedure performance is still unknown. This study aimed to compare image quality and radiation dose in five acquisition protocols to ascertain how different technical features can affect CT performance, to inform protocol optimisation and design.
Methods or Background: Five different acquisition protocols to investigate five different clinical tasks were investigated: intracranial haemorrhage, pulmonary embolism, acute abdomen, paediatric chest and paediatric acute abdomen.
The Catphan 600 (The Phantom Laboratory, USA), was imaged in 34 CT scanners from 20 hospitals across Italy, for a total of 251 datasets, using the locally optimised clinical protocols. The Catphan body external-annulus was used to assess the effects of tube current modulation for the different manufacturers.
Image quality was quantified in terms of the detectability index using an object contrast ranging 25-700 HU and lesion size ranging 5-25 mm, representative of clinical scenarios related to the examined protocols.
Results or Findings: Each investigated protocol showed differences between manufacturers in terms of dose levels and image quality values. The reconstruction algorithm’s generation proved another driving element for image quality. Nevertheless, variability between CTs was found, even from the same vendor and technology.
Conclusion: This study suggested that manufacturer and reconstruction algorithms play a main role in the optimal trade-off between diagnostic information and radiation dose levels. Furthermore, the other examined CT procedure technical features, have a non-negligible impact. All the potential sources of variability should be included when designing optimization actions in clinical CT.
Limitations: Data numerosity should be increased to improve the study strength, especially for paediatric exams.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: No information provided by the submitter.
7 min
Implementation of the German lung cancer screening protocol and image quality evaluation over three generations of CT systems
Mishal Ursani, Heidelberg / Germany
    Author Block: M. Ursani1, T. D. Do1, H-U. Kauczor1, H-P. Schlemmer1, T. Lasser2, S. O. Schönberg3, S. Sawall1, M. Kachelrieß1; 1Heidelberg/DE, 2Garching/DE, 3Mannheim/DEPurpose: To develop methods for the implementation of lung cancer screening using low-dose CT according to the German “Bundesamt für Strahlenschutz” (BfS) and to evaluate the resulting image quality in three generations of CT systems.Methods or Background: Experiments were performed using a thorax phantom (QRM, Germany) in three generations of CT systems (Somatom Flash/Force, Naeotom Alpha, Siemens Healthineers, Germany). The phantom and imaging protocols were tailored to meet BfS requirements for lung cancer screening. The phantom featured 1 cm sized spherical lung nodules with a contrast of 150 HU and fat extension rings mimicking different patient sizes, namely S (20×30 cm), M (25×35 cm), and L (30×40 cm). Acquisitions were performed at
  1. 3 mGy as defined as the maximum dose for a standard patient (BMI 26 kg/m²), employing patient-specific prefilters, if available. Images were reconstructed iteratively using the smoothest kernel ensuring a spatial resolution with a FWHM≤1 mm in both axial and longitudinal direction. Hence, resolution is only approximately matched among systems. Contrast-to-noise ratio (CNR) was used to assess the visibility of lung nodules in all images.
  2. Results or Findings: A dose of
  3. 3 mGy was achieved using tube settings of 20 mAs/120 kV, 306 mAs/100 kV+Sn, and 150 mAs/100 kV+Sn for Flash, Force and Alpha. Image noise varied between 115 HU to 330 HU over all systems and phantom sizes. CNR (S to L) ranged from 1.30 to 0.36, 1.03 to 0.65, and 1.76 to 0.71 for Flash, Force, and Alpha.
  4. Conclusion: A wide range of CT systems are capable of providing an image quality sufficient for lung cancer screening examinations. However, larger patients might require an increase in dose or slice thickness to account for the high image noise.Limitations: The study is limited to phantom acquisitions.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? Not applicableEthics committee - additional information: No information was provided by the submitter.
