Progress of intravoxel incoherent motion diffusion-weighted imaging in liver diseases

Abstract

Traditional magnetic resonance (MR) diffusion-weighted imaging (DWI) uses a single exponential model to obtain the apparent diffusion coefficient to quantitatively reflect the diffusion motion of water molecules in living tissues, but it is affected by blood perfusion. Intravoxel incoherent motion (IVIM)-DWI utilizes a double-exponential model to obtain information on pure water molecule diffusion and microcirculatory perfusion-related diffusion, which compensates for the insufficiency of traditional DWI. In recent years, research on the application of IVIM-DWI in the diagnosis and treatment of hepatic diseases has gradually increased and has achieved considerable progress. This study mainly reviews the basic principles of IVIM-DWI and related research progress in the diagnosis and treatment of hepatic diseases.

Key Words: Magnetic resonance imaging, Diffusion magnetic resonance imaging, Liver diseases, Liver cirrhosis, Carcinoma, Cholangiocarcinoma

Core tip: Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) utilizes a double-exponential model to obtain information on pure water molecule diffusion and microcirculatory perfusion-related diffusion, which makes up for the insufficiency of traditional DWI. In recent years, research on the application of IVIM-DWI in the diagnosis and treatment of hepatic diseases has gradually increased and has made great progress. This study mainly reviews the basic principles of IVIM-DWI and its research progress in the diagnosis and treatment of hepatic diseases.

INTRODUCTION

Magnetic resonance diffusion-weighted imaging (MR-DWI) uses a single exponential model to obtain the apparent diffusion coefficient (ADC) to quantitatively reflect the diffusion motion of water molecules in tissues[1]. However, voxels also contain diffusion information other than the diffusion motion of pure water molecules, namely, the microcirculatory perfusion of the capillary network in tissues[1-4]. The ADC values obtained by traditional DWI contain two types of information: Water molecule diffusion and microcirculatory perfusion. The intravoxel incoherent motion (IVIM)-DWI theory was initially proposed by Le Bihan et al[4,5] in the 1980s; IVIM-DWI can obtain quantitative information on pure water molecule diffusion and micro-circulatory perfusion-related diffusion and thus compensate for the shortcomings of traditional DWI. This modality can also more accurately and truly reflect the characteristics of tissue structure and pathological changes. Since 1999, when Yamada et al[6] for the first time applied IVIM-DWI technology to the liver, the research on liver IVIM-DWI has gradually increased and has made great progress. This study reviews the basic principles of IVIM-DWI and related research progress in the diagnosis and treatment of hepatic diseases.

BASIC PRINCIPLES OF IVIM-DWI

IVIM-DWI originated from diffusion-weighted magnetic resonance imaging (MRI). MR-DWI can noninvasively detect the diffusion motion of water molecules (water molecules in fluid exhibit microscopic erratic random movement, known as Brownian motion), which reflects the biological characteristics of the tissue. When the signal derives from biological tissues, some factors attenuate the diffusion motion according to different diffusion coefficients, representing the ADC, which is the sum of the contributions of all motion-related diffusion coefficients. The MR signal can be expressed by a single exponential equation:

Sb/S0 = exp(-b×ADC).

Where b is the diffusion-sensitive gradient with a unit of s/mm²; Sb represents the signal obtained at a particular b value; and S0 represents a signal that does not apply a diffusion gradient[7].

IVIM-DWI uses multiple b values and a double-exponential model for image acquisition and analysis, and simultaneously obtains information regarding pure water molecular diffusion and microcirculatory perfusion-related diffusion. IVIM-DWI measures perfusion based on the following assumption: The effects of incoherent motion and blood movement can be separated from other incoherent effects. The IVIM-DWI model assumes that other sources of incoherent signals (water exchange between blood and tissue) are negligible compared with blood flow. In the IVIM-DWI model, the relationship between the degree of local signal attenuation in the tissue and the b value can be expressed by the following formula[8]:

S_b/S₀= (1-f) ×exp(-b×D) +f×exp(-b×D*).

