Predicting Model for Tumor Budding Status using Radiomics Features of 18F-PET/CT and in Cervical Cancer

Objective The aim of this study was to compare radiomics feature on 18 F-FDG PET/CT and intratumoral heterogeneity according to tumor budding (TB) status and to develop predicting model for TB status using radiomics feature of 18 F-FDG PET/CT in patients with cervical cancer. Materials and methods A total of 76 cervical cancer patients who performed radical hysterectomy and preoperative 18 F-FDG PET/CT were included. We assessed the status of intratumoral budding (ITP) and peritumoral budding (PTB) in all available hematoxylin and eosin-stained specimens. Three conventional metabolic parameters and a total of 59 features were extracted and analyzed. Univariate analysis was used to identify signicant metabolic parameters and radiomics ndings for TB status. Predicting model for TB status was built by the LASSO regularization. The univariate analysis lead to the identication of 2 signicant metabolic parameters and 12 signicant radiomic features according to ITB status. Among these parameters, only compacity was remained in multivariate analysis for ITB status (odds ratio. 5.0047; 95% condence interval, 1.1636 – 21.5253; p = 0.0305). Five radiomics features (Kurtosis, Compacity, Short-Zone Low Gray-level Emphasis, Coarseness, Low Gray-level Run Emphasis) were selected by the LASSO regularization and the predicting model for ITB status had a mean area under curve of 0.810 in training dataset and 0.794 in validation dataset. Radiomics features on 18 F-FDG PET/CT was associated with ITB status. The predicting model using radiomics features successfully predicted TB status in cervical cancer. The predicting models for ITB status may contribute to personalized medicine in the management of cervical cancer patients.


Introduction
Tumor budding (TB) is de ned as a single neoplastic cell or cell cluster of up to four neoplastic cells at the invasive front of the tumor (peritumoral budding, PTB) or within the tumor mass (intratumoral budding, ITB) [1]. Several studies have demonstrated that TB was associated with lymphovascular invasion (LVI), lymph node metastasis, disease recurrence, and an unfavorable survival outcome, especially in colorectal cancer [2], esophageal carcinoma [3], and head and neck cancer [4]. Recently we evaluated prognostic roles of TB and correlation between TB and conventional pathologic parameters in gynecological cancers [5,6]. TB was associated with deep depth invasion, higher International Federation of Gynecologic Obstetrics (FIGO) stage, LVI, and lymph node metastasis in endometrial cancer [5].
Moreover, high TB was an independent prognostic factor for predicting survival outcomes in cervical cancer [6].
Currently, F-18 uorodeoxyglucose positron emission tomography/computed tomography ( 18 F-FDG PET/CT) has been widely used to detect lymph node involvement, distant metastasis, and recurrent disease in cervical cancer [7]. Various metabolic parameters of 18 F-FDG PET/CT have been reported as prognostic factors including maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG). Radiomics studies, which represent intratumoral heterogeneity have emerged as a new and exciting study area in recent years. The measurement of texture indices from tumor 18 F-FDG PET/CT images has been recently proposed as an adjunct to predict tumor response to therapy. There is emerging evidence that intratumoral metabolic heterogeneity on pre-treatment 18 F-FDG PET/CT might be a predictor of tumor recurrence after treatment in patients with lung cancer, esophageal cancer, head and neck cancer, and cervical cancer [8][9][10][11]. Furthermore, recent studies showed 18 F-FDG PET/CT radiomics using various textural features were potential biomarker to predict tumor recurrence and lymph node metastasis [12][13]. Tumor heterogeneity is de ned by the presence of different cell subpopulations or clones and has a fundamental role in growth, progression, and therapeutic resistance.
Tumor hypoxia, angiogenesis, necrosis, brosis, cell proliferation, and in ammation process play a role in tumor heterogeneity [14]. Epithelial-mesenchymal transition (EMT) of primary tumor tissues may lead loss of cell-to-cell adhesion and individual cells or small groups of cells acquired the ability to migrate and invade through the surrounding tissues. Moreover, TB may lead to a more aggressive clinicopathologic characteristics through similar mechanism of EMT such as increased extracellular matrix degradation, increased migration, and loss of cell adhesion [15]. These EMT processes are accompanied by changes in cell morphology [16] and may lead to change tumor heterogeneity. However, there has been no study which evaluate correlation between metabolic parameters and radiomics nding of 18 F-FDG PET/CT and TB in cervical cancer. So, we hypothesized TB may associate with higher metabolic parameters in 18 F-FDG PET/CT because of aggressive behavior of TB and radiomics nding may different according to TB status because TB is individual cells and small groups of cells.
The aim of this study was to compare radiomics feature on 18 F-FDG PET/CT and intratumoral heterogeneity according to TB status and to develop predicting model for TB status using radiomics feature of 18 F-FDG PET/CT in patients with cervical cancer.

