Objective: To evaluate a modified subtraction coronary computed tomography angiography (CCTA) technique with a two-breathhold approach in terms of image quality and stenosis grading of calcified coronary segments and in the detection of significant coronary stenosis in segments with severe calcification. Materials and Methods: The Institutional Review Board approved this study, and all subjects provided written consent. A total of 128 patients were recruited into this trial, of which 32 underwent subtraction CCTA scans and invasive coronary angiography (ICA). The average Agatston score was 356 +/- 145. In severely calcified coronary segments, the presence of significant (> 50%) stenosis was assessed on both conventional CCTA and subtraction CCTA images, and the results were finally compared with ICA findings as the gold standard. Results: For severely calcified segments, the image quality in conventional CCTA significantly improved from 2.51 +/- 0.98 to 3.12 +/- 0.94 in subtraction CCTA (p < 0.001). In target segments, specificity (70% vs. 87%; p = 0.005) and positive predictive value (61% vs. 79%; p < 0.01) were improved using subtraction CCTA in comparison with conventional CCTA, with no loss in the negative predictive value. The segment-based diagnostic accuracy for detecting significant stenosis was significantly better in subtraction CCTA than in conventional CCTA (area under the receiver operating characteristic curve, 0.94 vs. 0.85; p = 0.03). Conclusion: This modified subtraction CCTA method showed lower misregistration and better image quality in patients with limited breathhold capability. In comparison with conventional CCTA, modified subtraction CCTA would allow stenosis regrading and improve the diagnostic accuracy in coronary segments with severe calcification.