Multiphase CT radiomics nomogram for preoperatively predicting the WHO/ISUP nuclear grade of small (&lt; 4 cm) clear cell renal cell carcinoma

doi:10.21203/rs.3.rs-2612246/v1

Download PDF

Research Article

Multiphase CT radiomics nomogram for preoperatively predicting the WHO/ISUP nuclear grade of small (< 4 cm) clear cell renal cell carcinoma

https://doi.org/10.21203/rs.3.rs-2612246/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 09 Oct, 2023

Read the published version in BMC Cancer →

You are reading this latest preprint version

Background

To develop and validate a CT-based radiomics nomogram for preoperatively predicting WHO/ISUP nuclear grade in small (< 4 cm) clear cell renal cell carcinoma (ccRCC).

Methods

A total of 113 patients with histologically confirmed ccRCC were randomly assigned to the training set (n=67) and the testing set (n=46). The baseline and CT imaging data of the patients were evaluated statistically to develop a clinical model. A radiomics model was created, and the radiomics score (Rad-score) was calculated by extracting radiomics features from the CT images. Then, a clinical radiomics nomogram was developed using multivariate logistic regression analysis by combining the Rad-score and critical clinical characteristics. The receiver operating characteristic (ROC) curve was used to evaluate the discrimination of small ccRCC in both the training and testing sets.

Results

The radiomics model was constructed using six features obtained from the CT images. The shape and relative enhancement value of the nephrographic phase (REV of the NP) were found to be independent risk factors in the clinical model. The area under the curve (AUC) values for the training and testing sets for the clinical radiomics nomogram were 0.940 and 0.902, respectively. Decision curve analysis (DCA) revealed that the radiomics nomogram model was a better predictor, with the highest degree of coincidence.

Conclusion

The CT-based radiomics nomogram has the potential to be a noninvasive and preoperative method for predicting the WHO/ISUP grade of small ccRCC.

Clear cell renal cell carcinoma

Small renal mass

Radiomics nomogram

Computed tomography

WHO/ISUP grade

The increasing use of cross-sectional imaging in recent decades has lead to an increase in the incidence of renal cell carcinoma (RCC) [1]. RCC accounts for about 90% of renal tumors, with clear cell renal cell carcinoma (ccRCC) being responsible for almost 80% [2]. A small renal mass (SRM), defined as a tumor less than 4 cm in diameter, represents around 40% of all kidney tumor[3, 4]. Most SRMs are malignant, and ccRCC is the most common renal malignancy. Despite early detection and resection of RCC, the mortality rate has not decreased, indicating that small RCC is not the main cause of death from RCC [5, 6]. Several studies have shown that active surveillance of patients with small ccRCC who have a limited life expectancy or who do not wish to undergo surgical treatment does not result in a significant increase in mortality[7, 8].

Most small RCCs appear as benign tumors, but some can be highly aggressive and spread to the perirenal fat or distant locations [9, 10]. The World Health Organization/International Society of Urological Pathology (WHO/ISUP) criteria, established in 2016, is the most widely used classification system for ccRCC, with grades I and II considered low-grade and grades III and IV considered high-grade [11]. High-grade ccRCC is more aggressive, more likely to metastasize, and has a poorer outcome. Predicting the tumor grade in advance can assist in determining subsequent treatment strategies. Percutaneous puncture pathology biopsy is a commonly used method for preoperative grading of ccRCC, but it is an invasive procedure. Heterogeneity within the tumor may result in a low histological grade, which can delay treatment [12, 13]. A noninvasive and effective method for determining the histological grade of small ccRCC is necessary.

The most popular noninvasive diagnostic tool for determining whether a small renal mass is benign or malignant, and for identifying the histological grade of small ccRCC, is computed tomography (CT) [14–16]. However, poor interobserver agreement and inconsistent performance make it difficult to use these radiological characteristics in clinical practice. Additionally, there are many overlapping imaging features between high-grade and low-grade tumors[17].