7 min
Physics evaluation of a super resolution deep learning reconstruction algorithm in an anatomical background
Kirsten Boedeker, Los Angeles / United States
    Author Block: K. Boedeker, D. Shin, N. Akino, D. Mather; Otawara-shi, Tochigi/JPPurpose: Deep learning reconstruction algorithms for computed tomography (CT) are generally trained on anatomical data. Therefore, the purpose of this work is to characterise a super resolution deep learning reconstruction algorithm (DLR) compared to traditional hybrid iterative reconstruction (HIR) and filtered back projection (FBP) in terms of standard quantitative noise and spatial resolution measures, with the test objects inserted in both an anatomical background as well as in uniform phantom background.Methods or Background: A commercial super resolution DLR was assessed for cardiac mode. The modulation transfer function (MTF) was first measured using a standard edge-based approach on images acquired from the Catphan 600 with uniform background and reconstructed with DLR, HIR, and FBP. The cylindrical test objects used for assessment were next simulated via forward projection using a system model and inserted into raw patient data. The raw patient data was then reconstructed with DLR, HIR, and FBP. The MTF and standard deviation of noise were then measured. A 100 mm FOV was used for measurement. The Pearson correlation coefficient was determined for each reconstruction algorithm’s MTF in patient versus phantom.Results or Findings: The FBP MTF values in the anatomical background versus uniform phantom were within 10%. In both phantom and the anatomical background, noise reduction for DLR versus HIR was approximately 45-50%. The 10% of MTF was
  1. 6p/cm with DLR versus 8.4lp/cm with HIR in anatomical background. The Pearson correlation in anatomical background versus phantom was greater than 0.99 for all three reconstruction algorithms.
  2. Conclusion: Traditional in-plane MTF applied in an anatomical background demonstrates significant improvement over HIR and correlates well with phantom measurements.Limitations: An identified limitation is that a limited dose range was analysed.Funding for this study: Funding was received from Canon Medical System Corporation.Has your study been approved by an ethics committee? Not applicableEthics committee - additional information: No information provided by the submitter.
7 min
Influence of foreign bodies on image quality in dark-field chest radiography
Lennard Kaster, Munich / Germany
    Author Block: L. Kaster1, S. Karl1, T. Urban1, F. Gassert1, R. C. Schick1, T. Koehler2, A. Sauter1, F. Pfeiffer1, D. Pfeiffer1; 1Munich/DE, 2Hamburg/DEPurpose: Dark-field radiography is a novel x-ray method that visualises microstructural properties of the examined object. Our pioneering studies, the first conducted on humans worldwide, have demonstrated that this method outperforms conventional radiographs in diagnosing and staging pulmonary diseases, like COPD and Covid-
  1. It is known from conventional x-ray radiographs and CT that foreign bodies, such as metal implants, can lead to artefacts and thus impair image quality. Therefore, this research evaluates the influence of foreign bodies on x-ray dark-field chest radiographs.
  2. Methods or Background: Experiments were conducted using a clinical dark-field chest radiography prototype, including investigating patient images. The "LUNGMAN" thorax phantom was used and measurements were performed with and without foreign bodies like breast implants, medical compresses and a pacemaker unit. The impact of these foreign bodies on image quality was analysed.Results or Findings: The results demonstrated that the impact of a foreign body on image quality is contingent upon its material composition. Homogeneous materials, such as breast implants, do not contribute additional dark-field signals. Conversely, materials possessing a microstructure, exemplified by medical compresses, generate their own dark-field signals. Notably, these signals do not influence the surrounding areas.In the case of a pacemaker unit, despite the absence of a microstructure, a dark-field signal can be observed. Furthermore, areas adjacent to the foreign body are also affected.For both medical compresses and the pacemaker unit, the dark-field signal emanating from tissue within the same radiation path cannot be reliably evaluated.Conclusion: Foreign bodies have the potential to generate a substantial true or artificial dark-field signal, which can significantly impact the image quality and interpretability of dark-field radiographs. Therefore, it is crucial to consider these potential artefacts to ensure an accurate diagnosisLimitations: No limitations were identified.Funding for this study: Funding was received from the Federal Ministry of Education and Research (BMBF) and the Free State of Bavaria under the Excellence Strategy of the Federal Government and the Länder, as well as by the Technical University of Munich – Institute for Advanced Study.Has your study been approved by an ethics committee? YesEthics committee - additional information: The study was approved by the Ethics Commission of the Medical Faculty, Technical University of Munich, Germany (reference number: 166/20S).