24th CT users group meeting: 6/10/2022
The 24th CT Users Group meeting was held at Trent Vineyard, Nottingham, on 6th October 2022. The programme is shown below, links to some of the presentations will be available soon.
The day before (5th October), we ran our first one-day training course - described as “An introduction to the physics of computed tomography, including how images are produced and factors affecting radiation dose and image quality”, check out the CTUG training course page for more details.
Please note: information provided in the slides is not peer-reviewed, is for educational use only and is explicitly not to be used for sales or marketing purposes. Any of the authors can be contacted, via the CTUG if no contact information is provided in the slides, to discuss the contents.
| « Previous meeting | – | Meetings Index | – | Next meeting » |
|---|
Meeting Programme
Session 1 - Manufacturer Developments
10:00 Current developments in Canon’s CT technology - Mark Condron, Canon Medical Systems Ltd
10:25 Current developments in GE’s CT technology - Matthew Cormell, GE Healthcare
10:50 Current developments in Philips' CT technology - Richard Andrew, Philips
11:15 The technology behind Siemens' dose saving techniques - Siemens
Session 2 - Implementing New Technology
12:10 Implementing Siemens iterative reconstruction techniques in routine CT protocols - Jamie Dormand, Louise Giansante, Elly Castellano, Ed McDonagh, James Brymer, Cate Savidge, Jonathan Frazer-McRobert, Christian Kelly-Morland - Royal Marsden Hospital
The use of iterative reconstruction techniques in CT has been growing in clinical practice in the last few years, particularly due to its dose reduction potential. Iterative reconstruction (IR) techniques use fundamentally different processes to filtered back projection (FBP) and traditional methods of characterising image quality are not directly applicable.
All diagnostic CT scanners at the Royal Marsden in Chelsea and in Sutton include the SAFIRE IR algorithm [1], but most routine CT protocols at both sites use FBP. The Royal Marsden private care centre in Cavendish Square, London, has a Siemens Somatom Drive CT scanner, which is the first in the Trust with ADMIRE IR [2].
We present the methodology and results of an optimisation task to implement ADMIRE and SAFIRE iterative reconstruction techniques on Siemens CT scanners at The Royal Marsden. MTF and NPS measurements were performed using the Catphan [3] and Siemens water phantoms to understand the general characteristics of each IR technique compared to FBP. Contrast-to-noise ratio measurements of clinically relevant features in a Kyoto PBU-60 anthropomorphic phantom [4] were performed to investigate the effects of Quality Reference mAs on resulting IR noise properties. A consultant radiologist appraised the anthropomorphic phantom images in terms of clinical image quality.
While several groups are developing novel approaches to quantitatively appraise IR algorithms [5] these typically require specific phantoms, bespoke software and measurement protocols that are not in widespread use. It is nevertheless important to implement and optimise dose saving features such as IR, while considering the uncertainties and limitations in any tests used.
The results of our tests were used to inform the creation of ADMIRE IR protocols with a 15 % reduction in quality reference mAs compared to equivalent FBP protocols. Tips on performing optimisation tasks in similar circumstances will be presented and future developments of this work will be discussed.
References:
[1] Grant, K. & Raupach, R. (2012). SAFIRE: Sinogram Affirmed Iterative Reconstruction. Siemens Healthcare White Paper.
[2] Ramirez-Giraldo, J. C., Grant, K. L., & Raupach, R. (2018). Admire: Advanced modeled iterative reconstruction. Siemens Healthcare White Paper.
[3] The Phantom Laboratory. Catphan 700. Catphan 700 – The Phantom Laboratory, accessed 06 September 2022, https://www.phantomlab.com/catphan-700
[4] Kyoto Kagaku, C. O. (2020). LTD. Whole body phantom “PBU-60.”
[5] Smith, T. B., Solomon, J. B., & Samei, E. (2017). Estimating detectability index in vivo: development and validation of an automated methodology. Journal of Medical Imaging, 5(3), 031403.12:30 Initial Experience of Canon AiCE Deep Learning Algorithm - Jane Edwards - Royal Free London NHS Foundation Trust
Aim: Artificial intelligence is becoming more prevalent within diagnostic imaging and has the ability to reduce patient dose and improve image quality. We aim to present our initial experience with the Canon Advanced Intelligent Clear-IQ Engine (AiCE).
Method: A Canon Aquilion Prime SP CT Scanner with AiCE capabilities was commissioned by the Trust in July 2021. Image quality metrics, including SNR, CNR and image noise were assessed for different protocols both with and without AiCE reconstructions in clinical images. Image quality metrics were also compared to a second Aquilion Prime scanner without AiCE capabilities.
