Home About us Editorial board Ahead of print Current issue Search Archives Submit article Instructions Subscribe Contacts Login 

 Table of Contents  
ORIGINAL ARTICLE
Year : 2022  |  Volume : 18  |  Issue : 2  |  Page : 336-344

Computed tomography-based radiomics nomogram model for predicting adherent perinephric fat


1 Department of Radiology, Shandong Provincial Hospital Affiliated to Shandong University, Jinan City, Shandong Province, China
2 Department of Medical Imaging Interventional Therapy, Shandong Provincial Hospital Affiliated to Shandong University, Jinan City, Shandong Province, China
3 Department of Research Collaboration, R and D Center, Beijing Deepwise and League of PHD Technology Co., Ltd, Taian City, Shandong Province, China
4 Department of Radiology, Shandong Provincial Taian City Central Hospital, Taian City, Shandong Province, China

Date of Submission23-Aug-2021
Date of Acceptance20-Oct-2021
Date of Web Publication04-May-2022

Correspondence Address:
Yunchao Chen
Department of Radiology, Shandong Provincial Hospital Affiliated to Shandong University, No. 324, Jingwu Road, Huaiyin 250012, Jinan
China
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/jcrt.jcrt_1425_21

Rights and Permissions
 > Abstract 


Aims: We investigated the predictive value of a computed tomography (CT)-based radiomics nomogram model for adherent perinephric fat (APF).
Materials and Methods: The data of 220 renal carcinoma patients were collected retrospectively. Patients were divided into training (n = 153) and validation cohorts (n = 67). Radiomics features were extracted from plain CT scans, while radscore was generated by a linear combination of selected radiomics features and their weighting coefficients. Univariate logistic regression was used to screen clinical risk factors. Multivariate logistic regression combined with radscore was used to screen final predictors to construct a radiomics nomogram model. Receiver Operating Characteristic curves were used to evaluate the predictive performance of models.
Results: Thirteen radiomics features associated with APF achieved a good predictive effect. The overall area under the curve (AUC) of the radscore model was 0.966, and that of the training and validation cohorts was 0.969 and 0.956, respectively. Gender, age, hypertension, size, perinephric fat thickness, Mayo Adhesive Probability score, neutrophil-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, systemic inflammation response index, and systemic immune-inflammation index were risk factors for APF (P < 0.05). The overall AUC of the radiomics nomogram model based on radiomics features and clinical factors, the training, and validation cohorts was 0.981, 0.997, and 0.949, respectively. Both models had high diagnostic efficiency. However, their differential diagnostic accuracy was higher than that of the clinical model. Additionally, the radiomics nomogram model had higher AUC and specificity.
Conclusions: The radiomics nomogram model is a prediction tool based on radiomics features and clinical risk factors and has high prediction ability and clinical application value for APF.

Keywords: Adherent perinephric fat, computed tomography, nomogram, radiomics


How to cite this article:
Ma T, Cong L, Xu J, Huang C, Ma Q, Hua Q, Li X, Huang Z, Wang X, Chen Y. Computed tomography-based radiomics nomogram model for predicting adherent perinephric fat. J Can Res Ther 2022;18:336-44

How to cite this URL:
Ma T, Cong L, Xu J, Huang C, Ma Q, Hua Q, Li X, Huang Z, Wang X, Chen Y. Computed tomography-based radiomics nomogram model for predicting adherent perinephric fat. J Can Res Ther [serial online] 2022 [cited 2022 Jun 25];18:336-44. Available from: https://www.cancerjournal.net/text.asp?2022/18/2/336/344711

Teng Ma and Lin Cong contributed equally to this work.





