Quantitative CT-based biomarkers for predicting Renal cell carcinoma subtypes: a comparison of Dual-Energy CT, Perfusion CT, and CT texture parameters

Quantitative CT-based biomarkers for predicting Renal cell carcinoma subtypes: a comparison of Dual-Energy CT, Perfusion CT, and CT texture parameters | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Quantitative CT-based biomarkers for predicting Renal cell carcinoma subtypes: a comparison of Dual-Energy CT, Perfusion CT, and CT texture parameters ANJALI SAH, SNEHA GOSWAMI, AMIT GUPTA, SANIL GARG, NEEL YADAV, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5952087/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose To evaluate and compare the diagnostic performance of CT texture analysis (CTTA), perfusion CT (PCT), and dual-energy CT (DECT) in distinguishing between clear-cell renal cell carcinoma (ccRCC) and non-ccRCC. Methods This retrospective study included 66 patients with RCC (52 ccRCC and 14 non-ccRCC) who underwent DECT and PCT imaging before surgery between January 2017 and December 2022. Two independent radiologists measured DECT parameters (iodine concentration and iodine ratio) and PCT parameters (blood flow, blood volume, mean transit time, and time to peak) using circular ROIs placed on tumors. For CTTA, the largest tumor cross-section in the corticomedullary phase was manually annotated using the "labelme" tool, and texture features were extracted with Python libraries including "scipy" and "numpy." Multivariate logistic regression analysis was performed to assess the ability of PCT, DECT, and CTTA models to predict tumor subtypes. Results All three imaging modalities demonstrated high diagnostic accuracy, with F1 scores of 0.9107, 0.9358, and 0.9348 for PCT, DECT, and CTTA, respectively. Inter-reader agreement for PCT and DECT parameters was excellent (Pearson correlation > 0.85). None of the three models were significantly different (p > 0.05). While each modality could effectively differentiate between ccRCC and non-ccRCC, higher iodine ratio (IR) on DECT and increased entropy on CTTA were independent predictors of ccRCC, with F1 scores of 0.9345 and 0.9272, respectively (p < 0.001). The combined ML model integrating DECT, PCT, and CTTA parameters yielded the highest diagnostic accuracy, with an F1 score of 0.954. Conclusions The diagnostic accuracy of PCT, DECT, and CTTA in distinguishing between ccRCC and non-ccRCC tumors was equivalent and high. However, among these three methods, only IR on DECT and entropy on CTTA were identified as independent predictors of the RCC subtype; hence, these two quantitative markers may be more applicable in clinical practice. Clinical relevance: Accurate, non-invasive biomarkers are essential to differentiate RCC subtypes, aiding in prognosis and guiding targeted therapies, particularly in ccRCC, where treatment options differ significantly. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Renal cell carcinoma (RCC) accounts for approximately 85 to 90% of all renal tumors and is by far the most prevalent primary renal malignancy. Clear-cell renal cell carcinoma (ccRCC) is the most common and aggressive subtype of RCC, accounting for approximately 70% of cases ( 1 ). Known for its high morbidity and mortality, ccRCC requires timely diagnosis and targeted treatment to improve patient outcomes ( 2 ). The distinction between ccRCC and non-ccRCC is clinically significant, as it directly influences therapeutic decisions, particularly with the advent of targeted treatments such as tyrosine kinase inhibitors, vascular endothelial growth factor (VEGF) inhibitors, and immune checkpoint inhibitors ( 3 , 4 ). These treatments have shown marked efficacy in ccRCC due to its molecular underpinnings, while non-ccRCC treatment strategies often rely on extrapolated data from ccRCC trials, underscoring the need for better diagnostic and therapeutic specificity. Currently, tumor biopsy remains the gold standard for determining RCC subtype and guiding systemic therapy. However, this invasive procedure carries risks such as bleeding, infection, and sampling error, particularly for heterogeneous tumors or those in anatomically challenging locations ( 5 ). Furthermore, traditional imaging methods like multiphase CT and MRI often fail to reliably differentiate RCC subtypes due to overlapping radiological features. This limitation underscores the need for non-invasive, accurate, and reproducible imaging biomarkers that can distinguish RCC subtypes with greater precision. Quantitative imaging techniques such as Dual-Energy CT (DECT), Perfusion CT (PCT), and CT Texture Analysis (CTTA) have emerged as promising tools to address this diagnostic challenge. These modalities provide detailed insights into tumor vascularity, perfusion, and tissue heterogeneity, reflecting the underlying tumor biology. PCT, for instance, assesses perfusion parameters such as blood flow, blood volume, and mean transit time, which are particularly relevant given the highly vascular nature of ccRCC compared to less vascular non-ccRCC subtypes like papillary and chromophobe RCC ( 6 ). DECT, with its ability to analyze material-specific attenuation properties at different energy levels, offers unique quantitative metrics for tumor characterization and differentiation of RCC subtypes ( 7 ). CTTA is a powerful tool that analyzes the distribution and patterns of pixel intensities within a tumor on CT images, providing insights into tumor heterogeneity that are beyond what one’s eye can see ( 8 , 9 ). It evaluates features like entropy (irregularity), variance, skewness (asymmetry), and kurtosis (peakedness), which reflect the tumor's internal complexity and structure. It has shown great potential in characterizing and prognosticating various malignancies, including RCC, by quantifying tumor complexity and internal architecture ( 10 ). Unlike PCT, it can be applied retrospectively to existing imaging data, making it a practical option for personalized diagnosis and treatment planning. Incorporating these advanced imaging biomarkers could transform RCC diagnosis by offering a non-invasive alternative to biopsies and improving treatment stratification. The ability to differentiate ccRCC from non-ccRCC with high accuracy is crucial, given the variability in response to targeted therapies and the growing emphasis on personalized oncology. This study seeks to bridge this gap by assessing the diagnostic performance of quantitative parameters derived from PCT, DECT, and CTTA within the same patient cohort. By integrating these advanced imaging modalities, the study aims to establish a robust, non-invasive framework for RCC subtype differentiation, ultimately guiding personalized treatment strategies and improving clinical outcomes. MATERIALS AND METHODS This retrospective study was approved by the institutional review board and the need for informed consent was waived. The CT dataset was compiled by reviewing hospital records to identify patients diagnosed with RCC between January 2017 and December 2022 who had undergone both DECT and PCT prior to radical or partial nephrectomy. The details of patients inclusion and exclusion is shown in a flow chart (Fig. 1). We initially identified 85 patients with pathologically confirmed renal cell carcinoma (RCC) who underwent preoperative PCT and DECT between January 2017 and December 2022 (Fig. 1). Of these, 19 patients were excluded based on the following criteria: inadequate image quality due to poor timing of post-contrast phases or motion artifacts (n = 10), history of prior systemic therapy or surgery (n = 9). A total of 66 patients met the inclusion criteria, comprising 52 cases of ccRCC) and 14 cases of non-ccRCC. Perfusion CT protocol Perfusion CT (PCT) scans were performed in DECT mode using a 64-slice dual-source, dual-energy 2 x 128-section multi-detector CT scanner (Somatom Definition Flash, Siemens Healthineers, Germany). The current protocol in our department includes a non-contrast (NCCT) single scan-and-view acquisition performed at low dose with shallow breathing to ensure accurate positioning. Following this, 70–90 mL of non-ionic contrast medium (350 mg iodine/mL,Omnipaque 350, GE Healthcare) was injected at a rate of 5 mL/s, followed by a 40 mL saline chaser bolus. The dynamic imaging sequence started 5 seconds after the contrast injection, consisting of 34 acquisitions with a rotation time of 0.5 seconds and a cycle time of 1.5 seconds, lasting a total of 45 seconds. These images were acquired mostly in