Patients
After approval of the Institutional Review Board of Kyungpook National University Chilgok Hospital (KNUCH 2020-03-011), we reviewed the archival medical records and hematoxylin and eosin (H&E)stained slides of early-stage and locally advanced cervical cancer patients. The need for informed consent was waived because of the retrospective nature of the study. A total of 136 patients who underwent radical hysterectomy with pelvic and/or paraaortic lymphadenectomy for the treatment of early-stage and locally advanced cervical cancer were included. Among the 136 patients, 76 patients performed preoperative 18 F-FDG PET/CT and were enrolled in this study. The enrolled patients were semirandomly divided into a training dataset (51 patients) and a test set (25 patients) by using the "doBy" R package while preserving the distribution of ITB status. Patients with a history of preoperative chemotherapy, radiotherapy, or synchronous malignancies were excluded. The patients were clinically staged according to the 2009 International Federation of Gynecologic Obstetrics (FIGO) staging system [17].

Histopathological evaluation
Specimens were examined from multiple sections of the whole tumor areas and were stained with hematoxylin and eosin (H&E). For each case, all available specimens were independently reviewed for the detailed histopathological features and the quantitative assessment of TB by two pathologists (J.Y.P and J.Y.P) in a blinded manner without information about the clinicopathological data and outcomes.
The pathological parameters included tumor size, FIGO stage, histological subtype, deep stromal invasion, LVI, parametrial invasion, lymph node metastasis, and the number and distribution of TB. TB was de ned as an isolated single cancer cell or small cell clusters composed of 4 tumor cells or less located at the advancing edge (PTB) and within the tumor area (ITB). 18 F-FDG PET/CT Image Acquisition All patients fasted for at least six hours, and their blood glucose levels were determined before the administration of 18 F-FDG. Patients with blood glucose levels higher than 150 mg/dL were rescheduled for a later examination, and treatment was administered to maintain a blood glucose concentration of less than 150 mg/dL in all participants. Patients received intravenous injections of approximately 5.2 MBq of FDG per kilogram of body weight and were advised to rest for one hour before undergoing the acquisition of 18 F-FDG PET/CT images. The 18 F-FDG PET/CT scans were performed using Discovery 600 (GE Healthcare, Chicago, IL, USA). Before the PET scan, for attenuation correction, a low-dose CT scan was obtained without contrast enhancement from the skull base to the thigh while the patient was in the supine position and breathing quietly. PET scans were also obtained from the skull base to the thigh at 2.5 min per bed position. PET images were reconstructed with a 128 X 128 matrix and an ordered-subset expectation maximum iterative reconstruction algorithm.

Image Interpretation and PET Image Analyses
The 18 F-FDG PET/CT images were interpreted by two experienced nuclear medicine physicians, and a nal consensus was achieved for all patients. A positive nding was de ned as any focus with increased FDG uptake as compared with the surrounding normal tissue. Foci of FDG uptake due to normal physiology or benign variants, such as muscular exercise or an infectious pulmonary in ltration, were excluded from the analysis.
All image analyses were performed using the Advantage Workstation 4.5 software (GE Medical Systems, Waukesha, WI, USA). The primary tumor lesion was delineated by the volume of interest using an isocontour threshold method based on the SUV, and metabolic PET parameters were assessed. SUVmax values were based on body weight and were calculated using the following formula: SUVmax = maximum activity in the region of interest (MBq/g)/[injected dose (MBq) ÷ bodyweight (g)]. SUVmax was designated as the highest value of SUVmax of the primary tumor. The MTV was determined as the volume of voxels with a threshold SUV of the mediastinal blood pool, because the mediastinal blood pool has been regarded as the preferred site for measuring background activity [18]. The mean SUV of the mediastinal SUV values was determined by drawing a region of interest over contiguous slices on descending aorta carefully excluding the walls from the region of interest. The mean SUV of the mediastinal background plus 2 SDs was used as the threshold to automatically calculate MTV [17]. The TLG was calculated as the MTV multiplied by the SUVmean of the lesion. The MTV and TLG were also obtained for the primary tumor.

Statistical analysis
The differences between subsets were evaluated with a Student's t-test or Mann-Whitney test, and differences between proportions were compared with the chi square test or Fisher's exact test. To identify an optimal cutoff value of metabolic parameters and radiomics nding of 18 F-PET/CT for the prediction of ITB status, ROC curve analysis was performed. Multiple logistic regression analysis was used to evaluate metabolic parameters and radiomics nding of 18 F-PET/CT for ITB status. Estimated odds ratios (ORs) with 95% con dence intervals (95% CIs) were presented. All statistical tests were 2-sided, and a p-value of <0.05 was considered signi cant. Statistical analysis was performed using SPSS software version 22.0 (SPSS, Chicago, IL, USA) and Medcalc version 15.4 (Medcalc Software, Ostend, Belgium), and R version 3.6.3 (R foundation for Statistical Computing, Vienna, Austria). The R packages "caret", "glmnet", "MASS", and "pROC" were used for analysis.