Where D is the simple diffusion coefficient, also known as slow the apparent diffusion coefficient (D_slow), which reflects the diffusion motion of pure water molecules in the tissue of interest; D* is the pseudo-diffusion coefficient, also known as the fast apparent diffusion coefficient (D_fast), which reflects the microcirculatory perfusion in the capillary network in the area of interest; and f is the fraction of the fast apparent diffusion coefficient, which represents the volume ratio of the correlation effect of the local microcirculatory perfusion to the total diffusion effect and can be used to determine the blood volume in the area of interest. According to the formula, since D* is obviously higher than D, the information obtained from a low b value (< 200 s/mm²) mainly reflects the microcirculatory perfusion effect, while the signal attenuation obtained by a high b value (≥ 200 s/mm²) mainly reflects the diffusion effect[5,9]. Some researchers[10] believe that f × D* should be used as a new cyclic parameter and that its potential pathophysiological effects should be further elaborated in future studies.

Multiple b values are required for IVIM-DWI to separate tissue diffusion information from perfusion information. Theoretically, more b values correspond to a more accurate data fit and thus better image quality. However, a longerdata collection time corresponds to more clinical application limitations. The number of b values used in published studies ranges from 4 to16, with many studies using approximately 10 b values[6,9,11,12]. In terms of the distribution of b values, Cohen et al[13] studied the influence of low b values on D* and showed that at least two b values below 50 s/mm² are required to ensure that D* is not underestimated. Koh et al[14] proposed selecting fewer high b values and more low b values to focus data collection on the sensitive range of perfusion. At present, no consensus on the choice of b values is available, and optimization research on the number and distribution of b values is ongoing[11,15-17]. For clinical research, the choice of b values should be based on the different tissues tested and the purpose of the evaluation. The factors influencing the measurement results of IVIM-DWI are complex. In addition to b values, scanner and field intensities[11], the diffusion gradient and respiratory mode[18,19], the DWI sequence[20,21], other physiological activities (such as glandular secretions and fluid flow in glandular canals)[22], motion artefacts, fitting technology and region of interest (ROI) settings[23-28] can all affect the IVIM-DWI image quality and repeatability of the data. These should also be considered in the clinical study design.

NONALCOHOLIC FATTY LIVER DISEASE

In recent years, the prevalence of fatty liver disease in China has gradually increased, and the onset of the disease is trending to involve youngerpatients. Approximately 20% of noninvasive nonalcoholic fatty liver disease (NAFLD) casesdevelop into nonalcoholic steatohepatitis, which can progress to hepatic fibrosis or even cirrhosis[29]. Studies have shown that[30-33] the histological characteristics of nonalcoholic hepatic steatosis significantly affect the IVIM-DWI parameters. Parente et al[31] used IVIM-DWI to study type II diabetic patients with or without NAFLD and found that both the D value and D* value of fatty livers were significantly lower than those of non-fatty livers. The decrease in the D* value reflects the decrease in hepatic perfusion caused by hepatic steatosis, whereas the decrease in the D value may be caused by hepatocytomegaly, the deposition of intracellular fat and the simultaneous restriction of water molecule diffusion intracellularly and extracellularly[33]. The study by Joo et al[34] showed that the f value of NAFLD livers was significantly lower than that of normal livers, and this f value is associated with the severity of NAFLD.

HEPATIC FIBROSIS

Although percutaneous liver biopsy can determine the severity of hepatic fibrosis, this test is invasive, and its clinical application is limited. IVIM-DWI has important value in the evaluation of hepatic fibrosis[9,12,19,35-47].

Studies[12,44,45] have shown that the D, D* and f values of hepatic fibrosis patients were significantly lower than those in normal control groups. The IVIM-DWI parameters D, D* and f can be used to distinguish healthy people from patients with hepatic fibrosis, with the f value exhibiting the best diagnostic value[38].The study by Shiraga et al[36] showed that IVIM-DWI also identified a pre-fibrotic state of the liver in Fontan patients. The study included five consecutive Fontan patients and four age-matched healthy volunteers. The results showed that in the five Fontan patients, laboratory tests and ultrasound showed almost normal liver conditions, and cardiac catheterization and MRI showed good Fontan circulation, but the D, D* and f values of Fontan patients were significantly lower than those of the control group. However, some studies[9,41,46] have shown that the D value was not significantly different between normal liver tissue and hepatic fibrosis tissue. Dyvorne et al[19] showed that D* values were not significantly different between normal volunteers and patients with hepatic fibrosis.