Radiomics is a new research method that extracts features from medical images that are not visible to the human eye [18, 19]. Radiomics has been successfully in various areas related to SRM, including differentiating between benign and malignant SRM and grading small RCC [20–24]. Most research has focused on textural features, and has not considered the potential value of clinical data and imaging characteristics, which could improve the diagnostic accuracy of the models.

The purpose of this study is to evaluate the ability of multiphase CT-based radiomics nomogram analysis, which combines radiomics features and clinic-radiological characteristics, to predict the WHO/IUSP grade of small ccRCC.

Patients

The Institutional Ethics Committee approved this retrospective study and waived the need for patient consent. The study included patients who underwent abdominal CT scans and were diagnosed with a renal tumor at our institution between January 2016 and January 2022. The inclusion criteria were as follows: (1) Patients with ccRCC who underwent a partial or radical nephrectomy. (2) Patients who had plain and enhanced CT scans performed prior to surgery. (3) Patients with complete clinical information. The exclusion criteria were as follows: (1) Significant artefacts on CT images. (2) Tumors with a diameter greater than 4 cm. (3) Patients with a history of both kidney tumors and other tumors. (4) Patients who received treatment before the CT scan. A total of 113 patients were enrolled in the study, including 49 with high-grade small ccRCC and 64 with low-grade small ccRCC. In a ratio of 6:4, patients were randomly assigned to a training set (n = 67) and a testing set (n = 46). Figure 1 illustrates the workflow for enrolling the patient cohort.

Ct Imaging Acquisition

Routine clinical CT scans of the kidney are typically performed using 64-slice multidetector CT equipment. The CT scan parameters are as follows: the tube voltage is 120kV-140 kV; tube current is 250mA-400 mA; slice thickness is 5 mm. Approximately 80 to 100 mL of contrast agents is injected into the antecubital vein at a rate of 3.0 mL/s using a high-pressure injector. Four phases of CT images are obtained: the unenhanced phase (UP), the corticomedullary phase (CMP) which is acquired 30 seconds after contrast injection, the nephrographic phase (NP) which is acquired 90 seconds after contrast injection, and the excretory phase (EP) which is acquired 180 seconds after contrast injection.

Traditional Radiological Characteristics Analysis

Two radiologists, Reader 1 and Reader 2, carefully reviewed the CT images. If there was a disagreement, the two radiologists would discuss together to reach a consensus. Without access to clinicopathologic information, the radiologists jointly evaluated the CT findings, including the maximum diameter of the tumor on axial CT images, shape, location, boundary, calcification, necrosis, renal vein invasion and lymph node metastasis.

To determine the CT value of the tumor, a region of interest (ROI) was selected within the parenchyma of the tumor, excluding necrosis, calcification and vascularity. The ROIs in the study were chosen based on NP images as the tumor was clearly contrasted with the renal parenchyma in these images. Reader 1 selected three non-overlapping ROIs, took individual CT measurements for each, and then averaged the results. As CT scans are performed by different operators and patients, systematic inaccuracies in tumor CT value measurement may occur. To mitigate errors, the CT values of the cortex were measured in the cortical region of the kidney on the side of the tumor. Figure 2 shows an example of this method.

The average tumor attenuation value (TAV) for UP, CMP and NP was obtained from the ROI. The CT value measured in the renal cortex at each phase is known as the cortical attenuation value (CAV). The tumor enhancement value (TEV) and the cortex enhancement value (CEV) were calculated by subtracting the CT value of the UP: TEV_x = TAV_x – TAV₀ and CEV_x= CAV_x – CAV₀, where x indicates the phase (0, UP; 1, CMP; 2, NP). The ratio of TEV to CEV was used to define the relative enhancement value (REV): REV_x = TEV_x/CEV_x, representing the degree of enhancement within the tumor relative to the renal cortex[25].

Construction Of The Clinical Model

The differences between clinic-radiological characteristics of high-grade and low-grade small ccRCC were analyzed using univariate analysis. For categorical variables, the Chi-square test or Fisher exact test was used, while for continuous variables, the t-test or Mann-Whitney U test was applied. Statistically significant clinic-radiological characteristics were then used in a multivariate logistic regression analysis to identify the most valuable clinical factors and build a model. The odds ratio (OR) was calculated for each independent factor as a measure of relative risk prediction with a 95% confidence interval (CI).

