ESTRO 2024 - Abstract Book

S4826

Physics - Quality assurance and auditing

ESTRO 2024

Consequently, this work developed a phantom-based method of comparing the quality of radiotherapy planning CT images, head and neck (H&N) and prostate, in a multicentre intercomparison, with a view to benchmarking and sharing best practice for image quality improvement. The phantom image quality assessment method was clinically validated against the results from a review of clinical image quality.

Material/Methods:

CT scan datasets were collected from 10 UK radiotherapy centres for H&N (n=14) and prostate (n=13). Bespoke phantoms were designed and sent to participants: cylindrical water-filled phantoms of 30 cm and 19 cm diameter for noise-power spectrum (NPS) measurements; a 19 cm diameter water-filled phantom containing inserts of acrylic, Teflon and low-density polyethylene for measurement of target transfer function (TTF) and contrast on the two examination protocols. The same metrics were measured on local Catphan phantoms (Model 604) to determine whether future intercomparisons could utilise locally owned phantoms [2]. On each scanner, phantom measurements were made using the tube current fixed at the value delivered for a patient of 19 cm water-equivalent-diameter (WED) at mid brain level and of 30 cm WED at mid prostate level. Other scan parameters matched the clinical scan protocols. For NPS, additional images were acquired at 35 cm and 20 cm reconstruction field-of-view (FOV). Noise, noise from area under NPS curve and frequency of NPS peak were recorded. Contrast was measured from the acrylic insert and water in the TTF phantom and the polystyrene, acrylic and background material in the Catphan phantoms. Edge visibility of the 1% contrast, 15 mm diameter target in the Catphan was also scored. Noise in a sample of 10 or more clinical images per scanner was assessed using a bespoke Matlab code which modified a technique by Tian and Samei for use with large FOV radiotherapy CT scans [3]. This measurement determined whether scanner performance was consistent for different sized patients. CT dose index from the clinical images, normalised to 19 cm (head) and 30 cm (body) WED, was also collected. The clinical image quality review required 3 annonymised clinical image sets per scanner per exam type of a standardised size. 6 clinical experts reviewed all images and ranked them as ‘3’ (best), ‘2’ (average) and ‘1’ (worst) for radiotherapy contouring. Average clinical scores per scanner were compared with phantom results.

Results:

The scoring of clinical images showed a large range of results for different scanners for H&N examinations, see Figure 1. The two highest scoring centres (best image quality) had optimised their protocols. The differences were less pronounced for prostate examinations. Reviewers confirmed that, in 75% of cases, the images scoring a ‘3’ would give improved quality or speed of contouring compared to lower scored images.