Prediction of Microvascular Invasion and Recurrence Using Preoperative 18F-FDG PET/CT Metabolic Parameters in Hepatocellular Carcinoma

Background: Microvascular invasion (MVI) is very important in the evaluation of hepatocellular carcinoma (HCC),but diagnosis is determined by postoperative pathology; thus, preoperative non-invasive methods will play an active role. The purpose of this study was to assess the performance of metabolic parameters of preoperative 18 F-uorodeoxyglucose positron emission tomography/computerized tomography ( 18 F-FDG PET/CT) in the prediction of MVI and postoperative recurrence in primary hepatocellular carcinoma. Methods: We retrospectively collected 72 patients with HCC who have performed 18 F-FDG PET/CT scan before partial hepatectomy between 2016 and 2019. We used both normal liver tissue and inferior vena cava as the reference background, and combined with clinicopathological features, 18 F-FDG PET/CT metabolic and volumetric indices to predict MVI and postoperative recurrence of primary HCC before surgery. Results: Twenty-one of the 72 patients recurred, in recurrent cases showed higher maximum standard uptake value (SUV max ), TNR (ratio of tumor SUV max to mean SUV [SUV mean ] of the background tissue), metabolic tumor volume (MTV) and total lesion glycolysis (TLG) than nonrecurrence cases(p<0.001). All 18 F-FDG PET metabolic and volumetric indices for predicting postoperative HCC recurrence were signicant on receiver-operating-characteristic (ROC) curve analyses (p<0.05). TNR IVC , TNR NL , MTV, TLG IVC , and TLG NL were signicant factors for predicting MVI in HCC (p<0.05). On multivariate analyses, MVI, SUV max , TNR IVC , TNR NL , MTV, TLG IVC , and TLG NL (p<0.05) are independent risk factors for predicting postoperative HCC recurrence. TNR IVC is the most relevant PET/CT parameter for predicting MVI in HCC, and MTV is the most valuable for predicting postoperative HCC recurrence.


Background
Hepatocellular carcinoma (HCC) accounts for 85-90% of all primary liver cancers. Radical resection is the rst choice for liver cancer. However, the 5-year recurrence rate is approximately 70% even with R0 resection [1] . Therefore, early diagnosis and accurate prediction of postoperative liver cancer recurrence is of great signi cance. It has been reported that the main risk factors for the recurrence of liver cancer include liver function, hepatitis, tumor size, stage, grade, microvascular invasion (MVI), and treatment [2][3] . As an indicator of tumor aggressiveness, MVI is very important for postoperative evaluation. However, MVI is determined by postoperative pathological examination. Preoperative non-invasive diagnosis of MVI will play an active role in the determination of surgical resection range and other treatment planning, which can effectively reduce the recurrence rate.
As an important molecular imaging technique, positron emission tomography/computerized tomography (PET/CT) plays a unique role in the diagnosis and therapeutic decision of HCC. 18 F-uorodeoxyglucose ( 18 F-FDG) PET/CT re ects the metabolic activity and differentiation of HCC based on variable parameters, and has the ability for tumor staging, re-staging, evaluation of therapeutic effect, and determination of malignant potential and prognosis [4][5][6][7] . Metabolic tumor volume (MTV) and total lesion glycolysis (TLG) of 18 F-FDG PET/CT have been concerned as measures of tumor volume and burden with high glucose metabolism [8] . However, they have been mostly studied in lymphoma, lung cancer and intestinal cancer, and rarely in HCC. In addition, most of the published studies are based on liver background. As the uptake in normal liver tissue is unstable due to susceptibility to hepatitis, cirrhosis, and other factors, this study intends to use the inferior vena cava as a reference background to overcome the shortcomings of the existing methods. At present, no studies have been conducted to investigate whether multiple metabolic parameters of 18 F-FDG PET-CT could provide more comprehensive predictive information for preoperative evaluation of MVI and the tumor activity of HCC using the inferior vena cava as a reference background.