Tumor Segmentation And Extraction Of Radiomics Features

Figure 3 illustrates the key steps in a radiomics model for renal tumors. The tumor’s volumes of interests (VOIs) were manually defined in ITK-SNAP software (version 3.8, www.itksnap.org) by two radiologists with extensive abdominal diagnostic experience (Fig. 4).

The extraction of features was performed using the Artificial Intelligence Kit software (A.K. Software, version 3.3.0.R). To minimize variability in the radiomics features, prior to extraction, the following image preprocessing techniques were applied: gray-level discretization, intensity normalization and voxel resampling. Subsequently, 1595 features were extracted from the UP, CMP and NP CT images by the open-source PyRadiomics library, respectively.

The intraclass correlation coefficient (ICC) was calculated to evaluate the consistency and reproducibility of the features. Features with ICC greater than 0.75 in both intra- and inter-observer agreement analyses were included in further analysis. Reader 1 and Reader 2 randomly segmented CT images of 20 patients (8 high-grade small ccRCC and 12 low-grade small ccRCC). Two weeks later, Reader 1 segmented these 20 patients once again.

Construction Of The Radiomics Model

To reduce redundant features and prevent overfitting of the developed radiomics model. The following steps were taken: 1) Features with an ICC greater than 0.75 were selected, 2) Univariate logistic analysis was performed to identify the statistically significant features, 3) The most significant features were chosen through a Gradient Boosting Decision Tree (GBDT) and multivariate logistic analysis, and 4) The remaining features were utilized to compute the radiomics score (Rad-score). The radiomics model was finally established through multivariate logistic regression.

Construction Of Radiomics Nomogram And Evaluation Model Performance

Clinical variables and Rad-score were combined to create a nomogram. Calibration curves were utilized to evaluate the calibration of the nomogram. The Hosmer-Lemeshow test was applied to assess the nomogram’s goodness of fit. The receiver operator characteristic (ROC) curves were utilized to evaluate the discrimination ability of the prediction model for high/low small ccRCC. The clinical validity of the clinical radiomics nomogram was further evaluated through decision curve analysis (DCA).

Statistical analysis

Python software (v.3.6.0) and R software (v.3.5.1) were used to perform the statistical analysis. A statistically significant difference between the two was defined as p < 0.05.

Clinical characteristics and development of clinical model

The differences in clinical and radiological variables for the 113 patients are shown in Table 1. In 113 patients, Shape, TEV2, REV1 and REV2 were significantly different in high-grade ccRCC and low-grade ccRCC after univariate analysis (p < 0.05). After multivariate logistic regression analysis, shape (OR = 0.146, 95% CI = 0.027–0.790, p = 0.025) and REV2 (OR = 26.912, 95% CI = 1.389-521.396, p = 0.029) was an independent risk factor for identifying small ccRCC WHO/ISUP grade (Table 2).