Radiomics analysis
Radiomics features were extracted using the LIFEx package (http://www.lifexsoft.org) [19]. LifeX was set up using the following input parameters for calculation of features: 64 Gy levels to resample the ROI content which was performed in absolute terms between a minimum of 0 and a maximum of 20 [20]. A total of 59 features were extracted from the analysis of the volumes inspected. These indices included conventional parameters, shape and size features, histogram-based features, second and high orderbased features. The correction for the partial volume effect was not applied. In the analysis were included all primary tumor lesions, irrespectively of their volume but LifeX calculates the shape and size indices as Each feature value was normalized using z-score normalization (z = (x − mean(x))/SD(x)) (standard deviation [SD])). Feature selection process consisted of two steps in the training dataset. First, t-test was performed to screen potential features throughout radiomics features and conventional metabolic parameters. Only features with p < 0.05 were considered signi cant and entered further selection step. Next, the least absolute shrinkage and selection operator (LASSO) regression was used to select key features to build prediction model, with a 3-fold cross validation.
After feature selection process, the prediction models were constructed using a random forest (RF), a support vector machine (SVM), and a neural network (NN) using training dataset. The constructed model performance was validated independently in the test dataset by the area under the receiver operating characteristic (ROC) curve. The R packages "randomForest", "kernlab", and "neuralnet" were used for building prediction model.

Clinicopathologic features and treatment outcomes
The clinicopathologic characteristics of the study participants are listed in  (Table 3).

Discussion
In this study, radiomics features of 18  Radiomics is relatively new and evolving eld in medical imaging in which many features are extracted from medical images analysis and interpretation using bioinformatic approaches [14]. Furthermore, radiomics precision medicine and tumor heterogeneity have become one of the hottest topics in the oncological medical area in recent years [21]. At the biological level, it has been recognized that heterogeneity of tumor microenvironment might be re ected in medical images, with respect to cellular density, proliferation, angiogenesis, hypoxia, receptor expression, necrosis, brosis, and in ammation and these factors might contribute to a more aggressive phenotype and poor treatment responses [22]. Therefore, radiomics signature might represent segmentation of tumor subregions with different biological characteristics and contribute to treatment response and prognosis.
In colorectal cancer, high ITB correlated with higher tumor grade, higher pT stage, lymphatic invasion, vascular invasion, nodal metastasis, and shorter survival time [23]. ITB may lead to a more aggressive clinicopathologic characteristics through similar mechanism for PTB such as increased extracellular matrix degradation, increased migration, and loss of cell adhesion [15]. So, we hypothesized ITB may associate with higher metabolic parameters because of aggressive behavior of ITB. Moreover, radiomics nding may different according to ITB status because TB is individual cells and small groups of cells owing to loss of cell adhesion.
To date only one study recently demonstrated correlation between tumor cancer cell metabolism and morphologic features of aggressiveness as assessed by microscopy such as TB [24]. MTV was higher in TB group than in non-TB group with marginal signi cant (p = 0.06) in laryngeal and pharyngeal carcinoma [25]. In this study, SUVmax was signi cantly different according to ITB status (p = 0.0406). Moreover, SUVmax and TLG were associated with ITB status in univariate logistic regression analysis (p = 0.0146 and p = 0.0154, respectively). More aggressive nature of ITB may represent higher metabolic parameters in 18 F-FDG PET/CT.
EMT of primary tumor tissues may lead loss of cell-to-cell adhesion and occurrence ITB. So, ITB may lead segmentation of tumor subregions with different biological characteristics and these may contribute tumor heterogeneity. However, no study has been reported about correlation between ITB and radiomics ndings. In this study, 12 features were associated with ITB status and compacity was powerful biomarker which represent ITB status in multivariate logistic regression analysis. Previous study demonstrated that among the radiomics features, the most signi cant covariate was compacity for predict local control and survival in hepatocellular carcinoma [26].
Previous our study showed that tumors with high TB were signi cantly associated with LVI, deep stromal invasion, parametrial invasion and lymph node metastasis in cervical cancer [6]. Preoperative prediction of TB status may helpful to personalized medicine such as decision of radicality of surgery or extent of lymphadenectomy. However, TB status was nally determined by postoperative surgical specimens. To date, there was no modality with estimate TB status preoperatively. Therefore, predicting models were constructed for ITB status using both conventional metabolic parameters and radiomics features from 18 F-FDG PET/CT scans. Among the 48 features which were signi cant parameters in univariate logistic regression, 27 features were selected by the t-test and LASSO regularization. The AUC values of the prediction models were greater than 0.75 in the test dataset.
The main limitations of this study are its retrospective nature and small sample size, which may have contributed to selection bias. In addition, it is a single-center study; therefore, the generalization of our ndings are limited. Despite these limitations, our study offers some unique and signi cant ndings. Our study showed correlations between radiomics ndings and TB status for the rst time. Moreover, we established the predicting models for ITB status using radiomics feature in 18 F-FDG PET/CT.
In conclusion, higher metabolic quantities were observed in positive ITB group than in negative ITB group.
Radiomics ndings in 18 F-FDG PET/CT was associated with ITB status, and among these features, the most signi cant covariate was compacity for ITB status. Furthermore, prediction models for ITB status using radiomics ndings in 18

Con icts of interest
The authors declare that they have no con ict of interest.