This study intends to use the inferior vena cava as a reference, instead of the commonly used PET/CT reference background, to assess the prognostic value of the following PET/CT metabolic parameters for preoperative evaluation of MVI and postoperative recurrence of primary HCC: maximum standard uptake value (SUV max ), TNR (ratio of tumor SUV max to mean SUV [SUV mean ] of the background tissue), MTV, and TLG.

Materials And Methods
Patients Written informed consent for PET/CT was obtained from all patients in this study. This study was approved by the Ethics Review Committee of Fudan University Shanghai Cancer Center. A retrospective analysis was performed on 72 patients (58 males and 14 females) with HCC diagnosed after partial hepatectomy in the Department of Hepatobiliary Surgery, Fudan University Shanghai Cancer Center from January 2016 to February 2019.
Inclusion criteria 1) age > 18 years; 2) HCC was suspected on preoperative enhanced CT or MR or PET/CT, and con rmed by postoperative pathological analysis; 3) whole-body 18 F-FDG PET/CT was performed within two weeks before the operation, and resectability was con rmed preoperatively; 4) laboratory tests (heart, kidney, and blood etc.) were within normal limits; 5) no other antitumor treatment before and during the operation, and no other tumor history; 6) no postoperative complications (e.g., liver failure, bile leakage, upper gastrointestinal bleeding, hepatorenal syndrome, abdominal abscess, lung infection, and other infections); and 7) complete medical records and follow-up data.
Exclusion criteria 1) liver cancer with unresectable distant metastases; 2) liver cancer with clear Portal vein tumor thrombosis (PVTT); 3) a history of other malignant tumors or other anti-tumor treatments; 4) active infections; 5) elevated fasting blood glucose or any abnormal safety laboratory ndings; 6) perioperative death, very early relapse, and loss to follow-up (2 and 12 months); and 7) incomplete follow-up data. All patients were followed up for a minimum of 12 months. Liver function tests, alpha fetoprotein (AFP) measurement, and enhanced CT were performed every 3-6 months after hepatectomy to exclude recurrence. 18 F-FDG PET/CT PET/CT A Siemens Biograph 16 HR PET/CT scanner was used. 18 F-FDG with a radiochemical purity of > 95% was produced by a cyclotron and its synthesis module in our department. All patients were fasted for at least 6 h, rested in the supine position for 60 min, and injected intravenously with 18 F-FDG at a dose of 7.4 MBq/kg before the scanning. Blood glucose levels were controlled to be equal to or less than 7.78 mmol/L. The scanned area was between the calvarium and the top of the thighs. The CT parameters were as follows: voltage 120 kV, current 100 mA, slice thickness 3.75 mm, helical pitch 3.6, and tube rotation time 0.5 s. Immediately after CT, PET was performed with the same machine for the same scanned area as that of CT. 3D-PET emission data were acquired for 2 min/bed position for a total of 6-7 bed positions (data for the head and trunk were acquired separately). CT data corrected for attenuation were fused with images reconstructed by the iterative method to obtain PET/CT fusion images. Image Processing The PET/CT data was imported into the Siemens workstation, and read by two experienced PET/CT physicians independently. Standardized uptake value (SUV) was automatically calculated from manually drawn regions of interest (ROI) as follows: SUV = lesion concentration (kBq/ml)/injected FDG dose (MBq)/body weight (kg). SUV max was measured. MTV was measured automatically using an absolute threshold of SUV 2.5. SUV meanNL and SUV meanIVC were de ned as the mean SUV of three ROIs with the same area as the lesion placed at the level of normal liver tissue and normal inferior vena cava, respectively. The ratio of tumor to background SUV was calculated as follows: TNR NL = SUV max /SUV meanNL ; TNR IVC = SUV max /SUV meanIVC . TLG was calculated as follows: TLG NL = MTV × SUV meanNL ; TLG IVC = MTV × SUV meanIVC .
