Combined Texture Analysis of MR images and Glypican-3 to Identify the Differentiated Degree of Hepatocellular Carcinoma: a retrospective study

BACKGROUD To explore the clinical value of texture analysis of MR images (multiphase Gd-EOB-DTPA-enhanced MRI and T2 weighted imaging (T2WI) and the glypican-3 (GPC-3) to identify the differentiated degree of hepatocellular carcinoma (HCC). Each participant performed Gd-EOB-DTPA-enhanced Texture classification. The performance of texture operating There were no statistically significances for the expression of GPC-3 between poorly-, well- and moderately-differentiated HCC. The area under the curve (AUC) of the combined texture features between poorly- and well-differentiated HCC, poorly- and moderately-differentiated HCC, moderately- and well-differentiated HCC was 0.812, 0.879 and 0.808 respectively. With GPC-3 combined, the AUC was increased to 0.868, while accuracy was decreased, in poorly- verse well-differentiated HCC, and the AUC and accuracy were the same as those without GPC-3 combined in poorly- verse moderately-differentiated HCC. Although the AUC was increased to 0.818 with GPC-3 combined in moderately- verse well-differentiated HCC, there were no statistical significance for the value change (p>0.05). Texture analysis of Gd-EOB-DTPA-enhanced MRI and T2WI are valuable in identifying the differentiated degree of HCC. There is no significant effect of GPC-3 in identifying the differentiated degree of HCC, suggesting the promising value of texture analysis of MR images in the precise presurgical diagnosis of HCC. HCC, Hepatocellular carcinoma; TACE, transcatheter arterial chemoembolization; GPC-3, Glypican-3; OS, overall survival; AFP, alpha fetoprotein; ALT, alamine aminotransferase; AST, aspartate transaminase; T2WI, T2 weighted imaging; ROC, receiver operating characteristic; AUC, area the TBIL, bilirubin; Hepatic

grouped into poorly-differentiated HCC, moderately-differentiated HCC and well-differentiated HCC. According to the reports, the overall survival rate of the patients with moderately-differentiated and well-differentiated HCC was higher than that of the patients with poorly-differentiated HCC, while the recurrence rate was lower [8,9].
As we known, a precise pre-surgical evaluation of differentiated degree of HCC might affect the individual treatment schedule [10].
Currently, aspiration biopsy was the most common method to get the information of histopathology before surgery. However, it was criticized by many researchers due to its invasiveness and the probability of seeding metastasis [11,12]. Recently, many studies suggested the image characteristics of tumor might predict the differential degree of the HCC. For example, there were some reports found that the low density/intensity of HCC on the portal phase of CT and hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI might help to identify the differentiated degree of HCC [13,14].
Glypican-3 (GPC-3), a special marker of HCC, has been used to identify the HCC from other tumors and to evaluate the prognosis. GPC-3 is a member of the glypican family, which influences cell growth, differentiation, and migration [15]. GPC-3 is highly expressed in HCC, but not in cholangiocarcinoma or normal hepatic cells and is low express in well differential HCC [16,17]. Previous studies have shown that higher GPC-3 expression level in HCC is a risk factor for shorter overall survival (OS) and levels of GPC-3 expression in poorly-differentiated tumor cells were higher than moderately-and well-differentiated HCC [15,18].
Texture analysis was an established technique，which was beneficial to diagnoses, by extracting a large amount of texture information from medical images [19]. Until now, texture analyses have been used in identifying the differentiated degree and characteristics of tumor, evaluating the therapeutic effect, etc [20][21][22].
To the authors' knowledge, the texture analyses have not been used in identifying the differentiated degree of HCC yet. Thus, our aim of the present study is to evaluate the accuracy of the texture analysis of MR images and the effect of GPC-3 in discriminating the differentiated degree of HCC.

Patients
The present study received ethical approval from the Medical Ethics Review Committee of our institution and the relevant informed consent form was obtained in accordance with the Declaration of Helsinki. 104 participants were enrolled from 2015 to 2019, according to the following criteria:1) pathologically proved as HCC after hepatectomy or aspiration biopsy; 2) inpatients who have comprehensive clinic materials; 3) performed preoperative Gd-EOB-DTPA-enhanced MRI. The clinic data of the 104 participants were recorded in the Table 1, containing age, gender, alpha fetoprotein (AFP), alamine aminotransferase (ALT), aspartate transaminase (AST), ALT\AST, total bilirubin (TBIL), direct bilirubin and indirect bilirubin. The histopathlogical data of GPC-3 was obtained in 51 participants.
Exclusion criteria included:1) participants have been treated (transplantation, resection, ablation or embolization) before MR examination; 2) clinical data or pathological results were incomplete; 3) the lesions were not clearly displayed on the images due to the artifact.

