Image-matching digital macro-slide—a novel pathological examination method for microvascular invasion detection in hepatocellular carcinoma

Background Microvascular invasion (MVI) is a prominent risk factor of postoperative recurrence for hepatocellular carcinoma (HCC). The MVI detection rate of conventional pathological examination approaches is relatively low and unsatisfactory. Methods By integrating pathological macro-slide with whole-mount slide imaging, we first created a novel pathological examination method called image-matching digital macro-slide (IDS). Surgical samples from eligible patients were collected to make IDS. The MVI detection rates, tumor recurrence rates and recurrence-free survival were compared among conventional 3-Point and 7-Point baseline sampling protocols and IDS. Additionally, biomarkers to recognize MVI false negative patients were probed via combining conventional pathological sampling protocols and IDS. Receiver operating characteristic curve (ROC) analysis was used to obtain the optimal cutoff of biomarkers to distinguish MVI false negative patients. Results The MVI detection rates were 21.98%, 32.97% and 63.74%, respectively, in 3-Point, 7-Point baseline sampling protocols and IDS (p < 0.001). Tumor recurrence rate of patients with MVI negative status in IDS (6.06%) was relatively lower than that of patients with MVI negative status in 3-Point (16.90%) and 7-Point (16.39%) sampling protocols. Alpha-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) were selected as potential biomarkers to distinguish MVI false negative patients. Conclusions Our study demonstrated that IDS can help enhance the detection rate of MVI in HCC and refine the prediction of HCC prognosis. Alpha-fetoprotein is identified as a suitable and robust biomarker to recognize MVI false-negative patients in conventional pathological protocols. Supplementary Information The online version contains supplementary material available at 10.1007/s12072-022-10307-w.


Introduction
Hepatocellular carcinoma (HCC) is the fourth leading cause of cancer-related mortality worldwide and the leading cause of death among patients with cirrhosis [1]. HCC is characterized by an aggressive clinical course and dismal outcomes, with about two-thirds of patients diagnosed at advanced stage at first presentation. Up to 70-80% of HCC patients will develop disease recurrence within five years following initial curative treatments [1].
Microvascular invasion (MVI) refers to the presence of tumor cell clusters in a vascular lumen lined by endothelial cells under microscopic examination [2,3]. It has been repeatedly demonstrated to be a prominent risk factor of postoperative recurrence for HCC patient [4,5]. It is well known that pathological examination is the gold standard to diagnose and grade cancers via observing the morphological character, differentiation degree and growth pattern of tumor cells. Besides, pathological examination can detect some important biological factors around tumor tissues, including micro-satellites and MVI. Nevertheless, due to the limited scope of routine glass slides and randomness and bias of sampling, conventional pathological testing protocol has a tendency to under-report the incidence of MVI in HCC. Hence, constructing a novel pathological technique with greater MVI detection power is required.
Recently, based upon development of histological slide digitization and computational image processing, wholemount slides imaging (WSI) has been popularized in oncology studies. WSI refers to scanning a complete microscope slide, capturing many small high-resolution image tiles or strips and then montaging them to create a full image of a histopathological section. Many studies have highlighted the clinical significance of WSI in assisting pathologists investigating the whole spectrum of tumor biopsy specimens, and in identifying prognostic markers and histological subtypes of various cancers [6][7][8]. Macroscopic histological slide (macro-slide), different from routine small slides, expands to cover the whole section of tumor tissues and maintains the integrity of tumor specimens to the greatest extent. Macroslide has the advantage to exhibit more pathological features compared with conventional small slides.
In this study, by combining macro-slide and WSI technique, we create and first report the clinical utility of a novel pathological examination method called "image-matching digital macro-slide (IDS)" for MVI detection in HCC. We show that IDS has the capacity to remarkably increase MVI detection rates in HCC, guiding postoperative adjuvant therapies and surveillance protocols, thus reducing long-term recurrence rates. By analyzing IDS, we are able to get deeper insight into the comprehensive and most relevant features of tumors.