cortico-medullary (CM) phase. This was followed by the nephrographic (NP) phase at 60–70 seconds and excretory (DEL) phase images at 240 seconds. Each of the 34 acquisitions included three consecutive 5 mm slices reconstructed across an 11 cm craniocaudal region. Perfusion parameters such as tumor blood flow (TBF), tumor blood volume (TBV), mean transit time (MTT), and vascular permeability-surface area product were calculated. Blood flow (BF), blood volume (BV), and flow extraction product (K) were derived using the maximum-slope and delay-corrected modified Patlak approach and correlated with microvessel density (MVD). Data analysis was conducted using dedicated software (Siemens Healthineers, volume PCT and Syngo MMWP), with the arterial input defined by drawing a circular ROI on the abdominal aorta, after applying the inbuilt motion correction algorithm. Perfusion maps were generated using a deconvolution pharmacokinetic model. For each tumor, the largest cross-section was selected for ROI placement by two independent readers having 5 and 7 years of experience, who were blinded to histopathological results. In heterogeneous tumors, ROIs were drawn to include the most enhancing areas while excluding necrotic regions. Dual-Energy CT Protocol For DECT analysis, nephrographic (NP) images were selected from the above-mentioned protocol. The retrieved CT scans were anonymized during archiving and a new unique ID was given to the scans. Color-coded iodine maps were created using the iodine subtraction method (Liver VNC, Siemens Healthineers) and dual-energy software (Syngovia VB10A, Siemens Healthineers) for DECT data analysis. Similar regions of interest (ROIs) were drawn by the readers (mean size: 0.58 ± 0.67 cm², range: 0.1–5.7 cm²) within the tumor region exhibiting maximum iodine uptake on DECT. Iodine concentration (IC) in milligrams per milliliter and the iodine ratio (IC in tumor / IC in the aorta at the level of the renal artery supplying the kidney with the tumor) were measured to assess the tumor’s iodine enhancement pattern and vascularity. Texture analysis For each patient, a single axial CT slice representing the largest tumor cross-section was extracted from each available CT phase and saved as a JPEG file. A single radiologist (9 years of experience in body imaging) manually annotated the tumor on these CT images using the 'Polygonal ROI' function in the freely available ‘LabelMe’ tool. This annotation process was applied to all available CT phases for each patient, and the annotated data was saved as a JSON file for each tumor region. Image pre-processing was performed using the open-source OpenCV library. For the CTTA, texture parameters were computed using open-source Python libraries including “scipy,” “numpy,” and “opencv.” First, second, and third-order texture features were extracted from each tumor annotation, encompassing parameters such as mean, variance, energy, entropy, and smoothness. These texture features were used to capture tumor heterogeneity, providing valuable insights into the tumor microenvironment. The details of DECT, PCT, and CTTA analysis is shown in Fig. 2. The ML models for DECT, PCT, and CTTA: We employ Support Vector Machines (SVM), which are supervised learning methods, for machine learning ( 11 ). SVM is a technique that determines the most efficient way to draw a line—or, in more complicated situations, a plane or a hyperplane—between groupings of data points to distinguish them (in our example, cc-RCC vs non cc- RCC). This line is created such that it is as far away from the nearest points of each group as feasible, ensuring the most clear differentiation between the various types of tumors. This architecture is more resistant to outliers and has a faster convergence rate to its final model. CTTA, DECT, and PCT parameters obtained from each CT phase independently were used to train SVM models for RCC subtyping (ccRCC vs. non-ccRCC). The SVM model was developed in two steps namely normalization and feature scaling followed by model training. Using a technique known as min-max normalization, each parameter was normalized from 0 to 1. This phase is critical because it places all parameters on an equal scale, allowing them to contribute equally to the model regardless of their original scale or units. For model training, we did not use any prior models when creating our SVM model. Another parameter taken into account was the prediction label for each sample in the testing dataset. For training, validation, and test datasets, we employed a 70:20:10 split, accordingly. A standard machine learning library called sklearn, which offers resources for creating and implementing machine learning models, was used for the training process. Statistical analysis All statistical analyses were performed by using STATA 18 statistical software. Continuous variables were expressed as means ± standard deviations (SD) or medians with interquartile ranges (IQR), while categorical variables were presented as frequencies or percentages. Interreader agreement for the DECT and PCT parameters was calculated using the Pearson correlation (r) and Spearman correlation coefficient (ρ). Using the "matplotlib" and "seaborn'' libraries of Python, we ran multivariate logistic regression analyses on the PCT, DECT, and CTTA models to determine how well each feature predicted tumor type. The first-order features, such as mean, variance, standard deviation, etc., were computed on this image using the standard numpy library. The second-order features, such as energy, entropy, and smoothness, were calculated using the Scipy library. The corresponding p-values and F1 scores were recorded for CTTA, DECT, and PCT parameters. Box plots were made for each of the parameters to differentiate between ccRCC and non-ccRCC. The trained SVM models were also subjected to the same statistical analysis as the individual quantitative parameters from the CTTA, DECT, and PCT analyses. This study repetition is critical for validating and comparing the performance of our SVM models to the quantitative parameter analysis. Results This study included 66 patients with histopathologically confirmed renal cell carcinoma, comprising 52 cases of ccRCC and 14 cases of non-ccRCC. The study population included 40 males and 26 females, with a median age at presentation of 68 (range: 43–79 years). Demographic details of the patients are summarised in Table 1. The interreader agreement for DECT and PCT parameters is summarized in Table 2. Excellent agreement was observed for DECT parameters, with Pearson (r) and Spearman (ρ) correlation coefficients of 0.894 and 0.894 for iodine concentration (IC) and 0.889 and 0.862 for iodine ratio (IR), respectively (p < 0.0005). Among PCT parameters, blood volume (BV) and blood flow (BF) also demonstrated strong agreement (r = 0.935, ρ = 0.946 for BV; r = 0.837, ρ = 0.922 for BF, p < 0.05). However, MTT and TTP showed weaker agreement, with r = 0.35, ρ = 0.422 for MTT (p = 0.629) and r = 0.311, ρ = 0.373 for TTP (p = 0.937). The ML models for DECT, PCT, and CTTA showed strong diagnostic performance in differentiating ccRCC from non-ccRCC, as shown in Table 3. DECT achieved the highest F1 score among individual modalities (F1 = 0.935), followed closely by CTTA (F1 = 0.934) and PCT (F1 = 0.910). The combined ML model incorporating parameters from all three modalities yielded the highest diagnostic accuracy, with an F1 score of 0.954. Among both the DECT features, iodine ratio (IR) was the best-performing parameter, achieving an F1 score of 0.902 (Table 3).Iodine ratio (IR) was markedly elevated in ccRCC, with a mean value of 65.12 + 23.73 compared to 35.17 + 17.99 in non-ccRCC (p < 0.001). Similarly, iodine concentration (IC) was also notably higher in ccRCC, with a mean value of 2.3 + 0.79 compared to 1.42 + 0.75 in non-ccRCC (p < 0.05). Figure 3 shows a box plot comparing IR values across the two tumor subtypes. In texture analysis, entropy emerged as the most effective parameters for distinguishing ccRCC from non-ccRCC, achieving an AUC of 0.94 . Entropy was significantly higher in ccRCC ( 7.94 ± 0.336 ) compared to non-ccRCC ( 6.43 ± 0.297 , p < 0.001 ), with a threshold value of 7.21 yielding 84.8% accuracy , 83.0% sensitivity , and 92.3% specificity . Figure 4 shows the ROC curve. Variance also demonstrated strong performance, with ccRCC showing higher values ( 1440.0 ± 663.0 ) than non-ccRCC (505.0 +235.0, p < 0.001 ), achieving an AUC of 0.874 , 84.8% accuracy, 86.8% sensitivity, and 76.9% specificity at a threshold