Image analysis
The MRI images were reviewed in the picture archiving and communication system (PACS). Experienced radiologists, who were blinded to the pathological results, evaluated the MRI imaging features of the HCC. The imaging features of MRI (arterial enhancement, capsule appearance, the intensity of HBP，the margin and diameter of tumor, intra-lesion fat, intra-tumor vessel and etc.) were selected referring to the Liver Imaging-Reporting and Data System (LI-RADS 2017) (https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/LI-RADS) [23].

Texture analyses and features selection
MaZda software (version 4.6, quantitative texture analysis software, available from http://www.eletel.p.lodz.pl/mazda/) was used for texture analysis. All images were transformed into Bitmap (BMP) format considering for the application compatibility of MaZda. An experienced radiologist manually portrayed the region of interest (ROI) of the lesion on the slice which contained the maximum proportion of tumor. 104 ROIs (one ROI for each patient) on HBP images were analyzed firstly. Subsequently, the ROIs were copied onto T2, AP and EP images. Then, texture features were extracted and analyzed. Finally, the texture features were grouped into grey-level histogram, the grey-level co-occurrence matrix (GLCOM), the grey-level run-length matrix (GLRLM) and wavelet transform. More detailed texture features are listed in Table 2.
Feature selection algorithms included Fisher coefficient, mutual information [MI], and classification error probability combined with average correlation coefficients [POE + ACC]. 10 texture features were extracted by each of these algorithms. In order to enhance the discriminability, these three methods were combined, called "FPM", by which 30 texture features were extracted in total.

Histopathological analysis
Histopathological evaluation was available after hepatectomy or aspiration biopsy for the lesions. The specimens were routinely prepared with 4% formaldehyde. The specimens were evaluated by two experienced pathologists who were blind to MRI information. The eight slices of each lesion were analyzed and evaluated with slices ranging from 0.3 cm to 2.0 cm depending on the size of the lesion. The Edmonson method was used to categorize all the specimens. According to the differentiation degree of tumor cells, HCC were categorized into grades I to IV. Grades I and II were defined as well-differentiated HCC, grade III was defined as moderately-differentiated HCC, and grade IV was defined as poorly-differentiated HCC. The specimens were stained with GPC-3 antibodies. The results of immunohistochemical staining were considered positive if more than 10% of the tumor cells showed cytoplasmic staining, otherwise the results were considered negative.

Statistical analysis and misclassification rate
The statistical analysis was performed using Statistical Product and Service Software (SPSS ver. 20.0, Chicago, IL). In present study, the continuous variables which are abnormal distribution, such as age, ALT, AST, ALT\AST and texture features, were compared with Mann-Whitney U test. Categorical variables were compared using Pearson Test when the sample size is over 40 and the minimal expected frequency is over 5. Otherwise, the correction formula of chi-squared test will be chosen. And the R×C table was used when the dependent variable is over 2. The value of p<0.05 were considered statistically significant. And Bonferroni correction was used to adjust P values. In order to evaluate the diagnostic accuracy of texture features derived from T2, HBP, AP, and EP, the ROC analyses were performed and the area under the curve (AUC) were calculated by MedCalc (MedCalc statistical software, ver.15.8).
The B11, a module of MaZda (version 4.6), provides following four pathways, Principal component analysis (PCA), linear discriminant analysis (LDA), nonlinear discriminant analysis (NDA) and Raw data analysis (RDA), to classify and analyze the texture features. The b11 implements 1-NN classifier for non-linear supervised classification [24]. The misclassification rate is defined as total false samples divided by the total samples and the ratio indicates that the estimated group is different from the observed group. According to the misclassification rate, the classification results were separate into four levels: excellent (misclassification rates ≤10%), good (10% <misclassification rates ≤20%), moderate (20% <misclassification rates ≤30%), fair (30% <misclassification rates ≤40%), and poor (misclassification rates >40%) [25].