Patients
Consecutive patients who underwent radical hepatectomy of liver cancer at Eastern Hepatobiliary Surgery Hospital (EHBH) from October 2018 to December 2019 were enrolled. The follow-up date was censored on January 31, 2021. The inclusion criteria were as follows: (I) patients aged from 20 to 70 years old, (II) Child-Pugh class A-B7, (III) Eastern Cooperative Oncology Group (ECOG) performance score was 0-1, and (IV) underwent radical resection and had complete postoperative histopathological tissues. The exclusion criteria were as follows: (I) palliative-intend resection, (II) presence of extrahepatic metastasis or major vascular invasion, (III) underwent preoperative anti-cancer treatment, (IV) a history of other malignant tumors, and (V) pathologically confirmed intrahepatic cholangiocarcinoma (ICC) or combined HCC-ICC. The study was approved by institutional review board of our hospital. Written informed consent was obtained from patients for their data to be used for research purposes.

Pathological diagnosis standard
Two experienced pathologists identified HCC and MVI in all cases, and a third pathologist participated in the identification and gave the final result when there was ambiguity. HCC was diagnosed upon the Guidelines for the Diagnosis and Treatment of Hepatocellular Carcinoma (2019 Edition) [9]. MVI was defined as a cancer cell cluster composed of ≥ 50 cells in a microscopic vessel adjacent to the primary tumor [10,11].

Intraoperative marking
When intraoperative exploration, we determined the surgical resection range and marked the cutting edge and direction. It was recommended that the cross-section of the human body be taken as the sampling plane, marked with two asymmetric directions, and photographed for filing. Consequently, it was more convenient to distinguish the direction of the tumor specimen in vitro.

Specimen processing
We determined the section according to the maximum diameter of the tumor, half of which were taken according to the 3-Point and 7-Point baseline sampling protocols ( Fig. 1a;  Fig. 1b), the other half of which were taken using the IDS method (Fig. 1c). After the specimen was detached, 10% neutral formaldehyde should be injected as soon as possible (within 30 min) and fixed continuously for not less than 48 h. According to the direction of sampling and section, the thickness of the specimen was 0.5-1.0 cm, and the maximum size of the trimmed specimen was 6.0 cm × 5.0 cm. If the specimen was too large, it could be divided into several parts as required.

Grossing and paraffin embedding
The specimens were rinsed for formaldehyde with water for at least 2 h. Tissue automatic dehydrator was used to dehydrate specimens, which were then dried using slice drying machine. Next, the specimens were put into and soaked the liquid paraffin. The tissue specimens were embedded and shaped using paraffin embedding station and special paraffin embedding mold. The required sections were confirmed and adjusted again when shaping the tissues.

Serial sectioning and Hematoxylin-Eosin (HE) staining
When serial sectioning, the paraffin block was fixed with large holder. Paraffin slicing machine and matched special cutter head for big paraffin block were used to make sections with a thickness of 3 μm. After fished out, the slides were baked at 75 °C for 15 min with a baking machine, so that the tissues and the slides can be firmly adhered. The sections were stained with hematoxylin and eosin for histological examination.

Digital scanning and analysis
All tissue sections were scanned with a high-resolution slice scanner and corresponding image data were stored. The images were analyzed using the software matched with the scanner for parameters such as micro-metastases, stromal cell proportion, peripheral inflammatory changes, etc. The image data were matched with the scanned images, and the characteristics of pathological and imaging changes were analyzed again. Digital image signals in the region could be analyzed according to the research content.

Follow-up
Patients were followed up with laboratory tests including tumor biomarkers and liver biochemistry, abdominal ultrasonography, and contrast-enhanced CT once every 3-6 months. The diagnosis of intrahepatic recurrence was made by imaging findings alone if the tumor displayed typical enhancement characteristics; otherwise, the recurrent diseases were biopsied. Recurrence-free survival (RFS) time was calculated from the date of first surgery to the date when there was a clear evidence of recurrent disease.