Results of this study will be presented at this meeting.12:50 Evaluating the use of eye dose reduction technologies on a Siemens Go CT scanner and their effect on measured eye dose and image quality - Ben Grimes, Anne Hill, Teresa Lo, Juttalie Cole, Pippa Dunbar - University Hospital Bristol and Weston NHS Foundation Trust
This project evaluated the effects of organ tube current modulation (OTCM) and orbitomeatal line (OML) angulation on dose to the lens of the eye, and the effects on image quality for computed tomography (CT) head examinations using a Siemens Go CT scanner. The Siemens Go is the newest of 3 Siemens scanners at one of the UHBW sites, and the only scanner at this site that has OTCM (X-CARE). Radiologists had reported lower image quality for head scans when compared to the two older scanners. X-CARE is a key difference between the scanners; it was hypothesised that this could be the cause of the reported lower image quality.
Data was obtained by scanning an ACS angiographic head phantom under clinical protocols, using thermoluminescent dosimeters (TLDs) placed over the eyes to measure eye dose. The following combinations of dose reduction methodologies were used: OML angulation, X-CARE, both OML angulation and X-CARE, no dose reduction methods applied (baseline).
Eye dose for each combination was analysed. The noise in the resultant images was assessed using the coefficient of variation (CoV %) at three regions in the brain: the cerebellum, basal ganglia and centrum semiovale. Consultation with a Radiologist determined that these three regions have the most clinical significance for image quality analysis.
X-CARE was comparably effective at reducing eye dose compared to OML angulation. When both were used together, there was a 40% reduction in dose compared to baseline. A statistically significant reduction in image quality (15% increase in CoV) was seen at the basal ganglia when both X-CARE and OML-angulation were used. However, this was not replicated in the other two regions. OML angulation alone did not result in any significant change in image quality in any of the three regions.
This study suggests that OML angulation and OTCM used together may result in some increase in noise in some areas of the brain. Further work should be to investigate with the clinical team whether this correlates with the changes seen clinically and whether an adjustment can be made to X-CARE, as required.
In conclusion, the eye dose reduction methods tested are effective at reducing dose to the lens of the eye with some increase in image noise. X-CARE is a viable alternative to OML angulation for reducing eye dose when the latter is not achievable.Session 3 - Patient Dose and Image Quality
14:40 Comparison of patient effective doses from multiple CT examinations based on different calculation methods: An update - Simona Avramova-Cholakova[1], Iliya Dyakov[2], Hristomir Yordanov[3], James O’Sullivan[1] - [1] Imperial College Healthcare NHS Trust, London, [2] Acibadem City Clinic UMBAL, Sofia, [3] Technical University - Sofia, Sofia, Bulgaria
Aim: The aim of this study is to compare effective dose (E) estimations based on different methods for patients with recurrent computed tomography (CT) examinations.
Material and Methods: Patient data from two hospital organisations were retrospectively extracted using the dose management software (DMS) available at each site (DoseWatch or Radimetrics). Firstly, patients with recurrent CT examinations exposed to cumulative effective dose (CED) of 100 mSv and above were identified. These patients were split in three size groups, based on patient effective diameter, as provided by the DMS: small, normal (with sizes close to the median value), and high, separately for each organisation, and a total of 40 patients were selected for further analysis. Seventeen methods were used to determine the E of each phase as well as the total E of the CT examination. These included three groups of estimations: based on the use of published E, calculated from typical or patient-specific values of volume computed tomography dose index (CTDIvol) and dose-length product (DLP) multiplied by conversion coefficients, and based on patient-specific calculations or calculations on the basis of typical values with use of software.
Results and discussion: The E from a single phase of the examination varied with a ratio from 1.3 to 6.8 for small size patients, from 1.2 to 6.5 for normal size patients, and from 1.7 up to 18.1 for large size patients, depending on the calculation method used. The cumulative effective dose (CED) ratio per patient for the different size groups varied as follows: from 1.4 to 2.5 (small), from 1.7 to 4.3 (normal), and from 2.2 up to 6.3 (large). The minimum CED across patients varied from 38 up to 200 mSv, while the variation of maximum CED was from 122 up to 538 mSv. The methods based on published or typical values were found to generally provide an overestimation of E for small size patients (up to 87%) while large size patients had underestimated doses down to -71%. The methods based on particular patient data were overestimating E for most normal to large size patients (up to 106%) compared to the adopted as reference method based on NCI dosimetry system. Conclusions: Although E is recommended for population estimations, it is sometimes needed and used for individual patients in clinical practice. Its value is highly dependent on the method applied. Individual estimations of E can vary up to 18.1 times and CED estimations can differ up to 6 times. There is a strong need of standardization of the methods used in the clinical practice for effective dose estimation, and/or a better patient-specific risk metrics.15:00 Comparison of foetal doses estimated by different dose calculators - Morgan Caldecoat[1,2], Jenny Edge[1], Mary Smail[1], Anne Hill[1], Pippa Dunbar[2] - [1] University Hospital Bristol and Weston NHS Foundation Trust, [2] University of the West of England
Pregnant patients occasionally undergo diagnostic CT scans for a number of clinical reasons. Some CT scans will give a relatively low dose to the foetus; for example, a CTPA protocol where radiation dose arises from scattered radiation. Other scans, such as an abdomen/pelvis protocol, will deliver higher foetal doses as the foetus is directly irradiated by the primary beam.
