PURPOSE: To improve the quality of radiotherapy head and neck CT images through use of an additional image set reconstructed from the raw data of the primary scan, thus allowing parameters such as reconstruction field-of-view (FOV) and kernel to be optimised without impacting on the images used for treatment planning dose calculations. METHODS: Using a Catphan image quality phantom and a Toshiba Aquilion LB CT scanner, qualitative and quantitative measurements were made for different reconstruction kernels and FOV diameters. The preferred FOV diameter and kernels were selected. Clinical images from six patients were reconstructed using those kernels (FC13, FC41, FC44, FC64) and the chosen FOV, 200 mm. The images were ranked to choose the kernel which gave best image quality for organ delineation. The scanner workflow was adjusted to produce for every scan a second image set using the chosen kernel and FOV. Finally, for 10 patient scans, image quality was compared for the two reconstructed images. RESULTS: The second image set was produced using kernel FC44 and 200 mm FOV. The primary image set using 550 mm FOV and FC13 was unchanged and contours from the second image set merged onto the first. Oncologists reported increased confidence in contouring in all cases using the new procedure. CONCLUSION: Production of a second image set, using a reduced reconstruction FOV and a kernel which optimises contrast and sharpness, significantly improves the quality of head and neck CT images for contouring, and avoids any dose increase.