In addition, many studies have focused on the use of IVIM-DWI technology to determine the severity of hepatic fibrosis, but the conclusions are not completely consistent[12,37-45,47]. Hu et al[40,45] showed that IVIM-DWI-related parameters were negatively correlated with hepatic fibrosis stages, with the D, D* and f value decreasing with increases in the severity of hepatic fibrosis. Chen et al[45] used IVIM-DWI to scan 25 patients with hepatic fibrosis and 25 healthy volunteers. The results showed that the ADC, D, D* and f values in the hepatic cirrhosis group were significantly lower than those in the control group. Additionally, significant differences were found in the parameters of hepatic fibrosis stages (the F0-1 group and F2-4 group and theF0-2 group and F3-4 group). However, further studies have shown that some parameters are not correlated with the hepatic fibrosis stage[37-39,41-44]. For example, Chung et al[41-44] showed that a higher stage of hepatic fibrosis corresponded to lower D* and f values; however, no correlation was found between the D values and the stages of hepatic fibrosis. On the other hand, Hu et al[39] observed a difference only in the D* values in different fibrosis groups. Other studies[48-51] have shown that IVIM-DWI is not sufficient for staging hepatic fibrosis (Table 1). The different D and D* values in the same disease of the previous studies maybe due to the following reasons: (1) The number of patients with each stage of disease was different; (2) The number of b values or an uneven b value distribution may limit the fitting precision; (3) D* has high uncertainty and poor reproducibility; and (4) D values were influenced by steatosis and iron overload[19,45,50].

Table 1 Summary of hepatic fibrosis intravoxel incoherent motion diffusion-weighted imaging studies.

Ref.	Subject number	Key findings
Chen et al[45]	50	The ADC, Dslow, Dfast, and FF values of liver fibrosis stages between the groups F0-1 and F2-4, and the groups F0-2 and F3-4 showed significant differences. As the stage of fibrosis increased, the values decreased
Hu et al[40]	57	Significant differences in D, f, D* and ADC values were found between stages (stages F0–F4). Inverse correlations were identified between fibrosis stages and D, f, D*, and ADC values
Ichikawa et al[41]	129	D*, f and ADC values decreased with fibrosis stage
Wu et al[43]	49	Liver fibrosis exhibited significant correlations with Dfast and fvalues. Dfast values in the F4 group were significantly lower than those in the F0, F1 and F2 groups
Hu et al[39]	56	The mean, interquartile (IQ) range, and percentiles (50th, 75th and 90th) of D* maps were significantly different between the F0–1, F2–3 and F4 groups
Huang et al[48]	42	IVIM diffusion MRI has high diagnostic performance in detecting viral hepatitis B-induced liver fibrosis
Liang et al[49]	30	The D value decreased as the fibrosis level increased
Franca et al[50]	74	ADC and f values were lower with higher fibrosis stages

ADC: Apparent diffusion coefficient; IVIM: Intravoxel incoherent motion; MRI: Magnetic resonance imaging.

In summary, the IVIM-DWI parameters of hepatic fibrosis tissues show a decreasing trend compared with those of normal hepatic tissues, and IVIM-DWI can be used for the diagnosis of hepatic fibrosis; however, the value of applying IVIM-DWI for hepatic fibrosis staging still requires further exploration.