Quantitative data are expressed as ± s. T test was used for comparison. Enumeration data were expressed as percentage. The rst relapse after hepatectomy for HCC was taken as the endpoint. Recurrence-free survival (RFS) was de ned as the time between surgical resection and the rst relapse after surgery. The sensitivity, speci city, and cutoff values of SUV max , TNR NL , TNR IVC , MTV, TLG NL , and TLG IVC in predicting tumor recurrence were determined using receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC) was divided into three categories: low accuracy (0.50-0.70), medium accuracy (0.70-0.90), and high accuracy (0.90-1.00). Factors for the presence of MVI were carried out using univariate logistic regression analysis. Factors with statistically signi cant differences in univariate analysis were included in multivariate logistic regression analysis. RFS curves were constructed using the Kaplan-Meier method for the factors affecting recurrence. Differences between groups were analyzed by the univariate analysis. Factors with statistically signi cant differences in univariate analysis were included in the Cox multivariate regression model to further analyze their effects on postoperative recurrence of HCC. P < 0.05 was considered statistically signi cant.

Results
Clinical characteristics and PET/CT ndings Seventy-two patients with HCC were enrolled, including 58 males (80.56%) and 14 females (19.44%), aged from 18 to 77 years, with an average age of 56 ± 9 years. All patients were followed up for 12 to 49 months, with a median follow-up time of > 24 months. Twenty-one of the 72 patients recurred. The overall postoperative recurrence rate was 29.17% (21/72), and the RFS rate was 70.83% (51/72) ( Table 1).
Quantitative analysis of PET/CT metabolic parameters revealed higher SUV max , TNR NL , TNR IVC , MTV, TLG NL , and TLG IVC in recurrent cases (all p < 0.001) ( Table 2).     (Table 4). In particular, the largest AUC was found for MTV (AUC = 0.915), which provided a sensitivity and speci city of 92.8% and 85.5%, respectively, for predicting postoperative recurrence at a cutoff value of 96.66. Therefore, MTV has the greatest value in predicting postoperative HCC recurrence (p < 0.001). Moreover, the AUC was slightly larger for TNR IVC than for TNR NL , and for TLG IVC than for TLG NL (0.812 vs. 0.793, 0.892 vs. 0.874), and so did the sensitivity and speci city. Therefore, PET/CT parameters are more accurate in predicting postoperative HCC recurrence with inferior vena cava as a reference background than with normal liver tissue. predicting postoperative HCC recurrence (Table 5). In particular, TLG IVC has the highest HR (HR = 5.106, p = 0.001), which indicates that the higher the TLG IVC , the higher the probability of recurrence.  (Fig. 1). PET/CT images of typical cases were presented in Fig. 2. 0.002* HBV hepatitis B virus;SUV max maximum standard uptake value; TNR IVC maximum standard uptake value of the tumor -to-mean uptake value of the inferior vena cava ratio; TNR NL maximum standard uptake value of the tumor -to-mean uptake value of the normal liver tissue ratio; MTV Metabolic tumor volume; TLG IVC total lesion glycolysis with the inferior vena cava as a reference background; TLG NL total lesion glycolysis with the normal liver tissue as a reference background; CI con dence interval; NA not assessed; NS not selected as signi cant factor; CI con dence interval; NA not assessed; NS not selected as signi cant factor. *:p < 0.05

Discussion
The overall 5-year survival rate of liver cancer is only approximately 10%, and has not improved signi cantly in the past 20 years [9] . The most important reason is the high recurrence and metastasis rates of liver cancer even after satisfactory radical resection. The postoperative recurrence and metastasis of liver cancer is closely related to MVI [10][11] . Compared with traditional imaging modalities, 18 F-FDG PET/CT can evaluate tumor metabolism, and is more effective in predicting tumor prognosis [12] .
The results of this study showed that MVI, SUV max , TNR IVC , TNR NL , MTV, TLG IVC , and TLG NL were independent risk factors for predicting postoperative HCC recurrence. In particular, TNR IVC was the most relevant factor for predicting MVI, and MTV was the most valuable for predicting postoperative HCC recurrence. Patients with higher TLG IVC value were likely to have a higher recurrence risk. Moreover, the PET/CT parameters are more accurate in predicting postoperative HCC recurrence when obtained with inferior vena cava as a reference background than with normal liver tissue. PET/CT non-invasively provides more comprehensive predictive information on MVI and postoperative recurrence of liver cancer before surgery, which is of great signi cance for improving early treatment, prolonging survival, and increase survival rate.