of 930. Similarly, mean values were also higher in clear cell than non clear cell carcinoma (p=0.01). Energy and Smoothness was higher in clear cell than non clear cell subtypes but the difference was not significant. PCT parameters demonstrated lower diagnostic performance compared to DECT and CTTA. The maximum peak intensity (MIP) showed significantly higher values in ccRCC (158.0 ± 38.8) compared to non-ccRCC (96.8 ± 33.1), with a p-value < 0.001, demonstrating high diagnostic accuracy (AUC: 0.884), with sensitivity of 84.9%, and specificity of 84.6%. Blood flow (BF) also differed significantly between the two groups, with ccRCC exhibiting a mean value of 883 + 780 compared to 532.0 + 430.0 in non-ccRCC (p = 0.04). Lastly, blood volume (BV) was significantly higher in ccRCC (118.0 + 85.6) than in non-ccRCC (70.5 + 60.6), with a p-value of 0.02. Although the mean TTP and MTT were higher in clear cell compared to non clear cell RCC, but the difference was not significant (p>0.05) Discussion The differentiation of RCCs into ccRCC and non-ccRCC subtypes is crucial for personalized treatment strategies due to significant differences in their biology, prognosis, and therapeutic response. Our study aimed to evaluate and compare the diagnostic performance of three advanced imaging modalities: DECT, CTTA, and PCT. To the best of our knowledge, this is the first study to comprehensively analyze these three modalities and integrate their parameters into a machine learning (ML) framework for RCC subtype differentiation in the same patient cohort. The results of our study reveal that each modality contributes valuable diagnostic information with unique strengths and limitations, but the integration of their findings using an ML model achieved the highest diagnostic accuracy, with an F1 score of 0.954. This highlights the synergistic value of combining DECT, CTTA, and PCT for enhanced diagnostic performance. DECT has been widely studied for its ability to differentiate ccRCC from non-ccRCC based on iodine concentration (IC), a marker of angiogenesis. Dai et al. and Zarzour et al. highlighted the utility of iodine quantification in DECT, with thresholds demonstrating excellent sensitivity and specificity for distinguishing RCC subtypes, particularly clear cell RCC (ccRCC) from papillary RCC (pRCC) and complex cysts ( 7 , 12 ). Similarly, Mileto et al. and Marcon et al. emphasized the correlation of iodine concentration with tumor perfusion and microvascular density, reinforcing its diagnostic value in differentiating RCC subtypes and tumor grades.( 13 , 14 ). However, these studies focused exclusively on DECT and often analyzed only two RCC subtypes at a time (e.g., ccRCC vs. pRCC or ccRCC vs. chrRCC), thereby oversimplifying the clinical scenario. In contrast, our study evaluates DECT in a broader context, comparing its performance with PCT and CTTA while accounting for interactions between multiple imaging parameters. The strong diagnostic performance of DECT parameters, particularly the iodine ratio (IR), in our study aligns with findings by Manoharan et al., who reported that iodine ratio effectively differentiates ccRCC from papillary subtype due to the hypervascular nature of ccRCC ( 15 ). Similarly, studies by Marcon et al. demonstrated the utility of iodine concentration (IC) in distinguishing RCC subtypes, with ccRCC showing significantly higher iodine uptake compared to papillary RCC ( 14 ). Our study reinforces these observations, with IR and IC achieving F1 scores of 0.902 and 0.859, respectively, and significant differences between ccRCC and non-ccRCC. DECT’s advantage lies in its ability to provide high diagnostic accuracy with a significantly lower radiation dose compared to PCT, making it a safer and more practical choice for patients. Additionally, the use of normalized iodine ratios in our study minimized inter-scanner variability, further enhancing the reproducibility of DECT parameters. PCT parameters, including blood flow (BF), blood volume (BV) and maximum peak intensity (MIP) showed variable diagnostic utility in our study. The MIP values in ccRCC were significantly higher (158.0 ± 38.8) compared to non-ccRCC (96.8 ± 33.1), with a p-value of < 0.001. This finding aligns with the study by Sah et al., where MIP showed promising diagnostic accuracy in distinguishing ccRCC from non-ccRCC ( 16 ). Blood flow (BF ) and Blood volume (BV) in ccRCC were also significantly higher compared to non-ccRCC. This difference mirrors findings in Rashmi et al. and Manoharan et al., where BF and BV were significantly elevated in ccRCC compared to other RCC subtypes ( 15 , 17 ).The higher radiation dose associated with PCT remains a limitation, especially in patients requiring repeated imaging. Furthermore, while PCT excels in quantifying vascularity, it is less effective in capturing intratumoral heterogeneity—a critical marker of malignancy and aggressiveness. This limitation is addressed by CTTA, which analyzes tumor texture features to provide a more nuanced understanding of tumor microarchitecture. CTTA proved particularly valuable in identifying textural differences between ccRCC and non-ccRCC, reflecting the heterogeneity of tumor biology. Although CTTA has been less extensively studied in RCC compared to DECT and PCT, its role in predicting tumor grade and subtype has gained increasing recognition. CTTA demonstrated significant potential in differentiating RCC subtypes, with entropy, a measure of image randomness, performing exceptionally well in our analysis. The higher entropy values in ccRCC compared to non-ccRCC align with findings by Deng et al., Chen et al. and Scrima et al., reflecting increased heterogeneity and vascular complexity in ccRCC ( 18 – 21 ). Similarly, Gupta et al. highlighted entropy in the corticomedullary phase as the most effective parameter for distinguishing ccRCC from non-ccRCC, achieving an F1 score of 0.83, which closely mirrors our results ( 22 ). Entropy outperformed other CTTA-derived features, such as variance, smoothness, and energy, emphasizing its diagnostic value. These results reinforce the role of CTTA as a non-invasive method for assessing tumor heterogeneity and complementing DECT in RCC characterization. Our findings support its potential as a complementary tool to traditional imaging parameters. By analyzing features such as entropy and skewness, CTTA offers unique insights into tumor characteristics that are not captured by DECT or PCT alone. A key strength of this study is the integration of DECT, PCT, and CTTA parameters into a machine learning (ML) model, which achieved the highest diagnostic accuracy, with an F1 score of 0.9541. This combined approach highlights the complementary strengths of these modalities, where DECT captures vascular characteristics, CTTA assesses textural heterogeneity, and PCT provides insights into perfusion dynamics. Prior studies, such as Budai et al., have also demonstrated the potential of radiomics-based ML models in RCC subtype differentiation ( 23 ).Similarly, Gupta et al. explored the application of a 2D deep learning architecture (FocalNet-DINO) integrated with spatial and class consistency modules, which significantly improved recall and accuracy for RCC subtype differentiation by optimizing neural network architectures ( 24 ).While Budai et al. ( 23 ) focused on radiomics features from specific contrast phases and Gupta et al. ( 24 ) on deep learning advancements, our study differs by integrating multiple advanced imaging modalities—DECT, PCT, and CTTA—into a unified ML framework. This comprehensive approach not only enhances diagnostic accuracy but also addresses the limitations of individual modalities, such as PCT's radiation dose concerns and CTTA's dependence on texture features alone. The integration of these modalities within an ML model underscores their synergistic potential for improving RCC subtype differentiation, advancing beyond radiomics-only or deep learning-only approaches, and paving the way for more personalized and precise clinical decision-making. Despite its strengths, our study has certain limitations that warrant consideration. The relatively small sample size, though sufficient for statistical analysis, may limit the generalizability of our findings. Future studies with larger and more diverse populations are needed to validate our results and extend their applicability to rare RCC subtypes, such as medullary or collecting duct carcinoma. Additionally, we did not include benign renal lesions, such as angiomyolipomas, which