DIFFERENTIATION OF BENIGN AND MALIGNANT HEPATIC TUMORS

A large number of studies have discussed the application of IVIM-DWI in the diagnosis of hepatic tumors. Studies have shown that[6,52-58] IVIM-DWI can be used for the diagnosis and differentiation of benign and malignant hepatic tumours, and a consensus indicates that the ADC and D values have high efficacy for differential diagnosis. The D value is the only simple diffusion coefficient sensitive to water molecule diffusion. For malignant lesions characterized by diffusion restriction, the D value has higher diagnostic efficiency than the ADC[54-56]. Lou et al[54] performed IVIM-DWI on 74 patients with hepatic tumors and showed that the ADC, D and f values in the malignant group were significantly lower than those in the benign group, with a larger area under the curve for the D value (0.968) and higher differentiation sensitivity (92.30%); however, no statistically significant difference in the D* value was foundbetween the two groups. Furthermore, conflicting results have been reported[53,57,59]. The results of the study by Watanabe et al[57] showed that for the identification of benign and malignant hepatic tumors, the area under the receiver operating characteristic (ROC) curve of the ADC value was higher than that of the D value; and Doblas et al[59] suggested that compared with the ADC value, IVIM-DWI parameters could not improve the differentiation efficiency for hepatic focal lesions.

The use of the perfusion-related parameters D* and f for the differentiation of benign and malignant nodules is controversial[58,60]. Penner AH et al[60]showed that compared with the traditional ADC, IVIM-DWI parameters were more significantly different between lesions and normal liver tissue. Ichikawa et al[58] studied 84 cases of focal benign and malignant hepatic lesions and found that the D and D* values of malignant hepatic lesions were significantly lower than those of benign hepatic lesions, while the f values showed no difference between benign and malignant hepatic lesions. These results suggested that the D* and D values were inhibited in malignant lesions. In contrast, some studies[52-54,61] have suggested that the D* and f values are meaningless for differentiating benign and malignant hepatic lesions. A possible reason is that the D* value may fluctuate in a large range due to different blood supply types in hepatic hemangiomas[52,61], while the magnitude of the f value depends on the time of echo (TE) (a longer TE corresponds to a greater f value). In addition, D* and f values are not uniquely specific to perfusion and may also be sensitive to other flow information, such as the excretion, direction or pattern of diffusion of a granule or gland[53,57,62] (Figures 1 and 2, Table 2).

Figure 1 A 51-year-old male patient with hepatocellular carcinoma in the right lobe of the liver. The tumor was hypointense on T1-weighted image (T1WI) and hyperintense on T2-weighted image (T2WI). Theapparent diffusion coefficient (ADC), D, D*and f maps were obtained by post processing of IVIM-DWI images. A: T1WI; B: T2WI; C: ADC map; D: D map; E: D* map; F: F map.

Table 2 Summary of hepatic nodules intravoxel incoherent motion diffusion-weighted imaging studies.

Ref.	Subject number	Key findings
Luo et al[53]	27	ADCtotal, D, and D* values of HCC were lower than those of focal nodular hyperplasia
Luo et al[54]	75	ADCtotal, D, and f values of malignant lesions were lower than those of benign lesions
Klauss et al[55]	72	D and ADC values of HCC were lower than those of focal nodular hyperplasia
Yoon et al[56]	157	True diffusion and ADCtotal of malignancies were lower than those of benign malignant lesions
Watanabe et al[57]	120	D and ADC values were lower in the malignant group than in the benign group
Ichikawa et al[58]	84	D and D* values of malignant lesions were lower than those of benign lesions
Doblas et al[59]	86	The apparent and pure diffusion coefficients were lower in malignant than in benign tumors
Penner et al[60]	61	ROC analysis demonstrated the best discriminability between HCCs and focal nodular hyperplasia for ADCslow and f values
Zhu et al[61]	39	ADCtotal and D values of malignancies were lower than those of hemangioma

ADC: Apparent diffusion coefficient; HCC: Hepatocellular carcinoma.

Figure 2 A 28-year-old female patient with focal nodular hyperplasia in the left lobe of the liver. The tumor was isointense on T1-weighted image (T1WI) and T2-weighted image (T2WI) with a star-shaped area in the center of the lesion (hypointense on T1WI and hyperintense on T2WI). A: T1WI; B: T2WI; C: ADC map; D: D map; E: D* map; F: F map.