Statistical analysis
Continuous variables were expressed as mean (standard deviation) or median (interquartile range) and compared using student's t test or Mann-Whitney U test according to the distribution of variables. Categorical variables were compared using chi-square test or Fisher's exact test. Survival curves were generated using the Kaplan-Meier method and compared using the log-rank test. Receiver operating characteristic curve (ROC) analysis was used to obtain the optimal cutoff value of AFP and PIVKA-II to distinguish MVI false negative patients. Clinical performance of 3-Piont, 7-Point and IDS to identify MVI actual positive was assessed by the sensitivity, specificity, and predictive values. Statistical analyses were performed by SPSS 26.0 software

The detailed process of IDS
Preoperative MRI showed that the tumor was located in the posterior lower segment of the right lobe of the liver. The size of the tumor was 72 mm × 58 mm × 55 mm. T2W1 showed slightly higher, equal, and uneven equalhigh signals. Scanning after enhancement, the hepatic artery phase demonstrated obvious uneven enhancement. In the portal phase and delayed phase, the relative signal attenuation was in line with the primary liver cancer (Fig. 2a). The preoperative evaluation was in accordance with the inclusion criteria. During the operation, the tumor was completely resected (R0 resection), and the whole specimen was taken to produce macro-slide. After hematoxylin-eosin staining, the macro-slide was observed at different positions with various magnifications (Fig. 2b). Then, the pathological macro-slide was matched with tumor specimen and imaging data to obtain WSI including MVI positions (Fig. 2c). This total process was named as IDS. The detailed preparation and production process of IDS is shown in the flow diagram (

The baseline characteristics and long-term outcomes of eligible patients
A total of 110 primary liver cancer patients were collected in this study. 19 patients who were pathologically diagnosed as intrahepatic cholangiocarcinoma were excluded, and 91 HCC patients were finally included. Using surgically resected specimens from 91 HCC patients, 145 thick slices were produced and examined. Conventional pathology covers tumor in all three extents. In this study, 300 sections in 3-Point, 664 sections in 7-Point, and 145 sections in IDS were examined and compared.
The baseline clinicopathological characteristics are shown in Table 1. Almost all patients had hepatitis B or C virus infection background. The percentage of hepatitis B virus infection was 96.7%, and only one patient did not have hepatitis. The liver function of patients was all graded as Child-Pugh class A. The median tumor size was 3.80 cm. In 4 HCC patients associated with portal vein tumor thrombus (PVTT), PVTT existed in the branches of main portal vein and could be resected radically. The shortest and longest follow-up times were 13 and 28 months, respectively. 24 (26.37%) patients developed disease recurrence and 5 patients (5.49%) died during follow-up.

MVI detection rates in 3-Point, 7-Point baseline sampling protocols and IDS
As shown in Fig. 3a, the detection rates of MVI were 21.98%, 32.97% and 63.74%, respectively, in 3-Point, 7-Point and IDS (p < 0.001). Patients with MVI positive status in 3-Point and 7-Point were all included in MVI positive status in IDS. The populations of 3-Point and 7-Point were not exactly the same (Fig. 3b). Among patients with MVI negative status in 3-Point and 7-Point, the two populations were partly different, but they all included MVI negative status in IDS (Fig. 3c). Therefore, in this study, the specificity and sensitivity of IDS on MVI detection were both 100%, while the specificity of 3-Point and 7-Point on MVI detection were both 100%, and the sensitivity of 3-Point and 7-Point on MVI detection were only 34% and 52%, respectively ( Table 2). The above results showed that IDS had superior sensitivity and specificity for the detection of MVI than 3-Point and 7-Point.