can mimic RCC on imaging. Including such lesions in future research would provide a more comprehensive evaluation of the diagnostic performance of these modalities. Furthermore, while our ML model demonstrated exceptional diagnostic performance, its clinical implementation requires further validation and optimization. In conclusion, this study highlights the complementary roles of DECT, PCT, and CTTA in RCC subtyping, with DECT and CTTA demonstrating superior diagnostic performance. The integration of these modalities into a combined ML model provides a powerful approach for improving diagnostic accuracy and guiding clinical decision-making. Among individual features, entropy from CTTA and IR from DECT were the most reliable predictors for distinguishing ccRCC from non-ccRCC, emphasizing their potential clinical utility. By advancing our understanding of RCC imaging biomarkers, this study contributes to the growing body of evidence supporting non-invasive, imaging-based approaches to tumor characterization. Declarations Author Contribution ANJALI SAH- did the analysis and prepared main manuscriptSneha goswami- prepared the images and tablesSanil Garg - collected data and prepared the excel sheetNeel Yadav- collected data and prepared the excel sheetAmit Gupta - collected data and guided in preparing the manuscriptChandan J dAS- helped in preparing the manuscript References Bukavina L, Bensalah K, Bray F, Carlo M, Challacombe B, Karam JA, et al. Epidemiology of Renal Cell Carcinoma: 2022 Update. European Urology. 2022 Nov;82(5):529–42. Padala SA, Barsouk A, Thandra KC, Saginala K, Mohammed A, Vakiti A, et al. Epidemiology of Renal Cell Carcinoma. World J Oncol. 2020;11(3):79–87. Clark PE. The role of VHL in clear-cell renal cell carcinoma and its relation to targeted therapy. Kidney International. 2009 Nov;76(9):939–45. Mendiratta P, Rini BI, Ornstein MC. Emerging immunotherapy in advanced renal cell carcinoma. Urologic Oncology: Seminars and Original Investigations. 2017 Dec;35(12):687–93. Patel HD, Johnson MH, Pierorazio PM, Sozio SM, Sharma R, Iyoha E, et al. Diagnostic Accuracy and Risks of Biopsy in the Diagnosis of a Renal Mass Suspicious for Localized Renal Cell Carcinoma: Systematic Review of the Literature. J Urol. 2016 May;195(5):1340–7. Das CJ. Perfusion computed tomography in renal cell carcinoma. WJR. 2015;7(7):170. Dai C, Cao Y, Jia Y, Ding Y, Sheng R, Zeng M, et al. Differentiation of renal cell carcinoma subtypes with different iodine quantification methods using single-phase contrast-enhanced dual-energy CT: areal vs. volumetric analyses. Abdom Radiol. 2018 Mar;43(3):672–8. Lubner MG, Smith AD, Sandrasegaran K, Sahani DV, Pickhardt PJ. CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges. RadioGraphics. 2017 Sep;37(5):1483–503. Ganeshan B, Miles KA. Quantifying tumour heterogeneity with CT. Cancer Imaging. 2013;13(1):140–9. Zhang GMY, Shi B, Xue HD, Ganeshan B, Sun H, Jin ZY. Can quantitative CT texture analysis be used to differentiate subtypes of renal cell carcinoma? Clinical Radiology. 2019 Apr;74(4):287–94. Hearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines. IEEE Intell Syst Their Appl. 1998 Jul;13(4):18–28. Zarzour JG, Milner D, Valentin R, Jackson BE, Gordetsky J, West J, et al. Quantitative iodine content threshold for discrimination of renal cell carcinomas using rapid kV-switching dual-energy CT. Abdom Radiol. 2017 Mar;42(3):727–34. Mileto A, Marin D, Alfaro-Cordoba M, Ramirez-Giraldo JC, Eusemann CD, Scribano E, et al. Iodine Quantification to Distinguish Clear Cell from Papillary Renal Cell Carcinoma at Dual-Energy Multidetector CT: A Multireader Diagnostic Performance Study. Radiology. 2014 Dec;273(3):813–20. Marcon J, Graser A, Horst D, Casuscelli J, Spek A, Stief CG, et al. Papillary vs clear cell renal cell carcinoma. Differentiation and grading by iodine concentration using DECT—correlation with microvascular density. Eur Radiol. 2020 Jan;30(1):1–10. Manoharan D, Netaji A, Diwan K, Sharma S. Normalized Dual-Energy Iodine Ratio Best Differentiates Renal Cell Carcinoma Subtypes Among Quantitative Imaging Biomarkers From Perfusion CT and Dual-Energy CT. American Journal of Roentgenology. 2020 Dec;215(6):1389–97. Sah A, Gupta A, Garg S, Yadav N, Khan MA, Das CJ. Can quantitative perfusion CT-based biomarkers predict renal cell carcinoma subtypes?. Abdominal Radiology. 2024 Dec 17:1-9. Chauhan SS, Dixit R, Chowdhury V, Khurana N. Perfusion CT in the Evaluation of Renal Cell Carcinoma. JCDR.2022:13(1):2046-2056. Deng Y, Soule E, Samuel A, Shah S, Cui E, Asare-Sawiri M, et al. CT texture analysis in the differentiation of major renal cell carcinoma subtypes and correlation with Fuhrman grade. Eur Radiol. 2019 Dec;29(12):6922–9. Chen M, Yin F, Yu Y, Zhang H, Wen G. CT-based multi-phase Radiomic models for differentiating clear cell renal cell carcinoma. Cancer Imaging. 2021 Dec;21(1):42. Scrima AT, Lubner MG, Abel EJ, Havighurst TC, Shapiro DD, Huang W, et al. Texture analysis of small renal cell carcinomas at MDCT for predicting relevant histologic and protein biomarkers. Abdom Radiol. 2019 Jun;44(6):1999–2008. Lubner MG, Stabo N, Abel EJ, Del Rio AM, Pickhardt PJ. CT Textural Analysis of Large Primary Renal Cell Carcinomas: Pretreatment Tumor Heterogeneity Correlates With Histologic Findings and Clinical Outcomes. American Journal of Roentgenology. 2016 Jul;207(1):96–105. Gupta A, Garg S, Yadav N, Dhanakshirur RR, Jain K, Nayyar R, Kaushal S, Das CJ. Predicting Renal Cell Carcinoma Subtypes and Fuhrman Grading Using Multiphasic CT-Based Texture Analysis and Machine Learning Techniques. Indian Journal of Radiology and Imaging. 2024 Dec 11. Budai BK, Stollmayer R, Rónaszéki AD, Körmendy B, Zsombor Z, Palotás L, et al. Radiomics analysis of contrast-enhanced CT scans can distinguish between clear cell and non-clear cell renal cell carcinoma in different imaging protocols. Front Med. 2022 Oct 13;9:974485. Gupta A, Dhanakshirur RR, Jain K, Garg S, Yadav N, Seth A, Das CJ. Deep Learning for Detecting and Subtyping Renal Cell Carcinoma on Contrast-Enhanced CT Scans Using 2D Neural Network with Feature Consistency Techniques. Indian Journal of Radiology and Imaging. 2024 Dec 11. Tables Table 1 : Demographic details of the patients Characteristics Patients , n= 66 Age in years, median ( range) 68 ( 43-79) Sex -Male -Female 40 ( 60%) 26 (40%) Clear cell RCC 52 (78.7%) Non clear cell RCC - Papillary - Chromophobe 10 (15.1%) 4 (6%) Table 2: Inter-reader variability of various PCT and DECT parameters Parameters Pearson Correlation (r) Spearman Correlation (ρ) Chi Square Test P-value Perfusion BV 0.935 0.946 370.21 <0.0005 BF 0.837 0.922 721.51 <0.0005 MTT 0.35 0.422 88.89 0.629 TTP 0.311 0.373 73.96 0.937 DECT IC 0.894 0.894 7.29 <0.0005 IR 0.889 0.862 173.6 <0.0005 Table 3 Comparison of features between clear cell renal cell carcinoma (ccRCC) and non-clear cell renal cell carcinoma (non-ccRCC) across the three imaging modalities: Study Feature ccRCC non-ccRCC p-value F1 Score Mean Std. dev. Mean Std. dev. CT Texture Analysis (n = 66) Mean 167.30 24.7 142.10 27.89 0.018 0.766 Variance 1440.0 663.0 505.0 235.0 0.01 0.710 Energy 1.62 1.46 1.19 0.972 0.169 0.654 Entropy 7.04 0.298 6.33 0.295 <0.001 0.842 Smoothness 0.986 0.020 0.982 0.017 0.482 0.597 ML Model 0.937 0.165 0.417 0.243 <0.001 0.934 CT Perfusion (n = 66) BF 883 780 532.0 430.0 0.04 0.766 BV 118 85.6 70.5 60.6 0.03 0.728 MTT 11.17 3.33 12.86 4.17 0.53 0.548 TTP 11.46 2.85 13.52 3.87 0.17 0.643 MIP 158 38.8 96.8 33.1 <0.001 0.848 ML Model 0.93 0.15 0.55 0.25 <0.001 0.910 DECT (n = 66) IR 65.12 23.73 35.17 17.99 <0.001 0.910 IC 2.3 0.79 1.42 0.75 0.005 0.894 ML Model 0.95 0.14 0.41 0.29 <0.001 0.935 Overall ML Model 0.97 0.13 0.36 0.31 <0.001 0.954 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5952087","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":412112706,"identity":"be8692aa-1a16-40e5-9a34-2466ac28adb1","order_by":0,"name":"ANJALI SAH","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"ANJALI","middleName":"","lastName":"SAH","suffix":""},{"id":412112707,"identity":"8bc6c7ff-dd33-4c0b-88c3-3939f3993081","order_by":1,"name":"SNEHA GOSWAMI","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"SNEHA","middleName":"","lastName":"GOSWAMI","suffix":""},{"id":412112708,"identity":"b637b587-b516-4e96-85fc-07f43bc5e0f7","order_by":2,"name":"AMIT GUPTA","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"AMIT","middleName":"","lastName":"GUPTA","suffix":""},{"id":412112709,"identity":"8700bc19-c2fd-46aa-81ec-df1f60a0a5cd","order_by":3,"name":"SANIL