DIFFERENTIATION OF HEPATIC MALIGNANCIES WITH DIFFERENT PATHOLOGICAL TYPES

Few studies have investigated IVIM-DWI for the differentiation of hepatic malignancies of different pathological types. Previous studies have suggested that IVIM-DWI is beneficial for the differentiation of hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC)[63-66], but the differentiation of HCC and metastatic hepatic cancer requires further study[52]. Shao et al[64] discussed the role of IVIM-DWI in differentiating hepatitis B virus (HBV)-related ICC from HCC and found that the ADC and D values of ICC were significantly higher than those of HCC, while the f value of ICC was significantly lower than that of HCC, and the areas under the ROC curves of the ADC, D and f values were 0.724, 0.753 and 0.741, respectively. Wei et al[63] showed that the ADC and D value could differentiate ICC from HCC. However, the D* and f values showed no difference in the ICC and HCC groups. Choi et al[65]studied 161 cases of focal hepatic lesions using IVIM-DWI. The results showed no significant difference in the ADC value in any malignant tumor. The D_slow value of HCC was significantly lower than that of ICC, and the f value was significantly higher than those of ICC and metastatic carcinoma. A meta-analysis conducted by Wu et al[52] (six papers meeting the requirements) showed that the ADC, D, D* and f values were not significantly different between metastatic liver cancer and HCC. The reason for this finding may be that these metastases originated from some primary tumors such as lesions in the gastrointestinal tract, lung, breast or urogenital system, which may have caused some changes in cell density and microcirculation. In addition, different HCC histological classifications and compositions used in different studies may have also caused certain effects.

HCC HISTOLOGICAL CLASSIFICATION

Studies have shown that the IVIM-DWI parameters of HCC are correlated with histological classification[19,67-75]. Li et al[73] performed an IVIM-DWI analysis on a rat model of hepatic cancer and showed that the ADC and D values of higher-grade lesions were lower than those of lower-grade lesions, and the ADC and D values were negatively correlated with the tumor’s histological classification. The D* and f values of higher-grade lesions were higher than those of lower-grade lesions, and the D* and f values were positively correlated with the tumor histological classification. During tumorigenesis, as the degree of tumor differentiation becomes increasingly poor, the cellularity and nuclear-to-cytoplasmic ratio increase, and as the extracellular space decreases, water diffusion is restricted, resulting in reduced ADC and D values in poorly differentiated HCC[68,69].

Regarding the role of D* and f values in the histological classification of HCC, the results of available studies are not completely consistent. Sokmen et al[71] showed that the ADC and D values were negatively correlated with the histological classification of HCC, the f value was positively correlated with the histological classification of HCC tissue, and no differences in the D* value in different histological classifications of HCC were observed. Granata et al[74,75] observed the role of the ADC, D and f values in differentiating the histological classifications of HCC but did not observe any significance of the D* value for differentiation. Other studies showed only the roles of the ADC and D value in the differentiation of the histological classifications of HCC, and the roles of D* and f values in differentiation were not observed[19,67-69], which may be explained by differences in b-value groupings and patient populations as well as consequent variation in tumor grade distributions[67,75].

EVALUATION OF THE HCC TREATMENT RESPONSE

Studies have shown that IVIM-DWI has important application value in the treatment and follow-up of HCC.

Zhang et al[76] conducted IVIM-DWI on 157 patients with HBV-related HCC who underwent surgical resection. The results showed 47 cases of tumor recurrence, and a D value < 0.985 × 10^-3 mm²/s was found to be a risk factor for recurrence. The D value of the IVIM-DWI model has been suggested as a potential biomarker for preoperative prediction of hepatic cancer recurrence[77-79]. Zhao et al[78] reported that D is an independent predictor of microvascular invasion (MVI); when the cut-off D value was 1.16 × 10^-3 mm²/s, the sensitivity and specificity for predicting MVI were 66.7% and 88.9, respectively.