Tumor recurrence rates in three various pathological examination methods
To compare the impact of MVI, which was detected by 3-Point, 7-Point and IDS, respectively, on tumor recurrence, survival analyses were performed (Fig. 4a-c). Under the three methods, patients with MVI positive status were more likely to relapse (p < 0.001, p < 0.001, p = 0.001, respectively). We found that 12 (16.90%) HCC patients with MVI negative status in 3-Point had tumor recurrence, and 2 patients died due to disease progression. In 7-Point sampling method, 10 (16.39%) patients with MVI negative status had recurrence (relapse time range: 2.1-14.4 months), including 2 recurrence-related deaths. In IDS, 2 (6.06%) patients with MVI negative status recurred (relapse time: 13.13 and 14.40 months), and no deaths occurred. Next, we did subgroup analysis combining IDS with 3-Point and 7-Point. Results showed that patients with MVI positive status in both 3-Point and IDS, and in both 7-Point and IDS were most prone to recur. Patients with MVI positive status in IDS were more likely to relapse than patients with actual MVI negative status (p = 0.021, p = 0.016). (Fig. 4d-e). It is to say, 3-Point and 7-Point sampling protocols have a potential possibility to miss MVI, and patients with MVI false negative status in 3-Point and 7-Point are more likely to recur than patients with actual MVI negative status in all three MVI pathological testing methods.

Identification of potential biomarkers to distinguish MVI false negative patients in conventional pathological sampling protocols
In order to find out the clinicopathological characteristics of patients with missed MVI in 3-Point and 7-Point, IDS was matched with 3-Point and 7-Point. As shown in Table 3, there were significant differences in 3-Point in AFP, PIVKA-II, ALP, and tumor number. The medians of AFP between 3-Point negative IDS negative and 3-Point negative IDS positive groups were 6.10  Table 4, only AFP and PIVKA-II had significant differences in 7-Point. The medians of AFP between 7-Point negative IDS negative and 7-Point negative IDS positive groups were 6.10 (3.10, 20.30) ug/L and 160.55 (20.92, 1210.00) ug/L. The medians of PIVKA-II between 7-Point negative IDS negative and 7-Point negative IDS positive groups were 107.00 (33.00, 412.00) mAU/ mL and 460.50 (239.25, 3084.25) mAU/mL, respectively. It revealed that AFP and PIVKA-II could be potential biomarkers to distinguish MVI false negative patients in 3-Point and 7-Point.

Comparison of sensitivity and specificity of AFP and PIVKA-II in identifying MVI false negative patients
In order to study the sensitivity and specificity of AFP and PIVKA-II to identify MVI false negative patients, 71 patients with MVI positive and negative status in IDS but   (Fig. 5a-b). In 7-Point, when AFP was divided by a cutoff of 23.9 ng/mL, the AUC was 0.748 (0.617-0.879), the sensitivity was 0.75 (0.55-0.89), and the specificity was 0.79 (0.61-0.91). When using 267 mAU/mL as the cutoff of PIVKA-II, the AUC was 0.696 (0.558-0.833), the sensitivity was 0.75 (0.55-0.89), and the specificity was 0.67 (0.47-0.81) (Fig. 5c-d). Therefore, AFP is superior to PIVKA-II as a biomarker to distinguish MVI false negative patients.