Table 1

Clinic-radiological characteristics in training cohort and testing cohort.
Clinic-radiological characteristics	Training cohort(n = 67)		p	Testing cohort(n = 46)		p
Clinic-radiological characteristics	High-grade	Low-grade		High-grade	Low-grade
Gender			0.008			0.708
Male, n (%)	23(79%)	18(47%)		12(60%)	17
Female, n (%)	6(21%)	20(53%)		8(40%)	9
Age (years)	60(50–65)	55(48–67)	0.676	56(45–66)	55(43–64)	0.682
Maximum diameter (cm)	2.9(2.5–3.7)	3.0(2.7–3.3)	0.844	3.3(2.9–3.6)	3.1(2.3–3.5)	0.179
Shape			0.007			0.012
Round, n (%)	20(69%)	36(95%)		12(60%)	24(92%)
Not round, n (%)	9(31%)	2(5%)		8(40%)	2(8%)
Location			0.570			0.074
Left, n (%)	14(48%)	21(55%)		13(65%)	10(38%)
Right, n (%)	15(52%)	17(45%)		7(35%)	16(62%)
Boundary			0.075			1.000
Clear, n (%)	20(69%)	33(87%)		17(85%)	23(88%)
Blurred, n (%)	9(31%)	5(13%)		3(15%)	3(12%)
Calcification			0.184			0.572
Present, n (%)	2(7%)	0(0%)		2(10%)	1(4%)
Absent, n (%)	27(93%)	38(100%)		18(90%)	25(96%)
Necrosis			0.952			0.085
Present, n (%)	17(59%)	22(58%)		15(75%)	13(50%)
Absent, n (%)	12(41%)	16(42%)		5(25%)	13(50%)
Renal vein invasion			1.000			1.000
Present, n (%)	0(0%)	1(3%)		0(0%)	0(0%)
Absent, n (%)	29(100%)	37(97%)		20(100%)	26(100%)
Lymph node metastasis			0.645			0.435
Present, n (%)	3(10%)	2(5%)		1(5%)	0(0%)
Absent, n (%)	26(90%)	36(95%)		19(95%)	26(100%)
TEV1 (HU)	80(59–101)	95(62–125)	0.215	100(76–124)	104(87–128)	0.444
TEV2 (HU)	71(58–95)	92(64–111)	0.027	84(59–104)	97(78–114)	0.086
REV1	0.76(0.59–0.95)	0.94(0.60–1.22)	0.034	0.91(0.57–1.09)	1.00(0.71–1.21)	0.231
REV2	0.61(0.47–0.70)	0.74(0.61–0.86)	0.003	0.67(0.51–0.84)	0.75(0.62–0.86)	0.465
TEV, tumor enhancement value; REV, relative enhancement value; 1, corticomedullary phase; 2, nephrographic phase

Table 2

Multivariate logistic regression analysis of the clinic-radiological characteristics in predicting the WHO/ISUP grade of small ccRCC.
Clinical characteristics	Multivariate analysis
Clinical characteristics	OR	95% CI	P value
Shape	0.146	0.027–0.790	0.025
TEV2	1.001	0.965–1.037	0.976
REV1	1.473	0.122–17.812	0.761
REV2	26.912	1.389-521.396	0.029
TEV, tumor enhancement value; REV, relative enhancement value; 1, corticomedullary
phase; 2, nephrographic phase

Extraction Of Features And Filtering To Create A Radiomics Model

Of the 4785 radiomics features in the three phases, 2560 had good repeatability (ICC > 0.75), and the dimensionality reduction section was based on these features. A total of 597 features were significantly different by univariate analysis. The six most valuable features were selected from 597 features based on the Gradient Boosting Decision Tree and multivariate logistic analyses (Fig. 5). These features were used to establish a radiomics model. The area under curve (AUC) values in the training and testing sets are respectively 0.924 (95%CI, 0.868–0.969) and 0.869 (95%CI, 0.781–0.947) with the radiomics model. The Rad-score was calculated using six valuable features:

Rad-score = 0.9392–1.6715 × CMP-log_sigma_4_0_mm_3D_firstorder_Maximum

+ 1.7051 × UP-lbp_3D_m2_glrlm_ShortRunLowGrayLeveIEmphasis

– 1.0577 × CMP-wavelet_HHH_ngtdm_Complexity

+ 1.6901 × UP-wavelet-HLH_firstorder_Minimum

+ 1.2094 × NP-wavelet_HHL_firstorder_Median

+ 0.9787 × CMP-wavelet_HHL_glszm_SmallAreaLowGrayLeveIEmphasis

Figure S1 shows how the distribution of the Rad-score in the training and testing cohorts.

The Development Of Nomogram And Evaluation Of Model Performance

The clinic-radiological characteristics and Rad-score from the training cohort were subjected to a multivariate logistic regression analysis to obtain a radiomics nomogram score (Nomo-score): Nomo-score= -3.4699 + 0.9858 × Rad-score + 2.1449 × Shape + 2.4012 × REV2 (Fig. 6). The nomogram’s calibration curves in Figure S2 demonstrate the model’s strong clinical applicability. Table 3 shows the diagnostic effectiveness of the clinical model, the radiomics model and the radiomics nomogram. The ROC curves for the three models are shown in Fig. 7. The DCA curve illustrated in Fig. 8 showed the accuracy of the three models.

