OA026] Design of an optimised bow-tie filter for cone-beam CT image guidance for children

Abstract

Purpose Children can receive excessive imaging dose from cone-beam CT (CBCT) during image-guided radiotherapy due to use of protocols designed for adults. Lower dose CBCT is a high priority for reducing the risk of radiation induced late effects in children. The range of imaging doses can be reduced by lowering exposure using an optimal ‘child-sized’ bow-tie filter with additional attenuation and tailored shape. We present a filter design using measurements from paediatric CBCT data and simulate its impact on image quality and registration accuracy. Methods The filter was designed to produce uniform signal across the average paediatric patient and provide additional attenuation compared to a standard (adult) bow-tie filter. To estimate the average patient thickness across the detector plane, we used the signal intensity profiles of projection images acquired from 15 paediatric patients (ages 3–17 years) with various tumour sites. The intensity values were converted to an estimate of patient thickness using the Beer-Lambert law, with all tissues approximated as water. We produced a symmetrical design accounting for attenuation through the average patient across the detector. CBCT scans including the added noise resulting from attenuation of the new filter were simulated by modifying projection images before reconstruction. Simulated scan quality was assessed visually and by testing registration accuracy, comparing the table correction to that for the original higher dose images. Results The resulting filter had a central thickness of 15.53 mm of Aluminium, corresponding to a fivefold reduction in signal compared to the use of the standard filter (approximately decreasing dose from 1.5 mGy to 0.33 mGy per scan). The thickness increased to 29.87 mm at the outer edge, following a curvature described by a polynomial function relating filter thickness to patient thickness across the detector. Simulated scans showed increased noise but clear bone/soft tissue boundaries. Registration remained accurate; most scans had table correction vector discrepancies within 1 mm (max 1.09 mm). Conclusions We have designed an optimised bow-tie filter for children. This dedicated filter design should allow considerable dose reduction to the patient, maintaining sufficient image quality and accurate registration. Observer studies on the acceptability of very low dose images in clinical practice are underway.