 > Introduction Top


With the development of imaging, an increasing number of kidney cancers can be diagnosed early. The average size of the tumor at the first diagnosis shows a gradual decrease.[1] Consequently, the clinical treatment of small renal carcinoma (<4 cm in diameter) also changes significantly.[2] The American Urological Association renal tumor guidelines pointed out that partial nephrectomy (PN) was the preferred choice for renal tumors in stage cT1a.[3] Presently, clinical researches mainly focus on the preoperative evaluation of the feasibility and difficulty of PN. The anatomical characteristics of the tumor itself and the specific characteristics of the patient are closely related to the complexity of PN.[4] The presence or absence of an accessory renal artery, adherent perinephric fat (APF), and other factors greatly impact the choice and difficulty of PN, of which APF is the most important factor.[5],[6],[7] APF is defined as inflammatory tissue adhering to the kidney.[8] APF makes the dissociation of the kidney and exposure of the tumor more difficult, which not only complicates the operation and prolongs the operation time but also may lead to separation bleeding and renal capsule stripping,[4] resulting in increased intraoperative bleeding and affecting the surgical prognosis and quality of life of the patient. Predicting the existence of APF before surgery is of great significance to evaluate the surgery risk, guide patients in selecting the appropriate surgical methods, and improve patient prognosis. APF may also be related to the invasiveness of clear cell carcinoma.[9] APF prediction is based on clinical, imaging, and pathological examinations. The need for preoperative biopsy for APF remains contrasting. Although the area of such iatrogenic trauma is small, it may increase local inflammation and fibrosis due to its direct impact on the renal tumor area, making it more difficult to separate fat during surgery. Therefore, clinicians pay more attention to preoperative clinical and imaging factors. Studies have shown that the Mayo Adhesive Probability (MAP) score can be used as an independent APF predictor.[10],[11] However, in classifying and quantifying cord-like changes in perinephric fat imaging, excessive reliance on the subjective judgment of the observers may lead to inaccurate evaluation results. Radiomics can transform medical images into mineable data, extract quantitative image features from high-throughput data and perform quantitative analysis on this data. Artificial intelligence has been used to establish a clinical decision-making support model.[12] Radiomics is currently used in tumor characterization, clinical grading and staging, genetic analysis, efficacy evaluation, and prognosis prediction.[13],[14] This study intends to establish and verify a radiomics nomogram model for predicting APF based on clinical risk factors and radiomics features.


 > Materials and Methods Top


Participants

The hospital ethics committee approved this retrospective study. This study included patients with renal carcinoma who underwent laparoscopic PN in our hospital from January 2016 to December 2020. Their clinical data were analyzed retrospectively. The inclusion criteria were as follows: (1) Solitary renal tumor with the normal structure of the contralateral kidney; (2) computed tomography (CT) examination before surgery; (3) no history of kidney surgery or abdominal surgery; and (4) no history of malignant tumors. The exclusion criteria were as follows: (1) Complications of acute urinary tract infection, pulmonary infection, autoimmune disease, blood system disease, etc., before surgery; and (2) severe respiratory motion artifacts in CT images. Ultimately, 220 patients were included in the study, including 149 males and 71 females. Their ages ranged from 22 to 85 years, with an average age of 53.95 ± 11.82 years. The patients were divided into an APF group (n = 97) and a non-APF group (n = 123) according to the presence or absence of APF. For model training and verification, all patients were randomly divided into the training cohort and the validation cohort in a ratio of 7:3. The training cohort included 153 patients, including 67 in the APF group and 86 in the non-APF group. The validation cohort included 67 patients, including 30 in the APF group and 37 in the non-APF group. The flow chart is shown in [Figure 1].
Figure 1: Flow chart of the study population

Click here to view


Image acquisition and segmentation

Before surgery, all patients underwent bilateral renal CT plain scans and multiphase multidetector CT (MDCT) scans with Aquilion ONE 640-slice spiral CT (TOSHIBA, Japan). The scanning parameters were as follows: Tube voltage of 120 kV, tube current of 300 mAs, scanning layer thickness and layer spacing of 5 mm, pitch of 0.75–1.0, and scanning range of whole kidneys. After plain scanning, MDCT scanning was performed. As the contrast agent, Iohexol was injected through the elbow vein at a flow rate of 3 ml/s and a dose of 1.5 ml/kg. After contrast agent injection, cortical, medullary, and renal pelvic phase scans were performed at 25 s, 60 s, and 180 s. All patient images were loaded and processed in the original DICOM format and then entered into the postprocessing workstation. Two radiologists with more than 10 years of experience in abdominal CT diagnosis jointly delineated the region of interest (ROI) on the plain scan image. For manual delineation of the perinephric fat area, the widest layer of the renal vein in the transverse section of the affected kidney and two adjacent layers (3 layers in total) were selected. These two radiologists unanimously confirmed the delineated range after discussion and performed feature extraction and quantification of the volume of interest (VOI). The process of delineation is shown in [Figure 2].
Figure 2: Method used to select images for texture analysis in this study. For each patient, three axial images were contoured for texture analysis