Clinical data
There were 37 patients with poorly-differentiated HCC, 43 with moderately-differentiated HCC, and 24 with well-differentiated HCC in present study. As showed in Table 1, there were no significant differences for age and gender among the groups (p>0.05). There were significant differences for AFP and ALT value between the poor-and well-differentiated HCC (p=0.001, 0.006, respectively). The ALT was statistically different between well-and moderately-differentiated HCC (p=0.008). 51 participants were with GPC-3, among which, 20 were with poorly-differentiated HCC, 20 with moderately and 11 with well-differentiated HCC.
There were no statistical significances for expression of GPC-3 between poorly-, well-and moderately-differentiated HCC, as Table 1 showed.

MRI feature evaluation
The MRI imaging features of l04 patients are demonstrated in Table 3. As the table showed, the tumor size were statistically different between poorly-and moderately-HCC, with p=0.014(p<0.017). However, no statistical difference in the margin of tumor, the status of tumor's capsule and liver cirrhosis, the HBP hypointensity, intratumoral vessel, intralesion fat, rim-enhancement AP and lymphadenectasis, between poorly-, moderately-and well-differentiated HCC.

Texture analysis and tissue classification
As showed in Table 4, 262 texture features derived from T2, HBP, AP and EP images were obtained and categorized into histogram (n=10), GLCOM (n=220), GLRLM (n=20) and wavelet transform (n=12) [26]. The frequency of each feature category extracted by FPM were showed for T2-weighted images and each phase of Gd-EOB-DTPA enhanced images among poorly-differentiated, well-differentiated and moderately-differentiated HCC. The GLCOM-based texture features were most frequently extracted with three phases for poorly-verse well-differentiated HCC, poorly-verse moderately-differentiated HCC and well-verse moderately-differentiated HCC.
The tissue classification results were demonstrated across the T2, AP, EP and HBP in Table 5. The misclassification rate of NDA was excellent for each phase of the three groups, with the miscalssification rate ranging from 3.33% to 14.93%. The misclassification rate of LDA was rank secondly to NDA, with the classification rate range from 4.92% to 33.75%. Both of the mislassification results of RDA and PCA were fair or poor. The phase of enhanced-MRI may be an important factor which can impact the classification rate of texture analysis when compared poorly-, moderately-and well-differentiated HCC.

ROC-analysis
All of the AUC of the statistically significant texture features were calculated. The ROC curves of the best combined diagnoses were demonstrated in Figure 1-3. As showed in Figure 1, the combine AUC value (combining texture features from T2, AP and EP) was 0.812, higher than that of any single texture feature from each phase, to differentiate poorly-from well-differentiated HCC (accuracy=0.77). As showed in Figure 2, the combine AUC value was 0.879 (accuracy=0.85), to differentiate poorly-from moderately-differentiated HCC, and as showed in Figure 3, the combined AUC value was 0.808 (accuracy=0.746) to differentiate moderately-from well-differentiated HCC.   The ROC analyses of combined GPC-3 and texture analyses were demonstrated in Table 6. "COMBINE" presents the combination of texture features derived from different phases. As showed in Table 6, the COMBINE AUC value was higher than that of any single texture feature from each phase. Additionally, the AUC value of combined GPC-3 and texture analyses was increased from 0.818 to 0.868 while the accuracy was decreased, when compared poorly-with well-differentiated HCC. The AUC and accuracy of combined GPC-3 and COMBINE was the same as those of COMBINE AUC (AUC 0.880, accuracy 0.800), in poorly-verse moderately-differentiated HCC. When moderately-verse well-differentiated HCC，the AUC was increased from 0.777 to 0.818 with GPC-3 combined, but there were no statistical significance for the value change (p=0.508).

