Mammographic screening for breast cancer is unable to distinguish molecular differences between hydroxyapatite (HA) microcalcifications (μcals) that are associated with malignancy and calcium oxalate (CaOx) μcals that are benign. Therefore, the objective of this study was to investigate quantitative material decomposition of model breast μcals of clinically-relevant composition and size using spectral photon-counting computed tomography (PCCT). Model μcals composed of HA, CaOx, and dicalcium phosphate (DCP) were treated as materials containing spatially coincident elemental compositions of calcium (Ca), phosphorus (P), and oxygen (O). Elemental decomposition was performed using constrained maximum-likelihood estimation in the image domain. Images were acquired with a commercial, preclinical PCCT system (MARS Bioimaging) with five energy bins selected to maximize counts at low photon energies and spectral differences between Ca and P. Elemental concentrations of Ca and P within the each μcal composition were accurately identified and quantified with a root-mean-squared error < 12%. HA and CaOx μcals, < 1 mm is size, were accurately discriminated by the measured P content with an area under the receiver operating characteristic curve (AUC) > 0.9. The mole fraction of P, P/(Ca+P), was able to discriminate all three μcal compositions with AUC > 0.8 for μcals < 1 mm is size and AUC = 1 for μcals > 2 mm in size. The overall accuracy for the classification of μcal types and quantification of P was robust against different assumptions in the elemental decomposition calibration, but quantification of Ca was improved with assumptions that most accurately accounted for the molar volume of each element within μcal compositions. Thus, PCCT enabled quantitative molecular imaging of breast μcal composition, which is not possible with current clinical molecular imaging modalities.