Several publications have reported the usefulness of DWI for liver lesion detection [25, 31–36]. Most of these focused on the value of low b-value images (50–100 s/mm²) compared to T2WI and postcontrast T1WI. This distinction is due to the “black-blood” effect inherent in low b-values, which suppresses the background signal of vessels in the liver parenchyma while maintaining high SNR (Fig. 41.1). This is particularly useful in detecting small lesions (<1 cm) localized near small hepatic vessels, since on flow-sensitive FSE T2WI sequence, only the main branches with fast flow appear dark, while small vessels are bright and can mask adjacent small lesions. Moreover, there is reduced blurring with DWI compared to FSE T2WI or HASTE T2WI [37].

Coenegrachts et al. [38] showed a higher conspicuity of hemangiomas and metastases with RT-DWI using low b-value compared to conventional unenhanced MRI. Zech et al. [36] compared DWI (b50) with fat-suppressed T2WI and observed better image quality and better sensitivity for lesion detection with DWI (83 % vs. 61 %). Bruegel et al. [31] compared RT-DWI with five different either fat-suppressed T2WI or STIR sequences for the diagnosis of hepatic metastases at 1.5 T and found that DWI was significantly more sensitive (0.88–0.91) and accurate (0.90–0.91) than any of the T2WI sequences (sensitivity 0.45–0.62 and accuracy 0.47–0.67). Moreover, when only small (≤1.0 cm) metastatic lesions were considered, the difference in sensitivity between DWI (0.85) and T2WI (0.27–0.44) was even larger. The combination of these results, especially when combined with other similar studies already published, has led to the claim that low b-value DWI outperforms both fat-suppressed T2WI and STIR in the detection of focal liver lesions (Figs. 41.1 and 41.2). Nasu et al. reported decreased detection of liver metastases in the left hepatic liver lobe due to cardiac-related artifacts [39]; however, this was not observed by Parikh et al. [25].

Regarding the comparison of DWI with contrast-enhanced imaging, there is evidence to suggest that DWI has a better diagnostic accuracy in the detection of colorectal metastases in conjunction with mangafodipir trisodium [40] and no significant difference in performance on a per-lesion basis compared to nonspecific extracellular gadolinium-enhanced T1WI [41].

At this point, there are only a few published studies directly comparing the accuracy of gadoxetic acid-enhanced MRI to DWI for the detection of metastatic liver lesions [42, 43]. Shimada et al. [43] showed a better diagnostic performance using gadoxetic acid compared to DWI at 3 T (AUC 0.958–0.966 and 0.881–0.906, respectively; p = 0.04 for observer 1 and no significant difference for observer 2 for the detection of small metastatic lesions (≤2 cm), without significant differences in sensitivity and positive predictive values. Compared with previous studies using extracellular agents, Shimada et al reported a higher accuracy, likely due to the combination of gadoxetic acid with 3 T. In the study by Löwenthal et al. [42], performed at 1.5 T, breath-hold DWI was compared to gadoxetic acid-enhanced T1W. The authors demonstrated superiority of gadoxetic acid-enhanced MRI (detection rate, respectively, 94.4 and 100 %, 2 observers) compared to DWI (detection rate 78.3 and 97.5 %) and contrast-enhanced T1WI at the dynamic phase (detection rate 81.5 and 89.9 %). Despite the lack of a low b-value acquisition (b-values 0 and 500) and relatively low in-plane resolution (slice thickness 8 mm), subgroup analysis revealed a substantially higher sensitivity for small lesions (≤1 cm) and lower accuracy of DWI for the left lobe (7 out of 9 misclassified lesions were located in the left lobe).

The most straightforward approach for characterizing focal liver lesions relies on visual assessment of intermediate/high b-value images (500–1,000) that help distinguish between cystic and solid lesions, the latter usually retaining relative high signal intensity over the background liver. A lesion can be considered likely benign, when it is hyperintense at a low b-value and shows a progressive and substantial signal loss on intermediate/high b-values, with a corresponding higher ADC value. On the other hand, a lesion is more likely malignant, when there is no or minimal signal drop on intermediate/high b-value images, with corresponding ADC values lower or isointense to those of the surrounding liver parenchyma [15, 25, 29] (Figs. 41.1, 41.2, 41.3, and 41.4).

As described previously [15, 25, 44, 45], visual characterization of focal liver lesions can be performed with high accuracy and excellent interobserver agreement with DWI. The distinction between benign and malignant lesions is reported to be correct in 89–93.0 % of cases, even though in most of the studies benign lesions were predominantly cysts or hemangiomas. However, inhomogeneous appearance of metastases can reflect different components of tumors (cystic, necrotic, or solid) (Fig. 41.5), such as mucinous malignant lesions that may show a lower restriction to diffusion and high ADC, and can be misdiagnosed as benign lesions [29, 46].

Quantitative ADC measurements may also be used to differentiate between benign and malignant lesions owing to the generally lower ADC values in the latter as compared to the former, probably resulting from greater cellularity or nucleus/cytoplasm ratio. ADC quantification requires minimum of 2 b-values, minimum acceptable SNR at higher b-values, as well as minimal lesion size of 1.5 to 2× the in-plane resolution, so as to avoid partial volume effect. As a general rule, statistically higher ADC values have been demonstrated for benign lesions compared to malignant lesions, with variable degrees of overlap [15, 25, 45, 47, 48]. Although an average ADC cutoff of 1.4–1.6 × 10⁻³ mm/s² may provide sensitivity as high as 74–100 % and specificity in the range of 77–100 % in the distinction between benign and malignant liver lesions [46], other published ADC values show much wider variation, in the range of 0.94–2.85 × 10⁻³ mm/s² for metastases [16, 24, 33, 44, 49–52]. There is considerable overlap between the ADC values of cellular benign hepatic lesions, such as focal nodular hyperplasia or adenoma [53], and those of malignant lesions, such as metastases and HCC. Also, although cysts and solid lesions are easily distinguishable on the ADC map, sometimes metastases and hemangiomas can have similar DWI features. Also, cystic, necrotic, or mucinous components in a metastatic lesion may increase ADC values, which can generate false negatives when DWI data alone are considered (Figs. 41.5 and 41.6). Cystic, mucinous, or necrotic metastases (e.g., after systemic chemotherapy) are characterized by a low viscosity in central portion, so they may show relatively high ADC values, while occasionally hemangiomas display portions with low ADC values [48, 49, 54]. Conversely, liver metastases originated by neuroendocrine tumors, histopathologically constituted by highly concentrated small round cells, are usually characterized by low ADC values. In contrast, abscesses or inflammatory lesions show low ADC values, due to their inflammatory and viscous contents, potentially generating false positives (Fig. 41.7).

Furthermore, each study group uses their own scanning protocols, with different manufacturers, different field strength scanners, different b-values sets, breath-hold or respiratory-triggered techniques, different ADC calculation algorithms, and different measurement techniques, making it difficult to integrate results across studies and propose a standardized ADC evaluation. Therefore, DWI alone is currently not suitable for stand-alone characterization of focal liver lesions, but it shows better accuracy when used in conjunction with contrast-enhanced and conventional MRI sequences.