Table 3

Diagnostic performance of the clinical model, the radiomics model and the radiomics nomogram.
Model	Training cohort				Testing cohort
Model	AUC (95%CI)	Accuracy %	Specificity %	Sensitivity %	AUC (95%CI)	Accuracy %	Specificity %	Sensitivity %
Clinical model	0.754 (0.650–0.850)	67.2%	89.7%	50.0%	0.706 (0.554–0.845)	73.9%	60.0%	84.6%
Radiomics model	0.924 (0.871–0.968)	85.1%	89.7%	81.6%	0.869 (0.773–0.948)	76.1%	95.0%	61.5%
Radiomics nomogram	0.940 (0.894–0.977)	86.6%	93.1%	81.6%	0.902 (0.811–0.976)	84.8%	72.0%	92.3%

With the increased detection rate of small ccRCC, the frequent underestimation of the histological grade of tumors on puncture biopsy, and the widespread use of active surveillance for patients with small RCC in clinical practice, a reliable method is needed to differentiate the histological grade of small ccRCC. This study has high accuracy in constructing a nomogram for the WHO/ISUP nuclear grading of small ccRCC based on clinic-radiological characteristics and radiomics features, with AUC values of 0.940 (95%CI, 0.894–0.977) and 0.902 (95%CI, 0.811–0.976) in the training and testing sets, respectively.

Previous studies have shown that the diagnostic of SRM is primarily based on imaging characteristics. Takahashi et al.[15] and Sasaguri et al.[14] demonstrated that CT images are highly effective in differentiating between benign and malignant SRM. Cohi et al.[16] found that CT imaging features could predict the histological grading of small ccRCC. In our study, four imaging features, shape, TEV2, REV1 and REV2, were significantly different in identifying high-grade from low-grade small ccRCC after univariate analysis. The results of Ding et al.[26] are consistent with our findings that high-grade ccRCC are more irregular in shape and high-grade ccRCC are less enhanced than low-grade ccRCC. We believe this is mainly because high-grade ccRCC are more malignant and more likely to invade surrounding tissues, making the shape irregular. High-grade ccRCC are prone to internal necrosis, and because of the active growth of the tumor tissue is more likely to block the blood vessels of the tumor, reducing the blood supply to the tumor. However, Halefoglu et al.[27] showed that high-grade RCC tumor enhancement values were higher than low-grade RCC, which is inconsistent with our results. We believe that this discrepancy is mainly due to selection bias in choosing the sample for the study; we studied only one subtype, ccRCC, whereas Halefoglu et al. chose both ccRCC and pRCC. Some studies have shown that necrosis can be used as an essential risk factor to differentiate the histological grade of tumors. But in this research, necrosis was not significantly different between the two groups of tumors, which in our analysis may be because the sample selected was all < 4 cm[27, 28].

