This paper explores the combination of cone beam optical computed tomography with an N-isopropylacrylamide (NIPAM)-based polymer gel dosimeter for three-dimensional dose imaging of small field deliveries. Initial investigations indicate that cone beam optical imaging of polymer gels is complicated by scattered stray light perturbation. This can lead to significant dosimetry failures in comparison to dose readout by magnetic resonance imaging (MRI). For example, only 60% of the voxels from an optical CT dose readout of a 1 l dosimeter passed a two-dimensional Low's gamma test (at a 3%, 3 mm criteria, relative to a treatment plan for a well-characterized pencil beam delivery). When the same dosimeter was probed by MRI, a 93% pass rate was observed. The optical dose measurement was improved after modifications to the dosimeter preparation, matching its performance with the imaging capabilities of the scanner. With the new dosimeter preparation, 99.7% of the optical CT voxels passed a Low's gamma test at the 3%, 3 mm criteria and 92.7% at a 2%, 2 mm criteria. The fitted interjar dose responses of a small sample set of modified dosimeters prepared (a) from the same gel batch and (b) from different gel batches prepared on the same day were found to be in agreement to within 3.6% and 3.8%, respectively, over the full dose range. Without drawing any statistical conclusions, this experiment gives a preliminary indication that intrabatch or interbatch NIPAM dosimeters prepared on the same day should be suitable for dose sensitivity calibration.