Detection of lesions in planar mammograms is a difficult task, predominantly due to the masking effect of superimposed parenchymal breast patterns. Tomographic imaging of the breast can provide image slices through the breast, possibly reducing this masking effect. In recent years, there has been interest in developing CT mammography using flat-panel digital detectors in a truncated cone-beam geometry. In this study, we have developed a framework for determining optimal design and acquisition parameters for such a CT mammographic system. The ideal observer SNR is used as a figure-of-merit, under the assumptions that the imaging system is linear and shift-invariant, and that the noise is stationary. The ideal observer calculation uses mathematical models of signal and noise propagation through the flat-panel detector, and realistic models of the lesion detection task in breast imaging. It is used to investigate optimal kVp settings of a tungsten anode spectra for CT imaging of the uncompressed breast, given the constraint of an average glandular dose approximately equivalent to that of a two-view planar mammography study. It is observed that modeling a realistic mammographic background structure into the detection task can affect the optimal kVp settings suggested by the ideal observer SNR. Since the exposure/view in flat-panel CT mammography is considerably lower than for planar mammography, it is observed that electronic additive noise can also affect the optimal kVp setting. In general, the optimal kVp settings for the tungsten anode spectra studied here were in the range of 30-50 kVp.