GARG","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"SANIL","middleName":"","lastName":"GARG","suffix":""},{"id":412112710,"identity":"31fc6044-d8ac-4762-824d-5b949e0c0a29","order_by":4,"name":"NEEL YADAV","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"NEEL","middleName":"","lastName":"YADAV","suffix":""},{"id":412112711,"identity":"e2e689ad-ed00-44dd-ab05-b8c40b8e0d4c","order_by":5,"name":"CHANDAN J DAS","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsUlEQVRIiWNgGAWjYDCCA0AsUcHAYECiljMka2FsI0UL3/HjDz9Yzjssb87efIDhR8U2wlokz+QYS0huO2y4s+dYAmPPmduEtRgcyGFjAGph3HAjx4CZsY0YLeefP2OQnHPYngQtNxLMGCQbDicSr0XyxhtjCYlj6ckbzhxLOEiUX/jOpz/8LFFjbbvhePPBBz8qiNACAswSDM1gxgHi1AMB4weGOqIVj4JRMApGwQgEAG7qQUJa6vcLAAAAAElFTkSuQmCC","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":true,"prefix":"","firstName":"CHANDAN","middleName":"J","lastName":"DAS","suffix":""}],"badges":[],"createdAt":"2025-02-03 15:08:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5952087/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5952087/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":76113986,"identity":"c921f4e0-bc43-4779-9264-22dc68202d95","added_by":"auto","created_at":"2025-02-12 12:24:33","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":99253,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Slide1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5952087/v1/d6e814f5fa64f6d5cf084060.jpg"},{"id":76113969,"identity":"790d4847-bd6f-4482-bbde-ff992eae2650","added_by":"auto","created_at":"2025-02-12 12:24:30","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":137672,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Slide2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5952087/v1/8cc0e3deac540c544ceaef5d.jpg"},{"id":76113988,"identity":"6f661f10-9183-451a-89bf-0e53b6db68cb","added_by":"auto","created_at":"2025-02-12 12:24:33","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":82883,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Slide3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5952087/v1/62a1fb7d99d2c86dd39410d2.jpg"},{"id":76114383,"identity":"c2d1db96-01fc-4066-9f48-cc2d82f5f9c3","added_by":"auto","created_at":"2025-02-12 12:32:34","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":58247,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Slide4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5952087/v1/6aa7743091636312bc6c248c.jpg"},{"id":76115357,"identity":"b5f78b6a-1f38-4994-9b50-3e528736bd32","added_by":"auto","created_at":"2025-02-12 12:40:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1136625,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5952087/v1/83bba40f-f833-4401-8413-02c3707aff71.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eQuantitative CT-based biomarkers for predicting Renal cell carcinoma subtypes: a comparison of Dual-Energy CT, Perfusion CT, and CT texture parameters\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eRenal cell carcinoma (RCC) accounts for approximately 85 to 90% of all renal tumors and is by far the most prevalent primary renal malignancy. Clear-cell renal cell carcinoma (ccRCC) is the most common and aggressive subtype of RCC, accounting for approximately 70% of cases (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Known for its high morbidity and mortality, ccRCC requires timely diagnosis and targeted treatment to improve patient outcomes (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). The distinction between ccRCC and non-ccRCC is clinically significant, as it directly influences therapeutic decisions, particularly with the advent of targeted treatments such as tyrosine kinase inhibitors, vascular endothelial growth factor (VEGF) inhibitors, and immune checkpoint inhibitors (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). These treatments have shown marked efficacy in ccRCC due to its molecular underpinnings, while non-ccRCC treatment strategies often rely on extrapolated data from ccRCC trials, underscoring the need for better diagnostic and therapeutic specificity.\u003c/p\u003e \u003cp\u003eCurrently, tumor biopsy remains the gold standard for determining RCC subtype and guiding systemic therapy. However, this invasive procedure carries risks such as bleeding, infection, and sampling error, particularly for heterogeneous tumors or those in anatomically challenging locations (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Furthermore, traditional imaging methods like multiphase CT and MRI often fail to reliably differentiate RCC subtypes due to overlapping radiological features. This limitation underscores the need for non-invasive, accurate, and reproducible imaging biomarkers that can distinguish RCC subtypes with greater precision.\u003c/p\u003e \u003cp\u003eQuantitative imaging techniques such as Dual-Energy CT (DECT), Perfusion CT (PCT), and CT Texture Analysis (CTTA) have emerged as promising tools to address this diagnostic challenge. These modalities provide detailed insights into tumor vascularity, perfusion, and tissue heterogeneity, reflecting the underlying tumor biology. PCT, for instance, assesses perfusion parameters such as blood flow, blood volume, and mean transit time, which are particularly relevant given the highly vascular nature of ccRCC compared to less vascular non-ccRCC subtypes like papillary and chromophobe RCC (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). DECT, with its ability to analyze material-specific attenuation properties at different energy levels, offers unique quantitative metrics for tumor characterization and differentiation of RCC subtypes (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCTTA is a powerful tool that analyzes the distribution and patterns of pixel intensities within a tumor on CT images, providing insights into tumor heterogeneity that are beyond what one\u0026rsquo;s eye can see (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). It evaluates features like entropy (irregularity), variance, skewness (asymmetry), and kurtosis (peakedness), which reflect the tumor's internal complexity and structure. It has shown great potential in characterizing and prognosticating various malignancies, including RCC, by quantifying tumor complexity and internal architecture (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Unlike PCT, it can be applied retrospectively to existing imaging data, making it a practical option for personalized diagnosis and treatment planning.\u003c/p\u003e \u003cp\u003eIncorporating these advanced imaging biomarkers could transform RCC diagnosis by offering a non-invasive alternative to biopsies and improving treatment stratification. The ability to differentiate ccRCC from non-ccRCC with high accuracy is crucial, given the variability in response to targeted therapies and the growing emphasis on personalized oncology.\u003c/p\u003e \u003cp\u003eThis study seeks to bridge this gap by assessing the diagnostic performance of quantitative parameters derived from PCT, DECT, and CTTA within the same patient cohort. By integrating these advanced imaging modalities, the study aims to establish a robust, non-invasive framework for RCC subtype differentiation, ultimately guiding personalized treatment strategies and improving clinical outcomes.\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cp\u003e This retrospective study was approved by the institutional review board and the need for informed consent was waived. The CT dataset was compiled by reviewing hospital records to identify patients diagnosed with RCC between January 2017 and December 2022 who had undergone both DECT and PCT prior to radical or partial nephrectomy. The details of patients inclusion and exclusion is shown in a flow chart (Fig.\u0026nbsp;1). We initially identified 85 patients with pathologically confirmed renal cell carcinoma (RCC) who underwent preoperative PCT and DECT between January 2017 and December 2022 (Fig.\u0026nbsp;1). Of these, 19 patients were excluded based on the following criteria: inadequate image quality due to poor timing of post-contrast phases or motion artifacts (n\u0026thinsp;=\u0026thinsp;10), history of prior systemic therapy or surgery (n\u0026thinsp;=\u0026thinsp;9). A total of 66 patients met the inclusion criteria, comprising 52 cases of ccRCC) and 14 cases of non-ccRCC.