As a new form of artificial intelligence, radiomics is widely used in the diagnostic and differential diagnosis of SRMs[20, 21, 23, 24, 29]. It can extract information from medical images that the human eye cannot see. Feng et al.[21] developed a machine learning model for differentiating angiomyolipoma without visible fat (AMLwvf) and RCC based on 58 patients with SRMs, with accuracy, sensitivity, specificity and AUC of 93.9%, 87.8%, 100% and 0.955, respectively. Yang et al.[23] collected 163 patients with SRM, including 118 RCC and 45 AMLwvf, and the final classification model was constructed with an AUC of 0.9. These studies were mainly based on SRM for benign-malignant discrimination and extracted features based only on the largest dimension of the tumor and did not include the full 3D-ROI, thus not containing the complete information of the tumor. Hagi-Momenian et al.[22] constructed various machine learning models based on noncontrast phase, CMP and NP for histological grading and tumor subtyping of small pRCC, respectively. The models constructed based on the features extracted from CMP had the highest AUC values of 0.97-1.0. Haji-Momenian et al.[29] extracted six histogram features and 31 texture features from the noncontrast phase, CMP and NP images of small ccRCC patients. The analysis revealed no significant difference between the features extracted from the noncontrast phase and NP in high-grade and low-grade small ccRCC. Twenty-three features extracted from CMP were significantly different, and multiple machine learning models were built based on these features with a maximum AUC value of 0.97. This is in line with our findings, where we extracted a total of 4785 features from UP, CMP and NP, and filtered them to obtain six valid features, three of which were from CMP. The results suggest that CMP images are valuable in distinguishing the histological grading of small RCC. Previous studies extracted only a few dozen features from the images, which hardly reflect the actual information of the tumor, and thousands of features were extracted for the analysis in this study. Previous studies have focused only on models built through radiomics features, ignoring the information contained in the medical images themselves. The AUC value of the nomogram constructed by combining the radiomics features with the clinical characteristics was 0.940 in the training set and 0.902 in the testing set, which were both higher than those of the clinical model and the radiomics model.

Our study has several limitations. Firstly, the sample size for this study is small, and there is a lack of external validation data. This is due to the fact that we only included ccRCC tumors with a diameter less than 4 cm, and more cases could be collected prospectively for future studies. Secondly, the tumor segmentation in this study was based on manual segmentation, which is both time-consuming and subjective. It would be beneficial to investigate an automated segmentation method for kidney tumors in the future. Lastly, this study only classified ccRCC as high-grade and low grades, which is not a highly accurate classification system. Future studies could develop a model with four categories.

In conclusion, we have developed and validated a CT-based radiomics nomogram that incorporates a rad-score, shape and REV2 to predict the grading of small ccRCC preoperatively. This nomogram will assist clinicians in making informed diagnostic and treatment decisions.

Author’s contributions

Yankun Gao: The acquisition of data, analysis of data, and drafting the article;

Xia Wang: The acquisition of data, analysis of data, and drafting the article;

Xiaoying Zhao: The acquisition of data, analysis of data;

Chao Zhu: The acquisition of data;

Cuiping Li: The acquisition of data;

Jianying Li: The analysis and interpretation of data;

Xingwang Wu: Final approval of the version to be submitted.

All authors read and approved the final manuscript.

Funding

This work was supported by 2021 Medical Empowerment- Pilot Elite Research Project Special Fund (NO. XM_HR_YXFN_2021_05_19).

Availability of data and material

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethic approval and consent to participate

The Ethics Committee of The First Affiliated Hospital of Anhui Medical University approved this retrospective study, and the informed consent was waived. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and /or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Consent for publication

Not applicable.

Competing for publication

The authors declare that they have no conflict of interest.

Acknowledgements

Not applicable.