Click here to view


Image preprocessing and feature extraction

This study used the Dr. Wise Multimodal Research Platform (https://keyan.deepwise.com, V1.6.2) (Beijing Deepwise and League of PHD Technology Co., Ltd., Beijing, China) for radiomics analysis, including image annotation, feature extraction and selection, and modeling. All images were resampled to the same resolution using B-spline interpolation [1, 1, 1]. Then, the images were normalized by centering the voxel density values at the mean with the SD, also known as the ± 3 σ technique.[15] Finally, the gray value of the image was discretized. Meanwhile, high-pass or low-pass wavelet filters and Laplacian of Gaussian filters with different λ parameters were applied to the original images for preprocessing, and other image transformations were performed. Radiomics features were extracted from the original images and the preprocessed images, including first-order features based on the pixel value of the original images or preprocessed images, shape features describing the shape of the tumor, gray level co-occurrence matrix (GLCM) texture features describing the internal and surface textures of the tumor, gray level run length matrix (GLRLM) texture features, gray level size zone matrix (GLSZM) texture features, and gray level dependence matrix (GLDM) texture features.

Feature selection and modeling

1734 radiomics features were extracted for each ROI, and all features were standardized by Z-score (that is, divided by the SD after subtracting the average value). The radiomics features significantly related to APF were selected for subsequent analysis. The least absolute shrinkage and selection operator (LASSO) feature selection algorithm were used to screen the most informative image features extracted from all VOIs to avoid the “curse of dimensionality,” which would lead to many false positives. Most of the covariate coefficients were reduced to zero during the feature selection process, and the variables retaining a nonzero coefficient after the shrinking process were selected to construct the model. These variables were used to build a model for predicting APF. The machine learning logistic regression algorithm, which is often used for binary classification, was used to construct a predictive model. In this study, a logistic regression model was used to predict APF. All the patients in this study were divided into a training cohort and a validation cohort at a ratio of 7:3. In the training cohort, 5-fold cross-validation was used to select the best model, while the validation cohort was used to verify the effectiveness of the model. The calculation of the radscore is as follows:

Radscore = Sigmoid (β1 × x1+ β2 × x2+ β3 × x3+…. + βn × xn)

Sigmoid (x) =1/(1 + e-x)

Where β1.β2.β3.βn are the coefficients of each feature, while x1. x2. x3. xn are the values of the radiomics features.

Clinical model, radscore model, and radiomics nomogram model

The clinical model included only clinical variables in the multivariate logistic regression modeling, while the radscore model contained only radiomics features. A model including clinical factors and radscore was defined as the radiomics nomogram model.

Statistical analysis

Statistical analyses of clinical characteristics were performed using R software (version 3.3.4; https://www.r-project.org). Categorical variables were compared using the Chi-square test or Fisher's exact test, and continuous variables were compared using the Student's t-test or the Mann–Whitney U-test, as appropriate. One-way analysis of variance (ANOVA) screened the radiomics features that were significantly related to APF. In addition, univariate logistic regression analysis was conducted on all collected clinical variables and radscore. Subsequently, variables with a P < 0.05 in the univariate logistic regression analysis were included in the final multivariate logistic regression model; that is, multivariate logistic regression modeling was applied on the significantly related variables to APF. The receiver operating characteristic (ROC) curves were used to evaluate the predictive performance of the clinical, the radscore, and the radiomics nomogram models. The area under the curve (AUC), specificity, and sensitivity were calculated, while the DeLong test evaluated significant differences among them. A P < 0.05 was considered statistically significant.