Patients with MVI are more likely to experience recurrence of liver cancer after surgery [13][14] . Although many studies have shown that imaging features on contrast-enhanced CT, MRI and US [15][16][17] could predict the microvascular invasion of HCC. However, it is still di cult to diagnose MVI based on preoperative CT, MRI or US imaging modalities. It is worth noting that MVI cannot be directly observed by PET/CT before surgery due to limited resolution. However, as an imaging technique based on tumor biological activity, PET/CT has the potential to re ect the malignant biological ability of tumors invading blood vessels [18] . SUV max is the most commonly used semi-quantitative parameter of PET/CT. Several recent studies have suggested that a higher ratio of tumor SUV max to normal liver mean SUV on preoperative 18 F-FDG PET-CT could predict the MVI of HCC [19] .
Baek et al. [20] suggested that the increase of 18 F-FDG uptake in liver cancer was correlated with MVI and indicated a high risk of postoperative recurrence. Our results show that SUV max in ROI alone is not statistically signi cant for predicting MVI.
The reason may be that SUV max only represents the highest metabolic activity within the tumor, but not the overall metabolic activity of the lesion. Hence, it has a limited predictive value for the prognosis of liver cancer. At present, other PET/CT parameters, such as SUV mean , tumor-to-background ratio (TBR), MTV, and TLG, have been introduced to provide precise evidence for tumor treatment [21] . TBR represents the biological activity of tumors [22] . MTV represents the volume of the tumor with abnormally high uptake [23] . TLG is the product of MTV and SUV mean in ROI and re ects tumor burden [24] . Preliminary studies have been performed to investigate the prognostic value of one or two metabolic parameters in liver cancer before or after surgery. Lee et al. [25] demonstrated that the metabolic status of HCC re ected by 18 F-FDG PET was an important factor in predicting long-term survival. [1]Ahn et al. [19] believed that the TSUV max /LSUV mean ratio of > 1.2 was signi cantly correlated with MVI in liver cancer. Among the few studies on the use of PET/CT volumetric parameters in liver cancer, however, normal liver tissue was mainly taken as the reference background [26][27] . Single-parameter, single-background, and single- In predicting the postoperative recurrence of liver cancer, Lee et al. [28] found that MTV was closely related to progression-free survival and overall survival in patients with HCC. Univariate analysis in this study demonstrated that tumor size, MVI, SUV max , TNR IVC , TNR NL , MTV, TLG IVC , and TLG NL were all signi cant independent predictors of postoperative HCC recurrence, which is consistent with the results of some previous studies [29][30][31][32] . Moreover, higher sensitivity and speci city were obtained for predicting postoperative HCC recurrence with the inferior vena cava as a reference background than with normal liver tissue. The possible reason is that liver cancer is often complicated with fatty liver, hepatitis, and This study has the following limitations: (1) As this study is a retrospective study, there was a selection bias in the enrollment of patients. (2) The 72 patients in this study were not followed up long enough to analyze their 5-year and overall survival. (3) It is a single-center study with a small sample size, where patients with MVI and recurrence accounted for a small proportion. Therefore, the cutoff values established for the PET/CT parameters need to be con rmed by multicenter studies with a large sample size. (4) HCC recurrence was diagnosed by enhanced CT or MR instead of pathological examination. Hence, it might be underestimated in some patients, resulting in a diagnostic basis. (5) The joint diagnostic e cacy of clinicopathological indicators combined with PET/CT parameters for MVI and recurrence of HCC was not evaluated. In the future, studies will be performed with a larger sample size and longer follow-up time to investigate the value of PET/CT for predicting MVI and postoperative recurrence.

Conclusion
In summary, among the various PET/CT metabolic and volumetric parameters, TNR IVC , TNR NL , MTV, TLG IVC , and TLG NL are independent risk factors for predicting MVI and recurrence of HCC. Moreover, they are more accurate in predicting MVI and postoperative HCC recurrence before surgery when the inferior vena cava is used as a reference background as compared to normal liver tissue.