Hock L, Lynch J,Balaji K. Increasing incidence of all stages of kidney cancer in the last 2 decades in the United States: an analysis of surveillance, epidemiology and end results program data. The Journal of urology 2002; 167:57-60. 10.1097/00005392-200201000-00015
Motzer R J, Jonasch E, Michaelson M D, Nandagopal L, Gore J L, George S, et al. NCCN Guidelines Insights: Kidney Cancer, Version 2.2020. J Natl Compr Canc Netw 2019; 17:1278-1285. 10.6004/jnccn.2019.0054
Cooperberg M R, Mallin K, Ritchey J, Villalta J D, Carroll P R,Kane C J. Decreasing size at diagnosis of stage 1 renal cell carcinoma: analysis from the National Cancer Data Base, 1993 to 2004. J Urol 2008; 179:2131-2135. 10.1016/j.juro.2008.01.097
Nguyen M M, Gill I S,Ellison L M. The evolving presentation of renal carcinoma in the United States: trends from the Surveillance, Epidemiology, and End Results program. J Urol 2006; 176:2397-2400; discussion 2400. 10.1016/j.juro.2006.07.144
Wong M C S, Goggins W B, Yip B H K, Fung F D H, Leung C, Fang Y, et al. Incidence and mortality of kidney cancer: temporal patterns and global trends in 39 countries. Sci Rep 2017; 7:15698. 10.1038/s41598-017-15922-4
Hollingsworth J, Miller D, Daignault S,Hollenbeck B. Rising incidence of small renal masses: a need to reassess treatment effect. Journal of the National Cancer Institute 2006; 98:1331-1334. 10.1093/jnci/djj362
Rendon R A,Jewett M A. Expectant management for the treatment of small renal masses. Urol Oncol 2006; 24:62-67. 10.1016/j.urolonc.2005.07.003
Schieda N, Krishna S, Pedrosa I, Kaffenberger S D, Davenport M S,Silverman S G. Active Surveillance of Renal Masses: The Role of Radiology. Radiology 2022; 302:11-24. 10.1148/radiol.2021204227
Pahernik S, Ziegler S, Roos F, Melchior S W,Thuroff J W. Small renal tumors: correlation of clinical and pathological features with tumor size. J Urol 2007; 178:414-417; discussion 416-417. 10.1016/j.juro.2007.03.129
Thompson R H, Hill J R, Babayev Y, Cronin A, Kaag M, Kundu S, et al. Metastatic renal cell carcinoma risk according to tumor size. J Urol 2009; 182:41-45. 10.1016/j.juro.2009.02.128
Moch H, Cubilla A, Humphrey P, Reuter V,Ulbright T. The 2016 WHO Classification of Tumours of the Urinary System and Male Genital Organs-Part A: Renal, Penile, and Testicular Tumours. European urology 2016; 70:93-105. 10.1016/j.eururo.2016.02.029
Volpe A, Mattar K, Finelli A, Kachura J R, Evans A J, Geddie W R, et al. Contemporary results of percutaneous biopsy of 100 small renal masses: a single center experience. J Urol 2008; 180:2333-2337. 10.1016/j.juro.2008.08.014
Harris C R, Whitson J M,Meng M V. Under-grading of <4 cm renal masses on renal biopsy. BJU Int 2012; 110:794-797. 10.1111/j.1464-410X.2012.10944.x
Sasaguri K, Takahashi N, Gomez-Cardona D, Leng S, Schmit G D, Carter R E, et al. Small (< 4 cm) Renal Mass: Differentiation of Oncocytoma From Renal Cell Carcinoma on Biphasic Contrast-Enhanced CT. AJR Am J Roentgenol 2015; 205:999-1007. 10.2214/AJR.14.13966
Takahashi N, Leng S, Kitajima K, Gomez-Cardona D, Thapa P, Carter R E, et al. Small (< 4 cm) Renal Masses: Differentiation of Angiomyolipoma Without Visible Fat From Renal Cell Carcinoma Using Unenhanced and Contrast-Enhanced CT. AJR Am J Roentgenol 2015; 205:1194-1202. 10.2214/AJR.14.14183
Choi S Y, Sung D J, Yang K S, Kim K A, Yeom S K, Sim K C, et al. Small (<4 cm) clear cell renal cell carcinoma: correlation between CT findings and histologic grade. Abdom Radiol (NY) 2016; 41:1160-1169. 10.1007/s00261-016-0732-9
Orlhac F, Frouin F, Nioche C, Ayache N,Buvat I. Validation of A Method to Compensate Multicenter Effects Affecting CT Radiomics. Radiology 2019; 291:53-59. 10.1148/radiol.2019182023
Lambin P, Leijenaar R, Deist T, Peerlings J, de Jong E, van Timmeren J, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 2017; 14:749-762. 10.1038/nrclinonc.2017.141
Gillies R, Kinahan P,Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology 2016; 278:563-577. 10.1148/radiol.2015151169