DISCUSSIONS
As previous studies showed, the diameter of HCC was an important factor to predict the pathological grade of HCC. Lee et al [26] and Martins et al [27] suggested that the diameter of most moderately-differentiated HCC was larger than well-differentiated HCC. Our present study also proved the opinion that the diameter of poorly-differentiated HCC was larger than that of moderately-differentiated and well-differentiated HCC. However, there was no statistical difference of diameter between poorly and well-differentiated HCC in present study, which was not in consistence with the Martins'. Such differences may be due to the heterogeneity of tumor cells themselves and individual differences of tumor growing pattern, as well as the limited sample size.
The differential degree of HCC is the most important factor that affect the prognosis of the patients. In this study, the patients were grouped into poorly, moderately and well-differentiated group based on the histopathological grade and explored whether the texture features can successfully differentiate the subtypes of HCC. Texture analysis is a method that can quantize the information provided by the images. Some studies verified that texture analysis had the potential to identify the histopathological type of neoplasm, such as the breast cancer and renal tumor [25,28]. However, there are no researches yet to explore the value of texture features derived from mlutiphase of Gd-EOB-DTPA-enhanced MRI and T2WI in predicting the histopathological grades of HCC.
In recent years, researchers gradually realized that the substantial quantitative features was increasingly important in the tumor diagnoses, not merely the application of qualitative features such as margin, signal intensity, capsule of the tumor and so on [29]. Mazda is a software package which provides a complete path for quantitative analysis of image texture. It includes image analysis, texture features extraction, data classification, analysis automation and other functions [24]. Substantial information obtained by Mazda, might differentiate the pathological grade of tumor. Previous study analyzed the texture features to predict the OS of the patients with advanced HCC [30]. Our study attempted to identify the histopathological grade by texture analysis.
B11 module provides four procedures, RDA, PCA, LDA and NDA, to analyze the selected thirty features. In present study, the classification rate of NDA was excellent. It suggested that texture analysis was a reliable method to identify the poorly-, moderately-and well-differentiated HCC. Although LDA was recommended as an optical method, NDA was demonstrated much preciser than LDA in present study, which was in consistent with Li Y's study [31]. This may due to the non-linearity of the clinical data which was obtained in a random way.
And the inconformity of the misclassification rate from the texture analysis of different image sequences, may result from the different histological components and enhancement patterns among the subtypes of HCC [25].
The GLCOM-based features which describe the spatial dependence of gray value in image were most frequently extracted than other texture features of other categories regardless of the phase of MRI and groups [31,32]. It may imply that the different pathological grade can impact the gray value of the image. But it also may be caused by the tremendous number of GLCOM (n=220) [25]. Besides, the texture features of GLRLM is secondly selected by texture analysis which demonstrates the pixel runs with the same grey level values in a given direction and depicts intensity homogeneity in given direction [31]. The result may depict that the intensity homogeneity between poorly-, moderately-and well-differentiated HCC is different.
The GLCOM-based features generated from AP was noticeably different between groups. As previous study showed, HCC with a higher differentiated degree was prone to have lower arterial supply. The arterial supply differences may lead to the differences of texture features derived from AP [33]. The Correlat depicts the similarity of gray value between the adjacent pixels in the horizontal and vertical directions and the SumAverg indicates the average gray value of the pixels and reflects the degree of brightness of the image. Our findings suggested that the signal intensity of poorly-differentiated HCC was slightly higher than that of well-differentiated HCC on AP images, which was somewhat approved the abundant arterial supply of poorly-differentiated HCC.
As previous studies showed, GPC-3 was hardly expressed in most of the tissues. However, GPC-3 was a reliable mark of HCC [34,35].
There were some studies suggested that the expression of GPC-3 of poorly-differentiated HCC was higher than well-and moderately-differentiated HCC and the OS was shorter [16,17]. But there was no statistical significance of the expression of GPC-3 between poorly-, moderately-and well-differentiated HCC in present study. The small sample size was supposed to be the reason of this discrepancy.
Additionally, as showed in Table 6, the diagnosis effect was not improved with the combination of GPC-3 and texture features. It suggested that the texture analysis was a promising and reliable method in identifying the differentiated degree of HCC, even without GPC-3, an index from pathological tissue.
There were some limitations in our study. Although we adopted strict inclusion and exclusion criteria in this retrospective study, selection bias was still inevitably. Second, the sample size was relatively small which need to be enlarged in the future study. Third, the ROI (tumor contour) was manually delineated on the slice containing the maximum diameter, which led to the lack of the 3D information of the tumor.
However, the 3D information of the HCC rarely impacted our clinical diagnosis.
In conclusion, the texture analysis of multiphase Gd-EOB-DTPA-enhanced MRI and T2WI are noninvasive and reliable quantitative technique to differentiate the subtypes of HCC. There is no significant effect of GPC-3 in identifying the differentiated degree of HCC, suggesting the promising value of texture analyses of MR images in the precise presurgical diagnosis of HCC. 6、 Each author contributed substantially to the study design, data acquisition, analysis and interpretation, and the writing of the manuscript.
Each author gave final approval of the version to be published, and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. 7、Not applicable acknowledgement.     A: poorly-differentiated HCC, B: well-differentiated HCC, C: moderately-differentiated HCC; AP: arterial phase, EP: equilibrium phase images, and HBP: hepatobiliary phase.