 > Results Top


Patient characteristics

Basic information of the patients in the training cohort and validation cohort is shown in [Table 1].
Table 1: Clinicopathological characteristics and computed tomography radscore prediction of patients in the training and validation cohorts

Click here to view


Image feature selection and the radscore model

For each VOI, 1734 features were extracted, including 360 first-order features, 14 shape features, 440 GLCM texture features, 320 GLRLM features, 320 GLSZM features, and 280 GLDM features. Subsequently, ANOVA was applied to select the radiomics features related to APF. Finally, 1142 features were confirmed to be related to APF, and these features were used in the final screening to reduce the feature dimension further, as shown in [Figure 3]. Finally, 13 features with nonzero coefficients, including 2 first-order features, 1 shape feature, and 10 texture features, were selected with LASSO, as shown in [Figure 4].
Figure 3: Radiomics feature selection using the least absolute shrinkage and selection operator regression algorithm. (a) Least absolute shrinkage and selection operator coefficient profiles of the 13 radiomics features. A coefficient profile plot was generated versus the selected-log (λ) value. A vertical line was plotted at the optimal λ value. Each color line represents the change track of each feature coefficient. (b) The parameter (λ) selection in the least absolute shrinkage and selection operator model used 5-fold cross-validation. The y-axis indicates the mean square error. The x-axis indicates the log (λ). The black curve indicates the average error for each model with a given λ. Each color line represents the error for each model with a given λ. The vertical lines define the optimal λ value of 0.04 with log (λ) =1.41

Click here to view
Figure 4: Summary graph of the influence of features on the decision making of the radscore model and interactions between the features in the model. The figure shows not only the 13 most important features but also their impact on the model output. Each point represents a case in the dataset. The color of the point represents the value of the element. Blue represents the lowest range, while red represents the highest range. The horizontal axis shows the corresponding Shapley value of the feature. A positive Shapley value represents a decision conducive to adhesion and vice versa

Click here to view


Modeling was based on the 13 included features. In this study, a logistic regression model was used for modeling to determine the presence of APF. Among all randomly selected data, 70% were used for model training, and 30% were used to test model performance. The equation below was used to calculate the radscore of each lesion.

Radscore =

Sigmoid (logarithm_glszm_SizeZoneNonUniformity *0.97309787+

log-sigma-1-0-mm−3D_glszm_GrayLevelNonUniformity *0.43570987+

wavelet-LLH_glszm_GrayLevelNonUniformity *1.15429267+

original_glcm_Imc1* −0.60812169+

wavelet-LLL_glcm_Imc2 * 0.23720362+

original_shape_Elongation * −0.00970969+

original_glcm_Imc2 * 1.16335793+

wavelet-LHH_glszm_SizeZoneNonUniformity *1.09874191+

logarithm_glcm_JointAverage *0.66022883+

log-sigma-5-0-mm − 3D_firstorder_Mean * −0.67589553+

log-sigma-4-0-mm − 3D_firstorder_10Percentile * −0.53141462+

wavelet-LLH_glszm_SmallAreaEmphasis * −0.41975166+

gradient_glcm_Correlation * −0.86582574 + 0.03933869)

The results showed that the overall AUC of the radscore model was 0.966, and the specificity and sensitivity were 0.951 and 0.928, respectively. In the training and validation cohorts, the AUCs were 0.969 and 0.956, the specificities were 0.953 and 0.946, and the sensitivities were 0.940 and 0.900, respectively.

Clinical model and radiomics nomogram model

In the training cohort, the univariate logistic regression results showed that gender, age, hypertension, size, perinephric fat thickness, MAP score, neutrophil-to-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR), systemic inflammation response index (SIRI), and systemic immune-inflammation index (SII) were all clinical risk factors for APF (P < 0.05), among which the odds ratio (OR) values for gender, MAP score, and SII were 3.017, 3.455, and 3.588, respectively, as shown in [Table 2]. The clinical model based on clinical factors and the radiomics nomogram model based on clinical factors and the radscore were established, respectively, as shown in [Figure 5], using the multivariate logistic regression analysis. The radiomics nomogram model and radscore model showed a good fit, as shown in [Figure 6]. The overall AUC of the clinical model was 0.874 and that of the radiomics nomogram model was 0.981. In the training cohort, the AUC was 0.899 for the clinical model and 0.997 for the radiomics nomogram model. In the validation cohort, the AUC was 0.823 for the clinical model and 0.949 for the radiomics nomogram model, as shown in [Table 3] and [Figure 7]. The DeLong test was conducted on the ROC curve of each model, as shown in [Table 4]. The results showed that both the radscore and radiomics nomogram models had higher diagnostic efficacy and better identification ability than the clinical model in the training and validation cohorts. However, there was no significant difference between the diagnostic efficacy of the radscore model and that of the radiomics nomogram model. Although the AUC and specificity of the radiomics nomogram model were higher than those of the radscore model, the sensitivity of the two models was the same.
Figure 5: Nomogram based on the radiomics feature and clinical factors

