To investigate the patterns and diagnostic implications of coronary arterial lesion calcification by CT angiography (CTA) using a novel, cross-sectional grading method, we studied 371 patients enrolled in the CorE-64 study who underwent CTA and invasive angiography for detecting coronary artery stenoses by quantitative coronary angiography (QCA). The number of quadrants involving calcium on a cross-sectional view for ≥ 30 and ≥ 50 % lesions in 4,511 arterial segments was assessed by CTA according to: noncalcified, mild (one-quadrant), moderate (two-quadrant), severe (three-quadrant) and very severe (four-quadrant calcium). Area under the receiver operating characteristic curve (AUC) were used to evaluate CTA diagnostic accuracy and agreement versus. QCA for plaque types. Only 4 % of ≥ 50 % stenoses by QCA were very severely calcified while 43 % were noncalcified. AUC for CTA to detect ≥ 50 % stenoses by QCA for non-calcified, mildly, moderately, severely, and very severely calcified plaques were 0.90, 0.88, 0.83, 0.76 and 0.89, respectively (P < 0.05). In 198 lesions with severe calcification, the presence or absence of a visible residual lumen by CTA was associated with ≥ 50 % stenosis by QCA in 20.3 and 76.9 %, respectively. Kappa was 0.93 for interobserver variability in evaluating plaque calcification. We conclude that calcification of individual coronary artery lesions can be reliably graded using CTA. Most ≥ 50 % coronary artery stenoses are not or only mildly calcified. If no residual lumen is seen on CTA, calcified lesions are predictive of ≥ 50 % stenoses and vice versa. CTA diagnostic accuracy for detecting ≥ 50 % stenoses is reduced in lesions with more than mild calcification due to lower specificity.