OBJECTIVE To evaluate the performance of dual-source computed tomography (DSCT) for the visualization of the coronary arteries in a population with atrial fibrillation (AF) compared to single-source CT (SSCT) and to explore the impact of patients' heart rate (HR) on image quality (IQ) and reconstruction timing. METHODS Thirty consecutive patients (11 male, 19 female; 69.0+/-9.2 years old) with suspected coronary artery disease and permanent AF were examined on a DSCT scanner (120 kV, 400 mAs/rot, 0.33 s rotation time, 64 x 0.6 mm collimation, pitch 0.20-0.28, Siemens Somatom Definition). Patients were divided into two groups: low and medium HR group (HR<or=80 bpm, n=14) and high HR group (HR>80 bpm, n=16). Five of the patients also underwent conventional coronary angiography (CAG). The raw data from both tube detector arrays were reconstructed as DSCT data using a routine algorithm (temporal resolution of 83 ms). The raw data from one tube detector array was reconstructed separately on the same system using a routine single source algorithm (temporal resolution of 83-165 ms) and defined as virtual SSCT data. Image quality was assessed using a four-point grading scale from excellent (1) to non-assessable (4). RESULTS IQ of the DSCT data was significantly better than that of the virtual SSCT data (mean score 1.33+/-0.61 vs. 1.80+/-1.02; Z=-8.755, P=0.000). 98.6% of the segments shown in DSCT were diagnostic, compared with 89.9% of the segments in virtual SSCT, chi(2)=32.595, P=0.000. In DSCT group, IQ of low HR group was also better than that of high HR group, although the difference was not as big (mean score 1.25+/-0.52 vs. 1.38+/-0.66; Z=-2.227, P=0.026). The mean HR of low HR group and high HR group were 67.4+/-8.5 beats per minute (bpm) and 94.2+/-8.8 bpm (t=-8.499, P=0.000). The range of the variation of HR was higher in high HR group than in low HR group (mean difference between maximum and minimum HR 79.5+/-21.0 vs. 49.9+/-21.1 bpm; t=-3.845, P=0.001). In 23 (77%) patients optimal IQ was achieved within one phase for all three main arteries. In low HR group, the optimal phase was distributed evenly between diastole and systole; but in high HR group, the optimal phase shifted to systole in most cases (92%). In five cases these results were compared to CAG results to look for the ability to identify stenosis with a diameter reduction of more than 50% of the lumen. With DSCT it was possible to diagnose 20 of 21 stenosis and 48 of 49 non-stenosed vessel segments correctly. With SSCT 19 of 21 stenosis and 45 of 49 non-stenosed vessel segments were diagnosed correctly according to CAG. CONCLUSIONS Due to its high temporal resolution, DSCT provides images of full diagnostic image quality in patients with AF, which otherwise would be excluded from CT examinations. The number of assessable segments for DSCT is still somewhat less than in non-AF patients, but in opposition to SSCT still valid for routine diagnostic imaging. Patients' HR had impact on IQ and reconstruction timing.