To justify these scans, an understanding of the radiation dose to the foetus and their subsequent risk of developing childhood cancer is required. Published values and the ImPACT calculator are currently used at UHBW; these are based on a stylised geometric phantom. Other tools that use anatomical voxel and mesh phantoms have become available. Three different CT dose calculators, ImPACT, NCICT and CoDE, were compared to identify differences between them.
Foetal dose was estimated using each dose calculator with exposure factors and scan lengths from a Siemens Definition AS+ scanner, using the standard phantom for that dose calculator. Comparisons were performed for CTPA and abdomen/pelvis with contrast protocols at early pregnancy and the three trimesters.
Differences were found between calculators. These differences were investigated by using two different phantom sizes based on the MIRD-5 and ICRP 89 reference female phantoms. Differences between the scanner models available in each calculator were investigated by running each calculator with the same scanner model, a Siemens Sensation 16.
Overall, NCICT and CoDE returned higher foetal dose estimates for both CTPA and abdomen/pelvis protocols. Difference in early pregnancy for both protocols and in the first and second trimesters for the CTPA protocol were statistically significant (p<0.05), based on a two-way ANOVA. No statistically significant differences were observed in the later trimesters for the abdomen/pelvis protocol or when comparing the different phantom sizes or scanner models.
Differences between dose calculators have been observed in previous studies; these results support those findings. Although any phantom will not represent a real individual exactly, NCICT had anatomically realistic phantoms. NCICT was also found to be the most user-friendly. Put together, these make it likely to report more realistic foetal doses. UHBW aims to adopt this method for estimating foetal dose once the correct license for clinical use is obtained.15:20 Development of a Visual Grading Characteristics (VGC) Analysis tool - Niamh Banks - University Hospitals Bristol and Weston
Visual Grading Characteristics (VGC) Analysis is a statistical analysis technique for the assessment of clinical image quality. Clinical images are visually scored against a set of criteria, and the data is processed in a way similar to ROC analysis. The method evaluates the statistical significance of the results.
The typical approach to clinical image quality investigations is to take a mean value of the image scores. However, this method is not statistically valid given the data is ordinal, and the mean is only representative of a typical value if the data is normally distributed. In contrast, VGC analysis treats the scale steps as ordinal and does not make assumptions about the distribution of the data.
The Medical Physics department at University Hospitals Bristol and Weston have employed the VGC methodology as a tool for comparative image quality assessment, and demonstrated its use with previously acquired CT image scores.
The VGC tool has been used to compare image quality across four CT scanners in the hospital, using scores from 3 radiologists assessing 10 abdo-pelvis scans on each scanner. The VGC process was found to be a successful method for comparative image quality analysis, and has the potential to identify significant differences in CT scanner performance for larger data sizes.Discussion and updates
16:10 Interslice variation in noise - what is acceptable? - Cat Rivett
16:20 Anatomy of an MTF analysis routine - David Platten
16:30 Update on IPEM 32 Part 3 - Susan Doshi - Velindre University NHS Trust
16:40 Initial experience with detectability index - Katie Howard - East Anglian Regional Radiation Protection Service (EARRPS), Cambridge University Hospitals NHS Foundation Trust
Aim: To investigate the usefulness and usability of a detectability index (d’) for optimisation of computed tomography examination protocols.
Study: A Phantom Laboratory CatPhan® 600 was imaged on a Siemens Definition Edge CT scanner at a range of tube-current settings (50-350mAs per rotation). The acquisitions were reconstructed with a range of algorithms. The Siemens iterative reconstruction, ‘SAFIRE’, was used at each ‘strength’ from 1 to 5. Contrast-to-noise ratios (CNRs) were measured for the CatPhan images using ImageJ (v1.53s). Detectability indices for the CatPhan images were established using an ImageJ plugin (David Platten).
Typical results were demonstrated with manual measurements of CNR. Namely, that higher tube current or greater use (higher ‘strength’) of iterative reconstruction produced higher CNRs. The detectability index also clearly demonstrated the change in image quality across the acquisition and reconstruction parameters. Further, the detectability index introduces the ‘task function’ to assess image quality with respect to the contrast and resolution required of the particular imaging question.
Initial experimentation with detectability index has shown the parameter to demonstrate the expected results for known changes in image quality. Further work in this study will investigate the reliability of d’ in predicting if the image quality is sufficient to detect a lesion of specific size and contrast.