Click here to view
Figure 6: Calibration curve of the radiomics nomogram model (a) and radscore model (b) in the validation cohort. The calibration curve and Hosmer–Lemeshow test showed a good fit of the radiomics nomogram and radscore model

Click here to view
Figure 7: Decision curve analysis of the radiomics nomogram and radscore model (a). Receiver operating characteristic curves of the clinical, radscore, and nomogram models for the training cohort (b) and validation cohort (c)

Click here to view
Table 2: Univariate analysis of the clinical features of the training cohort

Click here to view
Table 3: Performance of the clinical model, radscore model and nomogram model

Click here to view
Table 4: DeLong test of the clinical model, radscore model and nomogram model

Click here to view



 > Discussion Top


As a reflection of the complexity of PN, APF has been extensively studied. Studies have shown that the presence of APF significantly prolongs the operation time, increases the amount of blood loss during the operation, and increases the risk of morbidity during the perioperative period.[4],[10],[16] The presence of APF brings greater difficulty during surgery, especially during minimally invasive surgery. During the operation, surgeons must perform a more detailed and cautious dissection of the adherent perinephric tissue, which takes more time. The removal of the tumor may require intraoperative ultrasound or other methods.[17] Surgeons should obtain objective data on APF and its severity before surgery to evaluate the surgical complexity or select alternative methods to treat small renal carcinoma, such as radical nephrectomy or percutaneous ablation.[18] There are currently various scoring systems used to predict APF. In recent studies, the MAP score is the strongest predictor of APF during PN.[10],[19] Borregales et al.[17] developed a risk scoring system with scores ranging from 0 to 4 to predict APF that included three variables: Diabetes, perinephric fat thickness, and perinephric stranding. Zheng et al.[20] suggested that the density of the perinephric tissue in the ROI at the level of the renal hilum could predict the difficulty of perinephric fat dissection caused by APF during PN. The results of Li et al.[21] showed that the multimode and multiparameter models of dual-energy CT could effectively be used to predict the presence of APF. The above-mentioned scoring systems are all based on imaging. However, their accuracy and reproducibility are yet to be determined. This is because the operator has a certain degree of bias and subjectivity when measuring perinephric fat thickness and perinephric fat density on the CT image, and the human eye can only partially recognize the image information. Moreover, doctors with different levels of experience have certain differences in evaluating and discriminating the imaging features of APF. In summary, the development of a tool that can accurately predict APF is of great significance.

With traditional imaging, morphological diagnosis is mainly performed visually using images. However, in the era of precision medicine, this traditional model can no longer meet the requirements of accurate diagnosis and disease treatment. Under these circumstances, radiomics came into existence. Medical images can be converted into mineable data by the quantitative extraction of high-throughput features, which can prevent the diagnostic errors caused by human factors to a certain extent and reflect the heterogeneity of the disease more comprehensively and objectively. In 2018, Khene et al.[18] used texture analysis technology for the first time to study APF. They compared 15 signs of perinephric texture in MSCT ROIs in APF and non-APF patients, with statistically significant differences in texture skewness, correlation, and entropy. In this regard, we hold the same view and believe that these characteristics can predict APF. The results of this study showed that gender, age, history of hypertension, tumor size, perinephric fat thickness, MAP score, and some inflammatory markers in the preoperative peripheral blood were all clinically independent risk factors for APF. The OR values for gender, MAP score, and SII were 3.017, 3.455, and 3.588, respectively. However, the mechanism of APF is still unclear and may be related to factors such as fibrosis, autoimmunity, and inflammation.[7],[16],[22] The relationship of the NLR, MLR, SIRI, and SII with APF, especially the relationship between the SII and APF in this study, further confirmed the relationship between the occurrence of APF and systemic inflammatory response. To the best of our knowledge, this is the first study to report the relationship between inflammatory markers and APF. In this study, 13 imaging features were selected by LASSO dimension reduction, including 2 first-order features, 1 shape feature, and 10 texture features. The GLSZM features all have a higher weight, among which gray level nonuniformity, size zone nonuniformity, and small area emphasis mainly reflect the inhomogeneity and complexity of perinephric fat density, revealing the spatial characteristics of perinephric inflammatory tissue.

