Since diffusion-weighted (DWI) magnetic resonance (MR) imaging was first applied to brain in 1986, it was widely spread out and became established as an essential tool to detect the area of cytotoxic edema caused by acute ischemia. Diffusion is a physical process such as random motion of water molecules. Unlike in pure water which water molecules can undergo free and thermally agitated with a three-dimensional Gaussian distribution, the movement or distribution of water molecules in tissue was restricted by micro-structure more well-known like cell membranes and referred as apparent diffusion. DWI can produce its unique image contrast on the basis of differences in the apparent diffusion coefficient between tissues. It was hoped that ingenious attempts will be made to the detection and characterization of pathological lesions. Recent advances in technology enabled DWI to abdominal imaging, particularly in the liver. The advent of parallel encoding techniques played an important role to improve the image quality of DWI, which previously hampered by various artifacts such as ghosting or distortion caused by respiratory or cardiac motion and neighboring pulmonary or intestinal air. Currently, various imaging techniques and parameters are used to DWI depending on institutions, MR vendors, intended pathologies, and preference. Detection and characterization of HCCs with DWI is still challenging due to surrounding cirrhotic changes of liver parenchyma, numerous dysplastic nodules, cavernous hemangiomas, and metastases which can similar diffusion signal or ADC values. Basically ADC values were widely overlapped among HCCs, metastases, and hemangiomas and not useful to differentiate them. Only the simple cysts were potentially differentiated using ADC values alone with the cut-off value of roughly 2.5 to 3.0 × 10-3 mm2/s. Detection of HCCs with DWI has been reported in other literature as low as 50 % in evaluation using explanted livers. Though hypervascular HCCs which have acquired hepatic arterial supply usually show high signal on DWI with decreased ADC values because of increased cellularity and prolonged T2 relaxation time, DWI is not sensitive in detecting and characterizing border-line lesions such as dysplastic nodules or early HCCs. It makes more sense to use DWI for the detection of liver metastases probably because of following two reasons. First, T2 relaxation time is usually fairly long with most metastases and histopathological architecture of metastases are totally different from that of surrounding liver parenchyma. Second, surround liver parenchyma is usually non-cirrhotic in patients with liver metastases and as such heterogeneous signal or susceptibility artifacts are usually less seen in these patients.

Common therapeutic options for HCCs include liver transplantation, surgical resection, transcatheter arterial chemoembolization (TACE), radiofrequency ablation (RFA), and radiation therapy. The assessment of HCC response to TACE with DWI was described in several reports. For prediction of complete tumor necrosis after TACE, one report showed an area under the curve of 0.85, sensitivity of 75%, and specificity of 88% with ADC evaluation which showed no statistical significant difference between DWI and gadolinium contrast enhanced image subtraction technique. For the diagnosis of HCC recurrence after TACE with the use of DWI, we observed lower performance of DWI compared with gadolinium contrast enhanced imaging, with sensitivity for detection of local HCC recurrence of 60.7% compared with 82% for contrast enhanced imaging. The assessment of liver metastases response to chemotherapy with DWI was described mainly in animal models, such as breast cancer, sarcoma, glioma, and prostate cancer. There reports have all shown a significant increase in ADC values after the therapy. It is expected that early increase of ADC values can be a new biomarker of tumor response to treatment which is reliable earlier than are decreasing in tumor size, because the increase in ADC values is now considered as a consequence of cellular damage leading to necrosis. These were also observed in human study including 87 hepatic metastases in 23 colorectal and gastric cancer patients. The authors classified lesions as either responding or nonresponding, according to changes in size at the end of chemotherapy and observed that pretherapy mean ADCs were significantly lower in responding lesions than in nonresponding lesions.

The diffusion of water can be basically regarded as a random process. Hence, the chance of a particular water molecule diffusing from one location to another in a given period of time is governed by a normal probability distribution. Simply, this distribution is so-called as a Gaussian distribution with its width proportional to the diffusion coefficient. When looking at the water distribution in the tissue which is the complex structure consisting of cell membranes, organelles, and water compartments, the diffusion displacement of water molecule is no longer governed by normal probability distribution which is called as non-Gaussian distribution. This deviation from Gaussian behavior can be quantified using a convenient dimensionless metric called the excess kurtosis and is thought to represent the microstructure of the tissue. The principal metric derived from diffusion kurtosis (DK) imaging in mean kurtosis (MK), which is thought to be an index of microstructural complexity. An interesting trial was described by Raab et al. They characterized the non-Gaussian diffusion patterns of cerebral glioma microstructure with respect to the different glioma gredes (World Health Organization [WHO] grade II astrocytomas, grade III astrocytomas, and grade IV glioblastomas multiforme) by using DK imaging. In their result, significant differences between astrocytoma grades WHO II and III were demonstrated only by DK values and also area under the receiver operating characteristic curve was highest for normalized MK (0.972) to discriminate between low- and high-grade gliomas. In our preliminary study in the liver, identification of complete necrotic HCCs after TACE were significantly better in mean kurtosis than ADC value.

This lecture looks back on past developments regarding acquisition parameters and post-processing technique, and introduces the uses of DWI serving as transition to the current DWI status, for focal liver lesion detection, characterization, tumor treatment response, and diagnosis of diffuse liver diseases. Possible state-of-the-art for quantification, fast imaging, and non-breath hold technique are also introduced.

Acknowledgements

All images were obtained in Department of Radiology, Gifu University Hospital.