Uhlig J, Biggemann L, Nietert M M, Beissbarth T, Lotz J, Kim H S, et al. Discriminating malignant and benign clinical T1 renal masses on computed tomography: A pragmatic radiomics and machine learning approach. Medicine (Baltimore) 2020; 99:e19725. 10.1097/MD.0000000000019725
Feng Z, Rong P, Cao P, Zhou Q, Zhu W, Yan Z, et al. Machine learning-based quantitative texture analysis of CT images of small renal masses: Differentiation of angiomyolipoma without visible fat from renal cell carcinoma. Eur Radiol 2018; 28:1625-1633. 10.1007/s00330-017-5118-z
Haji-Momenian S, Ricker R, Chen Z, Houser M, Adusumilli N, Yang M, et al. Prediction of histologic grade and type of small (< 4 cm) papillary renal cell carcinomas using texture and neural network analysis: a feasibility study. Abdom Radiol (NY) 2021; 46:4266-4277. 10.1007/s00261-021-03044-5
Yang R, Wu J, Sun L, Lai S, Xu Y, Liu X, et al. Radiomics of small renal masses on multiphasic CT: accuracy of machine learning-based classification models for the differentiation of renal cell carcinoma and angiomyolipoma without visible fat. Eur Radiol 2020; 30:1254-1263. 10.1007/s00330-019-06384-5
Scrima A T, Lubner M G, Abel E J, Havighurst T C, Shapiro D D, Huang W, et al. Texture analysis of small renal cell carcinomas at MDCT for predicting relevant histologic and protein biomarkers. Abdom Radiol (NY) 2019; 44:1999-2008. 10.1007/s00261-018-1649-2
Zhu Y H, Wang X, Zhang J, Chen Y H, Kong W,Huang Y R. Low enhancement on multiphase contrast-enhanced CT images: an independent predictor of the presence of high tumor grade of clear cell renal cell carcinoma. AJR Am J Roentgenol 2014; 203:W295-300. 10.2214/AJR.13.12297
Ding J, Xing Z, Jiang Z, Chen J, Pan L, Qiu J, et al. CT-based radiomic model predicts high grade of clear cell renal cell carcinoma. Eur J Radiol 2018; 103:51-56. 10.1016/j.ejrad.2018.04.013
Halefoglu A M,Ozagari A A. Tumor grade estimation of clear cell and papillary renal cell carcinomas using contrast-enhanced MDCT and FSE T2 weighted MR imaging: radiology-pathology correlation. Radiol Med 2021; 126:1139-1148. 10.1007/s11547-021-01350-y
Xv Y, Lv F, Guo H, Liu Z, Luo D, Liu J, et al. A CT-Based Radiomics Nomogram Integrated With Clinic-Radiological Features for Preoperatively Predicting WHO/ISUP Grade of Clear Cell Renal Cell Carcinoma. Front Oncol 2021; 11:712554. 10.3389/fonc.2021.712554
Haji-Momenian S, Lin Z, Patel B, Law N, Michalak A, Nayak A, et al. Texture analysis and machine learning algorithms accurately predict histologic grade in small (< 4 cm) clear cell renal cell carcinomas: a pilot study. Abdom Radiol (NY) 2020; 45:789-798. 10.1007/s00261-019-02336-1

No competing interests reported.

Supplementarymaterial.docx

Download PDF

Journal Publication

published 09 Oct, 2023

Read the published version in BMC Cancer →

Editorial decision: Major revision
09 May, 2023
Reviews received at journal
12 Apr, 2023
Reviewers agreed at journal
04 Apr, 2023
Reviewers invited by journal
04 Apr, 2023
Editor assigned by journal
04 Apr, 2023
Editor invited by journal
02 Mar, 2023
Submission checks completed at journal
02 Mar, 2023
First submitted to journal
21 Feb, 2023

You are reading this latest preprint version

Multiphase CT radiomics nomogram for preoperatively predicting the WHO/ISUP nuclear grade of small (< 4 cm) clear cell renal cell carcinoma

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Materials And Methods

Patients

Ct Imaging Acquisition

Traditional Radiological Characteristics Analysis

Construction Of The Clinical Model

Tumor Segmentation And Extraction Of Radiomics Features

Construction Of The Radiomics Model

Construction Of Radiomics Nomogram And Evaluation Model Performance

Statistical analysis

Results

Clinical characteristics and development of clinical model

Extraction Of Features And Filtering To Create A Radiomics Model

The Development Of Nomogram And Evaluation Of Model Performance

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1