Potential clinical utility of upper limit of AFP normal value to identify MVI false negative patients in conventional pathological sampling protocols
Because the cutoff values of AFP in 3-Point and 7-Point (22.5 and 23.9 ng/mL) were similar to the upper limit of normal value of AFP (20 ng/mL), the comparison and analysis were carried out using 22.5 ng/mL and 20 ng/mL, 23.9 ng/ mL and 20 ng/mL respectively. In 3-Point sampling method, the detection rate of patients with MVI false negative status was 68.4% (26/38), and there was a significant difference (Fig. 5e, Table 6). In 7-Point sampling method (Fig. 5f,   Fig. 3     of MVI in HCC was relatively low in previous studies, ranging from 15.0 to 33.8% after liver resection or transplantation [12][13][14][15]. Because of the limited range of ordinary glass slices and the varied criteria in pathological sampling protocol, the conventional pathological testing approach tends to undervalue the presence of MVI in HCC. Therefore, establishing a novel and practical pathological examination approach which can increase MVI detection rate is urgently needed.
With the breakthrough of WSI techniques, high-resolution macro-pathological digital pictures can be acquired. After combining the digital pathological macro-slide data with the imaging informational of patients, the pathological information including the distribution and characteristics of MVI can be accurately mapped to the corresponding images. This novel pathological examination technique is named as IDS. The major difference of IDS and WSI is the slide formats and sizes. IDS sufficiently takes advantage of macroslide as its bridging vehicle of pathological and imaging information.
The present study first reported the clinical utility of IDS in detecting MVI in HCC. In this study, we showed that IDS had a higher MVI detection rate than classical pathological examination approaches. In our study, the detection rate of MVI using IDS was significantly higher than conventional pathological approaches (63.7 vs. 33.0% or 22.0%), indicating that IDS had superior power in MVI detection compared with classical method.
Additionally, we provided evidence that in patients diagnosed with MVI negative status, the recurrence rate was significantly lower using IDS compared with conventional pathological method (6.1 vs. 16.4% or 16.9%), suggesting it can guide postoperative surveillance and adjuvant treatments through screening those patients with MVI false-negative status (that is, MVI was detected by IDS but undetected by conventional pathological testing) in conventional pathological methods, thus reducing long-term recurrence rates. Furthermore, IDS and conventional pathological testing could complement with each other mutually. Our findings demonstrated that patients detected as MVI positive in both IDS and conventional pathological protocols were most likely to occur disease recurrence in the near future following initial hepatic surgery. The recurrence rate of patients with MVI false-negative status was substantially higher than that of patients with actual MVI negative status.
We also selected AFP as a suitable and robust biomarker to identify MVI false-negative patients in conventional pathological protocol. Approximately 70% of HCC patients who were diagnosed as MVI negative status in conventional pathological examination may be MVI positive status in IDS testing. This encouraging result showed that IDS outperformed conventional pathological method in MVI detection for patients with abnormal AFP. For patients with normal AFP level (≤ 20 ng/mL) and MVI negative status in conventional pathological method, IDS was also a crucial way to confirm the actual MVI status.
Previously, Sheng et al. [16] proposed a standardized pathological proposal for evaluating MVI of HCC. They concluded that the MVI detection rate determined by sevenpoint sampling protocol (SPSP) was significantly higher than that determined by 3-point sampling method (47.1 vs. 34.5%, p = 0.048). Nevertheless, there was no marked difference in MVI detection rate between SPSP and 13-point sampling method (47.1 vs. 51.3%, p = 0.517). Therefore, we suppose that it is a futile effort to simply increase the sampling numbers beyond 7 points in conventional pathological slides. Contrarily, IDS is a useful and promising pathological technique to further increase MVI detection rates on the basis of conventional pathological testing method, thus guiding the subsequent treatment strategies of patients with MVI.
The potential application of IDS in clinical practice goes beyond MVI detection. In our on-going study, the locations of MVI can be accurately positioned using IDS, providing a reliable way to explore the distribution patterns of MVI in HCC and other tumors. Additionally, IDS can facilitate in judging the degree of tumor necrosis and observing the infiltration of inflammatory and immune cells, which can improve the efficacy evaluation of non-surgical treatments, such as transarterial chemoembolization, tyrosine kinase inhibitors and immune checkpoint inhibitors.
This study has some limitations. First, the small sample size and observational nature of this study may potentially affect the results. Second, all of the patients included in this study had a background of HBV infection. Whether IDS is applicable to patients with other etiologies of HCC needs further investigation. Third, this study is based upon our antigen, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT γ-glutamyltransferase, GLU glucose, ALP alkaline phosphatase, WBC white blood cells, RBC red blood cells, PLT platelet, PT prothrombin time, PVTT portal vein tumor thrombus

Conclusions
This study demonstrates that imaging matching digital macro-slide (IDS) first implemented by our team can help improve the detection rate of MVI in HCC and refine the prediction of HCC prognosis. It highlights the importance of establishment of a novel pathological algorithm of IDS to study HCC more comprehensively and allow for a better understanding of number and distribution of MVI under microscope. In the future, combining with artificial intelligence-driven approaches, IDS has the opportunity to be the standardized pathological examination method in all types of cancers.