In this study, the radscore, clinical, and radiomics nomogram models were constructed. The results showed that both the radscore and radiomics nomogram models had high diagnostic efficacy. The overall AUC of the radscore model was 0.966, and the AUCs of the training and validation cohorts were 0.969 and 0.956, respectively. Meanwhile, the overall AUC of the radiomics nomogram model based on radiomics features and clinical factors was 0.981, and the AUCs of the training and validation cohorts were 0.997 and 0.949, respectively. The diagnostic efficacy of these two models was higher than that of the clinical model. This result indicates that radiomics has an advantage in predicting APF over clinical information, including MAP score, perinephric fat thickness, and other strongly correlated factors, and can further improve prediction accuracy. Although there was no significant difference between the diagnostic efficacy of the radscore model and that of the radiomics nomogram model, the overall AUC and specificity of the radiomics nomogram model were both higher than those of the radscore model. This may be the result of complementary effects of imaging features and clinical factors. Furthermore, in this study, the overall AUC of the radscore model reached 0.966 based only on radiomics features of plain CT scan images. These data fully demonstrate the feasibility and strong predictive ability of radiomics, which is expected to be a new method for APF prediction in the future.

Undeniably, there are some limitations in this study. First, instead of all perinephric fat, a two-dimensional texture analysis of the ROI was performed on only three consecutive sections with the widest renal vein as the center, which may cause a loss of information to some extent. Second, this study used only plain scan images to complete texture analysis and feature extraction. Third, the presence or absence of APF in patients was determined by urological surgeons according to the actual conditions observed during the operation, which may have certain subjectivity and thus affect the grouping of cases. Finally, this study was a retrospective study. The existence of some biases will inevitably affect the analysis. Thus, it is necessary to conduct prospective studies to control for confounding variables.


 > Conclusions Top


This study established a radiomics nomogram model based on radiomics features of plain CT images and clinical risk factors. The results showed that the radiomics nomogram model showed a high predictive ability and clinical application value in APF prediction. In future studies, the sample size will be expanded, and multicenter and prospective studies will be conducted to improve the diagnostic efficacy of the model further and obtain high-level evidence for clinical application.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.



 
 > References Top

1.
Basatac C, Akpinar H. Robot-assisted partial nephrectomy with segmental renal artery clamping: A single center experience. Urol J 2019;16:469-74.  Back to cited text no. 1
    
2.
Bukavina L, Mishra K, Calaway A, Ponsky L. Robotic partial nephrectomy: Update on techniques. Urol Clin North Am 2021;48:81-90.  Back to cited text no. 2
    
3.
Campbell S, Uzzo RG, Allaf ME, Bass EB, Cadeddu JA, Chang A, et al. Renal mass and localized renal cancer: AUA guideline. J Urol 2017;198:520-9.  Back to cited text no. 3
    
4.
Kocher NJ, Kunchala S, Reynolds C, Lehman E, Nie S, Raman JD. Adherent perinephric fat at minimally invasive partial nephrectomy is associated with adverse peri-operative outcomes and malignant renal histology. BJU Int 2016;117:636-41.  Back to cited text no. 4
    
5.
Kawamura N, Saito K, Inoue M, Ito M, Kijima T, Yoshida S, et al. Adherent perinephric fat in Asian patients: Predictors and impact on perioperative outcomes of partial nephrectomy. Urol Int 2018;101:437-42.  Back to cited text no. 5
    
6.
Martin L, Rouviere O, Bezza R, Bailleux J, Abbas F, Schott-Pethelaz AM, et al. Mayo adhesive probability score is an independent computed tomography scan predictor of adherent perinephric fat in open partial nephrectomy. Urology 2017;103:124-8.  Back to cited text no. 6
    
7.
Yao Y, Gong H, Pang Y, Gu L, Niu S, Xu Y, et al. Risk factors influencing the thickness and stranding of perinephric fat of mayo adhesive probability score in minimally invasive nephrectomy. Med Sci Monit 2019;25:3825-31.  Back to cited text no. 7
    
8.
Bylund JR, Qiong H, Crispen PL, Venkatesh R, Strup SE. Association of clinical and radiographic features with perinephric “sticky” fat. J Endourol 2013;27:370-3.  Back to cited text no. 8
    