IVIM-DWI parameters can distinguish the survival area, the fibrous area and the necrotic area of malignant hepatic tumors[80]. A study[81] including 15 patients undergoing resection of hepatic metastases of colorectal cancer found that the ADC and D values were associated with the degree of necrosis of metastatic hepatic tumors, suggesting that the specificity of free molecular proliferation of metastatic colorectal cancer increased after systemic chemotherapy.

In the early stages of local or targeted therapy for tumors, changes in tumor volume are often not detected by traditional imaging, and functional imaging is of great value in this field[82,83]. IVIM-DWI parameters can be used as biomarkers for early response evaluation[84-88]. Studies have shown that IVIM-DWI has important value in the evaluation of transarterial chemoembolization (TACE) treatment responses and the prediction of therapeutic effects in HCC[84,89-93]. The study by Peng et al[91] showed that after TACE treatment for HCC, the ADC and D values in tumor tissues significantly increased, the D* value decreased significantly, and the f value did not change significantly. The reason for the inconsistent changes in these two perfusion-related parameters may be that the D* and f values represent different aspects of perfusion, the former is mainly related to the capillary velocity of local tissues, while the latter is related to the blood volume of local tissues. The study by Park et al[89] showed that the D* value could reflect the blood supply of HCC and predict the deposition effect of iodide oil after TACE surgery; the D* value and arterial enhancement ratio (AER) of preoperative lesions in the group with better iodized oil deposition were significantly higher than those in the group with poor iodized oil deposition. Wu et al[92] believed that the ADC maybe the best independent predictor of TACE treatment responses.

The IVIM-DWI perfusion parameter f is significantly correlated with microvascular dysplasia (MVD) in HCC[88]. IVIM-DWI can be used to evaluate the anti-angiogenic effect of sorafenib on HCC[87,88,94-96]. Yang et al[95] established 35 nude mouse models of HCC and conducted IVIM-DWI at baseline and 7, 14 and 21 dafter sorafenib treatment. The results revealed that compared with the control group, the ADC and D values in the treatment group at each time point were significantly increased, while the f value was significantly decreased at 7 d and significantly increased at 21 d. In the treatment group, both the ADC and D values were significantly higher than the baseline values on days 7, 14 and 21, while the f value decreased significantly on the 7^th day and increased significantly on the 21^st day. In the treatment group, the ADC, D and f values were significantly correlated with the necrotic fraction (NF). These results suggested that IVIM-DWI parameters can be used as biomarkers to evaluate the therapeutic efficacy of sorafenib therapy for HCC. Shirota et al[87] conducted IVIM-DWI on nine patients with HCC treated with sorafenib, and the results showed that the baseline D value in the responsive group was significantly higher than that in the non-responsive group. The sensitivity and specificity of the differential therapeutic responses were 100% and 67%, respectively, and the other parameters showed no difference between the two groups, suggesting that the D value can be used to predict the therapeutic response of HCC to sorafenib.

OTHER DISEASES

The application of IVIM-DWI in hepatic ischemia-reperfusion injury (IRI)[97-99], hepatic alveolar echinococcosis[100], hepatic dysfunction[101], hepatic sinus obstruction syndrome[102] and other diseases is also gradually developing. For example, studies have found that IVIM-DWI-related parameters are significantly reduced after liver IRI[99].

CONCLUSION

In conclusion, IVIM-DWI seems to accurately reflect information related to the diffusion of simple water molecules and microcirculatory perfusion in tissues and it could have important application value in the diagnosis of hepatic fibrosis, the differentiation of benign and malignant hepatic lesions, the histological classification of HCC, the evaluation of local and targeted therapeutic response and the prediction of therapeutic efficacy. However, the value of IVIM-DWI in assisting in the histological classification of hepatic fibrosis and in differentiating malignant hepatic tumors with different pathological characteristics still requires further investigation. The standardization of IVIM-DWI scanning technology (such as the optimization and standardization of the number and distribution of b values) is an important topic, which warrants further study. In addition, the characteristics and diagnostic value of the D* and f values still require further exploration. With continuous innovation in IVIM-DWI technology and in-depth clinical research, IVIM-DWI will play a more important role in the diagnosis and treatment of hepatic diseases in the future.