A channelized hotelling trace collimator design method based on reconstruction rather than projections

In this study a new SPECT (single photon emission computed tomography) collimator design technique is investigated. In this feasibility study, the collimator hole size is the only design variable changed. Other collimator parameters are fixed to a low-energy-high-resolution (LEHR) design. The design is based on a task of imaging a small hot lesion with a uniform background. A channelized hotelling trace (CHT) is used to quantify the lesion detectability in a reconstructed image for a particular collimator parameter. The unique feature of this development is that the image quality is not evaluated for planar projection data, but rather is evaluated for the reconstructed tomographic image with an OS-EM (ordered-subset expectation-maximization) reconstruction algorithm. Our results suggest that a collimator hole size that is larger than the LEHR collimator hole size is preferable for lesion detection. This implies that the parameter that gives the optimal lesion detectability on the planar projections may not give the optimal lesion detectability in the reconstructed image. In SPECT, the optimal collimator design should be reconstruction algorithm dependent.