Herein, we observed that liver fibrosis stages were inversely associated with ADC values (r=−0.691, P<0.001), and positively associated with GPR parameters (r=0.502, P<0.001). The area under the curve for diagnostic efficacy of ADC values, GPR parameters, and their combination for F≥2 liver fibrosis was 0.831, 0.749, and 0.858, and for F≥3 was 0.872, 0.771, and 0.903, respectively. Using ADC values combined with GPR parameters was better than using ADC values or GPR parameters alone (all P<0.01). This demonstrated that the combined diagnostic tool may improve the accuracy of CHB-related liver fibrosis diagnosis, especially for F≥3.
Magnetic resonance DWI can be used to assess the physiological and pathological characteristics of the body tissue through the microscopic situation of water molecules [14]. It is an imaging technique particularly sensitive to the transverse dispersion motion of water molecules, and the size of the ADC value reflects the rapid and slow motion of tissue water molecules [15, 16]. Normal hepatocyte morphology and arrangement are orderly, and the hepatic parenchymal cells and hepatocyte matrix are stable. When hepatic fibrosis occurs in patients with chronic hepatic disease, hepatocyte degeneration, necrosis, abnormal hyperplasia of intrahepatic fiber tissue, and collagen fiber deposits in the stroma of liver cells occur, and water molecular activity is limited, resulting in the reduction of water molecular activity of the liver tissue, which decreases the ADC value of the pathological tissue.
However, some studies have reported that the accuracy of the MRI ADC value was relatively low in the diagnosis of segmentation point F≥2, which suggested that its ability to distinguish between mild and moderate fibrosis was relatively weak [17–19]. Thus, the cause may be related to the extremely low deposition of extracellular matrix and no hepatic fibrosis for grade F1, and confined peripheral sinus, confluence area and leaflet fibrosis, sparse distribution of collagen fiber, no fiber interval and no leaflet structure disorder, and the diffusion restriction of water molecules for grade F2 [20–22].
In contrast, the ability of GPR to diagnose hepatic fibrosis at different stages was relatively stable. Previous studies have reported on the noninvasive diagnosis of hepatic fibrosis by serology [23]. The GPR model established by Lemoine and his colleagues was better than the APRI and fibrosis-4 models in the evaluation of hepatic fibrosis in patients with CHB in West Africa [8]. More recent studies have also demonstrated the advantages of GPR model [24–30], which were also confirmed in our previous study [31]. However, GPR only reflects matrix changes in the serum, rather than liver parenchymal degradation, lacks specificity for the liver, and is easily affected by changes in γ-glutamyl transferase and platelet levels [32]. Hence, although the ADC values and GPR parameters have good clinical values for diagnosing significant and progressive hepatic fibrosis, they may miss the opportunity for early antiviral treatment due to their low sensitivity. Therefore, they still cannot completely replace liver puncture biopsy.
Nevertheless, our results demonstrate that the diagnostic efficacy of combining ADC values and GPR parameters for diagnosing hepatic fibrosis was superior to that of a single ADC value or GPR parameter. This compensated for the ADC values’ lack of ability to diagnose light and moderate liver fibrosis, although the combined tool having a low sensitivity is acceptable. Its high sensitivity should not miss diagnosis, when ADC values were used for independent diagnosis. Moreover, ADC values combined with GPR parameters were the most advantageous in diagnosing liver fibrosis with F≥3 (AUC is 0.903), which suggested that when diagnosing mild, moderate or severe liver fibrosis, joint diagnosis can be made. This can complement each other's advantages and greatly improve the diagnostic capabilities of both.
No non-invasive testing method or a single laboratory index that completely replaces the examination results of liver biopsy exists, so combining diagnostic models of non-invasive testing and laboratory indicators may overcome the shortcomings of using a separate method or indicator. In assessing the degree of hepatic fibrosis, the ADC values were more effective than the GPR parameters. The GPR parameters can be used as a prognostic evaluation of hepatic fibrosis staging. Combining the ADC values and GPR parameters can improve the diagnostic ability for hepatic fibrosis, and provide a more reliable diagnostic index for its early diagnosis, treatment, and prognosis. A non-invasive detection joint or multiple index model should be explored in future research.
This study has some limitations. First, the patients were enrolled from a single center, and the sample size was small, which make it difficult to accurately evaluate the overall situation of patients with HBV in China. Second, the study was designed to exclude concomitant diseases that could have effects on γ-GT or platelets, such as non-alcoholic fatty hepatic disease, alcoholic hepatic disease, combined with other hepatitis virus infections. Third, MRI examination is more expensive, require longer inspection time, and cannot be available due to contraindications. Finally, fat deposition can also affect the ADC value, so further research needs to be done in the future.
Recently, with the rapid development of artificial intelligence, machine learning methods (such as Naive Bayes, random forest, and neural network) play an increasingly important role in the medical field, especially in medical imaging [33, 34]. In a future study, we will expand the sample size and conduct a multi-center, prospective research, in which we will integrate multiple data, such as serum parameters, radiology imaging and other data, by machine learning methods and stage liver fibrosis.