Dr Craig Moore & Dr Tim Wood

1. Cone Beam CT Protocol Optimisation for Prostate Imaging with the Varian Radiotherapy OBI imaging system Dr Craig Moore & Dr Tim Wood

2. Background • With the increasing use of CBCT imaging alongside complex radiotherapy treatment regimes, it is becoming more important to understand the implications of current practice • On board CBCT daily imaging for verification of patient position is now common practice across the UK • It is not acceptable to simply dismiss these concomitant exposures as negligible in comparison with the radiotherapy treatment dose • Currently, all of our CBCT systems operate using Varian default settings • A single set of exposure factors for all patients is clearly not optimised! • Vital we have an idea of patient doses so that we can develop optimisation strategies

3. kV tube Treatment head – MV beams generated here kV detector

4. Aims • This talk will focus on: • Development of a computational method to estimate dose and risk for CBCT prostate imaging • Development of a strategy for patient sized protocol optimisation for CBCT prostate imaging

5. The first step… • The first phase of this project is to gain an understanding of the doses involved in CBCT imaging • Given the context of these procedures (i.e. as part of a RT treatment), simple risk estimates based on the effective dose are probably not sufficient in isolation • We need to start thinking about organ-at-risk tolerances and other healthy tissues that are not involved in the actual treatment • Hence, we need to develop a broader understanding of where the dose is being deposited, i.e. organ doses • What is the best way to do this? • TLDs? • A computational model? • A bit of both?

6. Developing a CBCT dose model with PCXMC • We have commercially available software (PCXMC) that is widely used for performing dose assessments for radiological examinations, etc • Allows you to rotate around a reference point within a mathematical (Christy) phantom (ideal for modelling RT imaging) • Only for simple uniform X-ray spectra

7. Only uniform beams PCXMC

8. Half-fan bow-tie filter = non-uniform beam Can we account this non-uniformity to make it ‘fit’ with PCXMC

10. 4 slithers used to correct for beam non-uniformity – treat independently for each projection The PCXMC model • To model the Varian CBCT system, 8 projections around the patient were used (at 45° intervals), with equal weighting for the final dosimetry • Each projection was split into 4 ‘slithers’ to account for non-uniformity of the x-ray beam • Treat each slither independently for each projection • PCXMC requires the correct air kerma and filtration for each slither to perform its calculation – need some beam profiling!!!!

11. CBCT beam profiling • Air-Kerma and tube filtration profiles were measured with the Unfors Xi chamber at the isocentre, and using the bed to step in 1 cm increments across the full width of the bow-tie profile • Air kerma taken directly from the Unfors Xi, filtration a little more tricky!!

12. S1 S1 S2 S2 S3 S3 S4 S4 Use this info to plug into PCXMC to calculate patient dose per slither

13. Model validation • Performed TLD dosimetry on two linear accelerators (RT treatment machines), with Rando phantom loaded with 80 TLD-100H chips in the positions of the various important organs in and around the scan volume • Liver & stomach were most superior organs measured (well outside the primary beam) • Uterus & ovaries – I know prostate patients don’t have these, but it was useful for validation purposes! • Bladder, prostate & testes – these were all fully irradiated by the primary beam • Small and large intestine – partially irradiated by the primary beam • Rando was positioned with the ‘prostate’ at the isocentre, and three CBCT ‘Pelvis’ scans performed

14. Model validation – TLD dosimetry * Measured Air Kerma corrected for ratio of (μen/ρ)ICRU soft tissue/(μen/ρ)air

15. Large distance from beam Not bad given the inherent errors associated with TLD dosimetry Model validation – The comparison • So how do these compare with the PCXMC model?

16. Effective dose? • Using PCXMC to calculate the effective dose, taking out contribution to ovaries and uterus (not applicable to our prostate patients!), and the prostate (which is the target of the RT treatment, so probably should not be included in the calculation) • Effective dose = 6.0 mSv per scan • Using TLD dosimetry with Rando • Effective dose = 5.9 mSv per scan • Good agreement!! • For daily prostate imaging we get up to 222 mSv for a 37 fraction treatment regime , risk of fatal cancer: • 1 in 150 for a healthy 60 year old male (using organ specific risk factors) • 1 in 90 using generic 5% per Sv • Not insignificant!!!

17. Organ doses? • Total individual organ doses for daily imaging with 37 fractions (ignoring prostate); • Bladder > 1.2 Gy • Testicles > 1.4 Gy • Large Intestine > 0.3 Gy • These don’t feel insignificant!

18. Size specific CBCT • Currently all Pelvis exposures use the same factors (125 kVp/80mA/13ms/650 projections ~ 680 total mAs) • No compensation for patient size means the organ/effective dose reduces as the patient gets bigger • But, we should probably be increasing exposure factors for the biggest patients to ensure we get acceptable images • We have it on good authority that these patients are difficult to image • Equally, smaller patients should have a lower dose protocol

19. Protocol Optimisation • Have started looking at patient size specific exposure protocols • We have used the CT AEC phantom Tim discussed in his talk earlier today • Scanned this at the default exposure setting • 125 kVp, 80 mA, 13 ms per projection, 650 projections, • Total of 680 mAs • Decreased the mA to assess the effect on image noise: • 60 • 40 • 20 • 10 • Wanted to increase mA as well but 80 mA is its upper limit!!! • Also scanned with increased/decreased ms: • 7 • 13 • 14 • 15 • 16 • 17 • 20 • 23 • 26

20. Patient thickness Protocol Optimisation – Effect of mA (dose)

21. Protocol Optimisation – Effect of mA (dose) • As expected decrease in noise as the mA (dose) increases, for a given patient thickness • Also, increase in noise as the patient gets thicker, for a given mA (dose) • There is definitely scope to optimise the mA for average and thinner patients • Possibly as low as 40 mA for the very thin ones?? • Even scope to decrease mA for thicker patients • 60 mA is not too different in terms of noise compared to 80 mA

22. Large patients Small patients Protocol Optimisation – Effect of ms

23. Protocol Optimisation – Effect of ms • As patient size increases noise increases • Less obvious with thinner patients • As ms increases noise decreases • May be able to decrease to 7ms for very thin patients (with 80 mA) • Given that we have been told larger patient images can be poor, and that we can’t increase the mA (max 80mA which is the default), it may be possible to increase the ms for larger patients to improve image quality. • Probably go to 26ms for same noise as average patient

24. Hot off the press!!! • Very large patient scanned with default settings led to images that were not usable • We recommended they use 26 ms and image quality had improved such that images are now acceptable for the clinical intent

25. Summary • Developed a PCXMC model that simulates CBCT organ doses for pelvic (prostate) imaging • Organ doses are not insignificant for daily CBCT imaging!!! • Developing size specific protocols should be possible • Increase/decrease in mA • Increase decrease in ms • Future work will include • Adopt size specific protocols into clinical practice • Looking in more detail at the organ specific risks of cancer induction • Create some written justification protocols for the use of CBCT with dose information for size specific scans • Looking at other anatomical sites