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePerfusion CT protocol\u003c/h2\u003e \u003cp\u003ePerfusion CT (PCT) scans were performed in DECT mode using a 64-slice dual-source, dual-energy 2 x 128-section multi-detector CT scanner (Somatom Definition Flash, Siemens Healthineers, Germany). The current protocol in our department includes a non-contrast (NCCT) single scan-and-view acquisition performed at low dose with shallow breathing to ensure accurate positioning. Following this, 70\u0026ndash;90 mL of non-ionic contrast medium (350 mg iodine/mL,Omnipaque 350, GE Healthcare) was injected at a rate of 5 mL/s, followed by a 40 mL saline chaser bolus. The dynamic imaging sequence started 5 seconds after the contrast injection, consisting of 34 acquisitions with a rotation time of 0.5 seconds and a cycle time of 1.5 seconds, lasting a total of 45 seconds. These images were acquired mostly in cortico-medullary (CM) phase. This was followed by the nephrographic (NP) phase at 60\u0026ndash;70 seconds and excretory (DEL) phase images at 240 seconds. Each of the 34 acquisitions included three consecutive 5 mm slices reconstructed across an 11 cm craniocaudal region. Perfusion parameters such as tumor blood flow (TBF), tumor blood volume (TBV), mean transit time (MTT), and vascular permeability-surface area product were calculated. Blood flow (BF), blood volume (BV), and flow extraction product (K) were derived using the maximum-slope and delay-corrected modified Patlak approach and correlated with microvessel density (MVD). Data analysis was conducted using dedicated software (Siemens Healthineers, volume PCT and Syngo MMWP), with the arterial input defined by drawing a circular ROI on the abdominal aorta, after applying the inbuilt motion correction algorithm. Perfusion maps were generated using a deconvolution pharmacokinetic model. For each tumor, the largest cross-section was selected for ROI placement by two independent readers having 5 and 7 years of experience, who were blinded to histopathological results. In heterogeneous tumors, ROIs were drawn to include the most enhancing areas while excluding necrotic regions.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDual-Energy CT Protocol\u003c/h3\u003e\n\u003cp\u003eFor DECT analysis, nephrographic (NP) images were selected from the above-mentioned protocol. The retrieved CT scans were anonymized during archiving and a new unique ID was given to the scans.\u003c/p\u003e \u003cp\u003eColor-coded iodine maps were created using the iodine subtraction method (Liver VNC, Siemens Healthineers) and dual-energy software (Syngovia VB10A, Siemens Healthineers) for DECT data analysis. Similar regions of interest (ROIs) were drawn by the readers (mean size: 0.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.67 cm\u0026sup2;, range: 0.1\u0026ndash;5.7 cm\u0026sup2;) within the tumor region exhibiting maximum iodine uptake on DECT. Iodine concentration (IC) in milligrams per milliliter and the iodine ratio (IC in tumor / IC in the aorta at the level of the renal artery supplying the kidney with the tumor) were measured to assess the tumor\u0026rsquo;s iodine enhancement pattern and vascularity.\u003c/p\u003e\n\u003ch3\u003eTexture analysis\u003c/h3\u003e\n\u003cp\u003eFor each patient, a single axial CT slice representing the largest tumor cross-section was extracted from each available CT phase and saved as a JPEG file. A single radiologist (9 years of experience in body imaging) manually annotated the tumor on these CT images using the 'Polygonal ROI' function in the freely available \u0026lsquo;LabelMe\u0026rsquo; tool. This annotation process was applied to all available CT phases for each patient, and the annotated data was saved as a JSON file for each tumor region. Image pre-processing was performed using the open-source OpenCV library.\u003c/p\u003e \u003cp\u003eFor the CTTA, texture parameters were computed using open-source Python libraries including \u0026ldquo;scipy,\u0026rdquo; \u0026ldquo;numpy,\u0026rdquo; and \u0026ldquo;opencv.\u0026rdquo; First, second, and third-order texture features were extracted from each tumor annotation, encompassing parameters such as mean, variance, energy, entropy, and smoothness. These texture features were used to capture tumor heterogeneity, providing valuable insights into the tumor microenvironment.\u003c/p\u003e \u003cp\u003eThe details of DECT, PCT, and CTTA analysis is shown in Fig.\u0026nbsp;2.\u003c/p\u003e\n\u003ch3\u003eThe ML models for DECT, PCT, and CTTA:\u003c/h3\u003e\n\u003cp\u003eWe employ Support Vector Machines (SVM), which are supervised learning methods, for machine learning (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). SVM is a technique that determines the most efficient way to draw a line\u0026mdash;or, in more complicated situations, a plane or a hyperplane\u0026mdash;between groupings of data points to distinguish them (in our example, cc-RCC vs non cc- RCC). This line is created such that it is as far away from the nearest points of each group as feasible, ensuring the most clear differentiation between the various types of tumors. This architecture is more resistant to outliers and has a faster convergence rate to its final model. CTTA, DECT, and PCT parameters obtained from each CT phase independently were used to train SVM models for RCC subtyping (ccRCC vs. non-ccRCC). The SVM model was developed in two steps namely normalization and feature scaling followed by model training. Using a technique known as min-max normalization, each parameter was normalized from 0 to 1. This phase is critical because it places all parameters on an equal scale, allowing them to contribute equally to the model regardless of their original scale or units. For model training, we did not use any prior models when creating our SVM model. Another parameter taken into account was the prediction label for each sample in the testing dataset. For training, validation, and test datasets, we employed a 70:20:10 split, accordingly. A standard machine learning library called sklearn, which offers resources for creating and implementing machine learning models, was used for the training process.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed by using STATA 18 statistical software. Continuous variables were expressed as means\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviations (SD) or medians with interquartile ranges (IQR), while categorical variables were presented as frequencies or percentages. Interreader agreement for the DECT and PCT parameters was calculated using the Pearson correlation (r) and Spearman correlation coefficient (ρ). Using the \"matplotlib\" and \"seaborn'' libraries of Python, we ran multivariate logistic regression analyses on the PCT, DECT, and CTTA models to determine how well each feature predicted tumor type. The first-order features, such as mean, variance, standard deviation, etc., were computed on this image using the standard numpy library. The second-order features, such as energy, entropy, and smoothness, were calculated using the Scipy library. The corresponding p-values and F1 scores were recorded for CTTA, DECT, and PCT parameters. Box plots were made for each of the parameters to differentiate between ccRCC and non-ccRCC.\u003c/p\u003e \u003cp\u003eThe trained SVM models were also subjected to the same statistical analysis as the individual quantitative parameters from the CTTA, DECT, and PCT analyses. This study repetition is critical for validating and comparing the performance of our SVM models to the quantitative parameter analysis.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eThis study included 66 patients with histopathologically confirmed renal cell carcinoma, comprising 52 cases of ccRCC and 14 cases of non-ccRCC. The study population included 40 males and 26 females, with a median age at presentation of 68 (range: 43\u0026ndash;79 years). Demographic details of the patients are summarised in Table 1. The interreader agreement for DECT and PCT parameters is summarized in Table 2. Excellent agreement was observed for DECT parameters, with Pearson (r) and Spearman (\u0026rho;) correlation coefficients of 0.894 and 0.894 for iodine concentration (IC) and 0.889 and 0.862 for iodine ratio (IR), respectively (p \u0026lt; 0.0005). Among PCT parameters, blood volume (BV) and blood flow (BF) also demonstrated strong agreement (r = 0.935, \u0026rho; = 0.946 for BV; r = 0.837, \u0026rho; = 0.922 for BF, p \u0026lt; 0.05). However, MTT and TTP showed weaker agreement, with r = 0.35, \u0026rho; = 0.422 for MTT (p = 0.629) and r = 0.311, \u0026rho; = 0.373 for TTP (p = 0.937).