Diffusion weighted imaging (DWI) plays an important role in detecting and characterizing liver lesions or tumors. Due to respiratory motion, three kinds of liver DWI acquisitions are commonly used in hepatic DWI: breath-hold (BH), respiratory-triggered (RT) and free-breathing (FB) ^[1]. Due to its collection in all phases of the respiratory cycle, the ADC value obtained by FB DWI is often not accurate. While the BH solves the issue of respiratory cycle, the NEX is limited by the time of holding breath, which would lead to a low SNR and hence degrades the clinical value. The RT acquisition ensures consistent respiratory phase of the DWI data with enough NEX, however is time consuming which may restrict its widely application ^[2]. In this work, those three types of the DWI techniques were tested in the detection of focal liver lesions to find an optimal option. One hundred and fifty-two patients with focal liver lesions underwent routine MR and the breath-hold, respiratory-triggered and free-breathing DWI scans of the liver on a 3.0 T system (GE MR750). Single shot SE-EPI sequence based DWI was performed with the following parameters: axial plane, slice thickness/gap = 6 mm/1 mm, FOV = 36 cm×36 cm, matrix = 128×128. Patients were divided into two groups with different DWI scanning protocols. Seventy-five of them were examined by method 1: those three techniques were performed with b values of 100 and 800 s/mm² and identical parameters except for signal average (two for respiratory-triggered, one for breath-hold, and four for free-breathing). Twenty-four slices were scanned with the whole liver coverage. The scan time was 20 s for breath-hold, 2 min for respiratory-triggered and 1 min for free-breathing respectively. Another seventy seven patients were examined by method 2: The scanning parameters of the three techniques are consistent, the numbers of excitation were fixed to 2，only ten slices were scanned with the coverage of local lesions. Overall scan times were controlled in 20 s for breath-hold, 25s for free breathing, 1 min for respiratory-triggered DWI, b value were still 100 and 800 s/mm². Two radiologists evaluated the image artifacts, subjective image quality scores, signal-to-noise ratio, and ADC value by using the GE AW4.6 workstation. The SPSS17.0 statistical software has been used for the data analysis. All the three DWI images require multiple comparisons, on the basis of Bonferroni correction, P < 0.0167 was considered statistically significant.In method 1，the ADC values of the normal liver parenchyma in three kinds of DWI are not matched when b = 800 (P < 0.0167), consistency of the performance was moderate (ICC = 0.425). However, the ADC values had no statistical difference when b = 100 (P > 0.0167) and they showed a relative strong consistency (ICC = 0.646). All three acquisitions, the focal liver lesions’ detection and characterization rates have no statistical differences (P > 0.05). But the focal lesions’ the ADC values of the three groups showed high correlation (when b = 800 and 100, the ICC were 0.970 and 0.887, respectively), and the CNR, SNR of breath-hold is lower than the respiratory-trigger and free-breathing DWI (P < 0.0167), more details reference to Table.1. In method 2, the three acquisitions’ lesion detection (90.1 ~ 95.6%) and characterization rate (85.4 ~ 89.7%) showed no statistical significant difference (P > 0.05), and the ADC values of normal liver (b = 800 and 100, the ICC was 0.701 and 0.701, respectively) and focal lesions (b = 800 and 100, the ICC was 0.947 and 0.947, respectively) showed high correlation, but the free-breathing DWI’s CNR, SNR was significantly lower than the respiratory-triggered and breath-hold DWI (P < 0.0167),more details reference to Table.2.For respiratory-triggered DWI of the 3T MRI system can achieve higher SNR and CNR, hence it would be most advantageous. Breath holding is the best choice for patients with focal lesions that whole liver DWI is to be completed in a short time; for the patients who need to do the DWI scan to evaluate the lesion and the responds of treatment, the breath-hold acquisition is optimal.