9.
Thiel DD, Davidiuk AJ, Meschia C, Serie D, Custer K, Petrou SP, et al. Mayo adhesive probability score is associated with localized renal cell carcinoma progression-free survival. Urology 2016;89:54-60.  Back to cited text no. 9
    
10.
Davidiuk AJ, Parker AS, Thomas CS, Leibovich BC, Castle EP, Heckman MG, et al. Mayo adhesive probability score: An accurate image-based scoring system to predict adherent perinephric fat in partial nephrectomy. Eur Urol 2014;66:1165-71.  Back to cited text no. 10
    
11.
Yang B, Ma LL, Qiu M, Xia HZ, He W, Meng TY, et al. A novel nephrometry scoring system for predicting peri-operative outcomes of retroperitoneal laparoscopic partial nephrectomy. Chin Med J (Engl) 2020;133:577-82.  Back to cited text no. 11
    
12.
Bi WL, Hosny A, Schabath MB, Giger ML, Birkbak NJ, Mehrtash A, et al. Artificial intelligence in cancer imaging: Clinical challenges and applications. CA Cancer J Clin 2019;69:127-57.  Back to cited text no. 12
    
13.
Papanikolaou N, Matos C, Koh DM. How to develop a meaningful radiomic signature for clinical use in oncologic patients. Cancer Imaging 2020;20:33.  Back to cited text no. 13
    
14.
Verma V, Simone CB 2nd, Krishnan S, Lin SH, Yang J, Hahn SM. The rise of radiomics and implications for oncologic management. J Natl Cancer Inst 2017;109:djx055.  Back to cited text no. 14
    
15.
Collewet G, Strzelecki M, Mariette F. Influence of MRI acquisition protocols and image intensity normalization methods on texture classification. Magn Reson Imaging 2004;22:81-91.  Back to cited text no. 15
    
16.
Khene ZE, Peyronnet B, Mathieu R, Fardoun T, Verhoest G, Bensalah K. Analysis of the impact of adherent perirenal fat on peri-operative outcomes of robotic partial nephrectomy. World J Urol 2015;33:1801-6.  Back to cited text no. 16
    
17.
Borregales LD, Adibi M, Thomas AZ, Reis RB, Chery LJ, Devine CE, et al. Predicting adherent perinephric fat using preoperative clinical and radiological factors in patients undergoing partial nephrectomy. Eur Urol Focus 2021;7:397-403.  Back to cited text no. 17
    
18.
Khene ZE, Bensalah K, Largent A, Shariat S, Verhoest G, Peyronnet B, et al. Role of quantitative computed tomography texture analysis in the prediction of adherent perinephric fat. World J Urol 2018;36:1635-42.  Back to cited text no. 18
    
19.
Yao Y, Xu Y, Gu L, Liu K, Li P, Xuan Y, et al. The mayo adhesive probability score predicts longer dissection time during laparoscopic partial nephrectomy. J Endourol 2020;34:594-9.  Back to cited text no. 19
    
20.
Zheng Y, Espiritu P, Hakky T, Jutras K, Spiess PE. Predicting ease of perinephric fat dissection at time of open partial nephrectomy using preoperative fat density characteristics. BJU Int 2014;114:872-80.  Back to cited text no. 20
    
21.
Li G, Dong J, Huang W, Zhang Z, Wang D, Zou M, et al. Establishment of a novel system for the preoperative prediction of adherent perinephric fat (APF) occurrence based on a multi-mode and multi-parameter analysis of dual-energy CT. Transl Androl Urol 2019;8:421-31.  Back to cited text no. 21
    
22.
Gorin MA, Mullins JK, Pierorazio PM, Jayram G, Allaf ME. Increased intra-abdominal fat predicts perioperative complications following minimally invasive partial nephrectomy. Urology 2013;81:1225-30.  Back to cited text no. 22
    


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
 
 
    Tables

  [Table 1], [Table 2], [Table 3], [Table 4]



 

Top
 
 
  Search
 
Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  >Abstract>Introduction>Materials and Me...>Results>Discussion>Conclusions>Article Figures>Article Tables
  In this article
>References

 Article Access Statistics
    Viewed933    
    Printed40    
    Emailed0    
    PDF Downloaded81    
    Comments [Add]    

Recommend this journal


[TAG2]
[TAG3]
[TAG4]