\u003c/p\u003e\n\u003cp\u003eThe ML models for DECT, PCT, and CTTA showed strong diagnostic performance in differentiating ccRCC from non-ccRCC, as shown in Table 3. DECT achieved the highest F1 score among individual modalities (F1 = 0.935), followed closely by CTTA (F1 = 0.934) and PCT (F1 = 0.910). The combined ML model incorporating parameters from all three modalities yielded the highest diagnostic accuracy, with an F1 score of 0.954.\u003c/p\u003e\n\u003cp\u003eAmong both the DECT features, iodine ratio (IR) was the best-performing parameter, achieving an F1 score of 0.902 (Table 3).Iodine ratio (IR) was markedly elevated in ccRCC, with a mean value \u0026nbsp;of 65.12\u003cu\u003e+\u003c/u\u003e23.73 compared to 35.17 \u003cu\u003e+\u003c/u\u003e 17.99 in non-ccRCC (p \u0026lt; 0.001). Similarly, iodine concentration (IC) was also notably higher in ccRCC, with a mean value of 2.3 \u003cu\u003e+\u003c/u\u003e 0.79 compared to 1.42 \u003cu\u003e+\u003c/u\u003e 0.75 in non-ccRCC (p \u0026lt; 0.05). Figure 3 shows a box plot comparing IR values across the two tumor subtypes.\u003c/p\u003e\n\u003cp\u003eIn texture analysis, \u003cstrong\u003eentropy\u003c/strong\u003e emerged as the most effective parameters for distinguishing ccRCC from non-ccRCC, achieving an \u003cstrong\u003eAUC of 0.94\u003c/strong\u003e. Entropy was significantly higher in ccRCC (\u003cstrong\u003e7.94 \u0026plusmn; 0.336\u003c/strong\u003e) compared to non-ccRCC (\u003cstrong\u003e6.43 \u0026plusmn; 0.297\u003c/strong\u003e\u003cstrong\u003e, \u003cstrong\u003ep \u0026lt; 0.001\u003c/strong\u003e\u003c/strong\u003e), with a threshold value of 7.21 yielding \u003cstrong\u003e84.8% accuracy\u003c/strong\u003e\u003cstrong\u003e, \u003cstrong\u003e83.0% sensitivity\u003c/strong\u003e\u003c/strong\u003e, and \u003cstrong\u003e92.3% specificity\u003c/strong\u003e. Figure 4 shows the ROC curve. \u003cstrong\u003eVariance\u003c/strong\u003e also demonstrated strong performance, with ccRCC showing higher values (\u003cstrong\u003e1440.0 \u0026plusmn; 663.0\u003c/strong\u003e) than non-ccRCC (505.0 +235.0, \u003cstrong\u003ep \u0026lt; 0.001\u003c/strong\u003e), achieving an \u003cstrong\u003eAUC of 0.874\u003c/strong\u003e, 84.8% accuracy, 86.8% sensitivity, and 76.9% specificity at a threshold of 930. Similarly, mean values were also higher in clear cell than non clear cell carcinoma (p=0.01). Energy and Smoothness was higher in clear cell than non clear cell subtypes but the difference was not significant.\u003c/p\u003e\n\u003cp\u003ePCT parameters demonstrated lower diagnostic performance compared to DECT and CTTA.\u0026nbsp;The \u003cstrong\u003emaximum peak intensity (MIP)\u003c/strong\u003e showed significantly higher values in ccRCC (158.0 \u0026plusmn; 38.8) compared to non-ccRCC (96.8 \u0026plusmn; 33.1), with a p-value \u0026lt; 0.001, demonstrating high diagnostic accuracy (AUC: 0.884), with sensitivity of 84.9%, and specificity of 84.6%. Blood flow (BF) also differed significantly between the two groups, with ccRCC exhibiting a mean value of 883 \u003cu\u003e+\u003c/u\u003e 780 compared to 532.0 \u003cu\u003e+\u003c/u\u003e 430.0 \u0026nbsp;in non-ccRCC (p = 0.04). Lastly, \u003cstrong\u003eblood volume (BV)\u003c/strong\u003e was significantly higher in ccRCC (118.0 \u003cu\u003e+\u003c/u\u003e 85.6) than in non-ccRCC (70.5 \u003cu\u003e+\u003c/u\u003e 60.6), with a p-value of 0.02. Although the mean TTP and MTT were higher in clear cell compared to non clear cell RCC, but the difference was not significant (p\u0026gt;0.05)\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe differentiation of RCCs into ccRCC and non-ccRCC subtypes is crucial for personalized treatment strategies due to significant differences in their biology, prognosis, and therapeutic response. Our study aimed to evaluate and compare the diagnostic performance of three advanced imaging modalities: DECT, CTTA, and PCT. To the best of our knowledge, this is the first study to comprehensively analyze these three modalities and integrate their parameters into a machine learning (ML) framework for RCC subtype differentiation in the same patient cohort. The results of our study reveal that each modality contributes valuable diagnostic information with unique strengths and limitations, but the integration of their findings using an ML model achieved the highest diagnostic accuracy, with an F1 score of 0.954. This highlights the synergistic value of combining DECT, CTTA, and PCT for enhanced diagnostic performance.\u003c/p\u003e \u003cp\u003eDECT has been widely studied for its ability to differentiate ccRCC from non-ccRCC based on iodine concentration (IC), a marker of angiogenesis. Dai et al. and Zarzour et al. highlighted the utility of iodine quantification in DECT, with thresholds demonstrating excellent sensitivity and specificity for distinguishing RCC subtypes, particularly clear cell RCC (ccRCC) from papillary RCC (pRCC) and complex cysts (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Similarly, Mileto et al. and Marcon et al. emphasized the correlation of iodine concentration with tumor perfusion and microvascular density, reinforcing its diagnostic value in differentiating RCC subtypes and tumor grades.(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). However, these studies focused exclusively on DECT and often analyzed only two RCC subtypes at a time (e.g., ccRCC vs. pRCC or ccRCC vs. chrRCC), thereby oversimplifying the clinical scenario. In contrast, our study evaluates DECT in a broader context, comparing its performance with PCT and CTTA while accounting for interactions between multiple imaging parameters. The strong diagnostic performance of DECT parameters, particularly the iodine ratio (IR), in our study aligns with findings by Manoharan et al., who reported that iodine ratio effectively differentiates ccRCC from papillary subtype due to the hypervascular nature of ccRCC (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). Similarly, studies by Marcon et al. demonstrated the utility of iodine concentration (IC) in distinguishing RCC subtypes, with ccRCC showing significantly higher iodine uptake compared to papillary RCC (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Our study reinforces these observations, with IR and IC achieving F1 scores of 0.902 and 0.859, respectively, and significant differences between ccRCC and non-ccRCC. DECT\u0026rsquo;s advantage lies in its ability to provide high diagnostic accuracy with a significantly lower radiation dose compared to PCT, making it a safer and more practical choice for patients. Additionally, the use of normalized iodine ratios in our study minimized inter-scanner variability, further enhancing the reproducibility of DECT parameters.\u003c/p\u003e \u003cp\u003ePCT parameters, including blood flow (BF), blood volume (BV) and maximum peak intensity (MIP) showed variable diagnostic utility in our study. The MIP values in ccRCC were significantly higher (158.0\u0026thinsp;\u0026plusmn;\u0026thinsp;38.8) compared to non-ccRCC (96.8\u0026thinsp;\u0026plusmn;\u0026thinsp;33.1), with a p-value of \u0026lt;\u0026thinsp;0.001. This finding aligns with the study by Sah et al., where MIP showed promising diagnostic accuracy in distinguishing ccRCC from non-ccRCC (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). Blood flow (BF ) and Blood volume (BV) in ccRCC were also significantly higher compared to non-ccRCC. This difference mirrors findings in Rashmi et al. and Manoharan et al., where BF and BV were significantly elevated in ccRCC compared to other RCC subtypes (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e).The higher radiation dose associated with PCT remains a limitation, especially in patients requiring repeated imaging. Furthermore, while PCT excels in quantifying vascularity, it is less effective in capturing intratumoral heterogeneity\u0026mdash;a critical marker of malignancy and aggressiveness. This limitation is addressed by CTTA, which analyzes tumor texture features to provide a more nuanced understanding of tumor microarchitecture.\u003c/p\u003e \u003cp\u003eCTTA proved particularly valuable in identifying textural differences between ccRCC and non-ccRCC, reflecting the heterogeneity of tumor biology. Although CTTA has been less extensively studied in RCC compared to DECT and PCT, its role in predicting tumor grade and subtype has gained increasing recognition. CTTA demonstrated significant potential in differentiating RCC subtypes, with entropy, a measure of image randomness, performing exceptionally well in our analysis. The higher entropy values in ccRCC compared to non-ccRCC align with findings by Deng et al., Chen et al. and Scrima et al., reflecting increased heterogeneity and vascular complexity in ccRCC (\u003cspan additionalcitationids=\"CR19 CR20\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). Similarly, Gupta et al. highlighted entropy in the corticomedullary phase as the most effective parameter for distinguishing ccRCC from non-ccRCC, achieving an F1 score of 0.83, which closely mirrors our results (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). Entropy outperformed other CTTA-derived features, such as variance, smoothness, and energy, emphasizing its diagnostic value. These results reinforce the role of CTTA as a non-invasive method for assessing tumor heterogeneity and complementing DECT in RCC characterization. Our findings support its potential as a complementary tool to traditional imaging parameters. By analyzing features such as entropy and skewness, CTTA offers unique insights into tumor characteristics that are not captured by DECT or PCT alone.\u003c/p\u003e \u003cp\u003eA key strength of this study is the integration of DECT, PCT, and CTTA parameters into a machine learning (ML) model, which achieved the highest diagnostic accuracy, with an F1 score of 0.9541. This combined approach highlights the complementary strengths of these modalities, where DECT captures vascular characteristics, CTTA assesses textural heterogeneity, and PCT provides insights into perfusion dynamics. Prior studies, such as Budai et al., have also demonstrated the potential of radiomics-based ML models in RCC subtype differentiation (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e).Similarly, Gupta et al. explored the application of a 2D deep learning architecture (FocalNet-DINO) integrated with spatial and class consistency modules, which significantly improved recall and accuracy for RCC subtype differentiation by optimizing neural network architectures (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e).While Budai et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e) focused on radiomics features from specific contrast phases and Gupta et al. (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e) on deep learning advancements, our study differs by integrating multiple advanced imaging modalities\u0026mdash;DECT, PCT, and CTTA\u0026mdash;into a unified ML framework. This comprehensive approach not only enhances diagnostic accuracy but also addresses the limitations of individual modalities, such as PCT's radiation dose concerns and CTTA's dependence on texture features alone. The integration of these modalities within an ML model underscores their synergistic potential for improving RCC subtype differentiation, advancing beyond radiomics-only or deep learning-only approaches, and paving the way for more personalized and precise clinical decision-making.\u003c/p\u003e \u003cp\u003eDespite its strengths, our study has certain limitations that warrant consideration. The relatively small sample size, though sufficient for statistical analysis, may limit the generalizability of our findings. Future studies with larger and more diverse populations are needed to validate our results and extend their applicability to rare RCC subtypes, such as medullary or collecting duct carcinoma. Additionally, we did not include benign renal lesions, such as angiomyolipomas, which can mimic RCC on imaging. Including such lesions in future research would provide a more comprehensive evaluation of the diagnostic performance of these modalities. Furthermore, while our ML model demonstrated exceptional diagnostic performance, its clinical implementation requires further validation and optimization.\u003c/p\u003e \u003cp\u003eIn conclusion, this study highlights the complementary roles of DECT, PCT, and CTTA in RCC subtyping, with DECT and CTTA demonstrating superior diagnostic performance. The integration of these modalities into a combined ML model provides a powerful approach for improving diagnostic accuracy and guiding clinical decision-making. Among individual features, entropy from CTTA and IR from DECT were the most reliable predictors for distinguishing ccRCC from non-ccRCC, emphasizing their potential clinical utility. By advancing our understanding of RCC imaging biomarkers, this study contributes to the growing body of evidence supporting non-invasive, imaging-based approaches to tumor characterization.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eANJALI SAH- did the analysis and prepared main manuscriptSneha goswami- prepared the images and tablesSanil Garg - collected data and prepared the excel sheetNeel Yadav- collected data and prepared the excel sheetAmit Gupta - collected data and guided in preparing the manuscriptChandan J dAS- helped in preparing the manuscript\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBukavina L, Bensalah K, Bray F, Carlo M, Challacombe B, Karam JA, et al. Epidemiology of Renal Cell Carcinoma: 2022 Update. European Urology. 2022 Nov;82(5):529\u0026ndash;42.\u003c/li\u003e\n\u003cli\u003ePadala SA, Barsouk A, Thandra KC, Saginala K, Mohammed A, Vakiti A, et al. Epidemiology of Renal Cell Carcinoma. World J Oncol. 2020;11(3):79\u0026ndash;87.\u003c/li\u003e\n\u003cli\u003eClark PE. The role of VHL in clear-cell renal cell carcinoma and its relation to targeted therapy. Kidney International. 2009 Nov;76(9):939\u0026ndash;45.\u003c/li\u003e\n\u003cli\u003eMendiratta P, Rini BI, Ornstein MC. Emerging immunotherapy in advanced renal cell carcinoma. Urologic Oncology: Seminars and Original Investigations. 2017 Dec;35(12):687\u0026ndash;93.\u003c/li\u003e\n\u003cli\u003ePatel HD, Johnson MH, Pierorazio PM, Sozio SM, Sharma R, Iyoha E, et al. Diagnostic Accuracy and Risks of Biopsy in the Diagnosis of a Renal Mass Suspicious for Localized Renal Cell Carcinoma: Systematic Review of the Literature. J Urol. 2016 May;195(5):1340\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eDas CJ. Perfusion computed tomography in renal cell carcinoma. WJR. 2015;7(7):170.\u003c/li\u003e\n\u003cli\u003eDai C, Cao Y, Jia Y, Ding Y, Sheng R, Zeng M, et al. Differentiation of renal cell carcinoma subtypes with different iodine quantification methods using single-phase contrast-enhanced dual-energy CT: areal vs. volumetric analyses. Abdom Radiol. 2018 Mar;43(3):672\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eLubner MG, Smith AD, Sandrasegaran K, Sahani DV, Pickhardt PJ. CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges. RadioGraphics. 2017 Sep;37(5):1483\u0026ndash;503.\u003c/li\u003e\n\u003cli\u003eGaneshan B, Miles KA. Quantifying tumour heterogeneity with CT. Cancer Imaging. 2013;13(1):140\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eZhang GMY, Shi B, Xue HD, Ganeshan B, Sun H, Jin ZY. Can quantitative CT texture analysis be used to differentiate subtypes of renal cell carcinoma? Clinical Radiology. 2019 Apr;74(4):287\u0026ndash;94.\u003c/li\u003e\n\u003cli\u003eHearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines. IEEE Intell Syst Their Appl. 1998 Jul;13(4):18\u0026ndash;28. \u003c/li\u003e\n\u003cli\u003eZarzour JG, Milner D, Valentin R, Jackson BE, Gordetsky J, West J, et al. Quantitative iodine content threshold for discrimination of renal cell carcinomas using rapid kV-switching dual-energy CT. Abdom Radiol. 2017 Mar;42(3):727\u0026ndash;34.\u003c/li\u003e\n\u003cli\u003eMileto A, Marin D, Alfaro-Cordoba M, Ramirez-Giraldo JC, Eusemann CD, Scribano E, et al. Iodine Quantification to Distinguish Clear Cell from Papillary Renal Cell Carcinoma at Dual-Energy Multidetector CT: A Multireader Diagnostic Performance Study. Radiology. 2014 Dec;273(3):813\u0026ndash;20. \u003c/li\u003e\n\u003cli\u003eMarcon J, Graser A, Horst D, Casuscelli J, Spek A, Stief CG, et al. Papillary vs clear cell renal cell carcinoma. Differentiation and grading by iodine concentration using DECT\u0026mdash;correlation with microvascular density. Eur Radiol. 2020 Jan;30(1):1\u0026ndash;10.\u003c/li\u003e\n\u003cli\u003eManoharan D, Netaji A, Diwan K, Sharma S. Normalized Dual-Energy Iodine Ratio Best Differentiates Renal Cell Carcinoma Subtypes Among Quantitative Imaging Biomarkers From Perfusion CT and Dual-Energy CT. American Journal of Roentgenology. 2020 Dec;215(6):1389\u0026ndash;97.\u003c/li\u003e\n\u003cli\u003eSah A, Gupta A, Garg S, Yadav N, Khan MA, Das CJ. Can quantitative perfusion CT-based biomarkers predict renal cell carcinoma subtypes?. Abdominal Radiology. 2024 Dec 17:1-9.\u003c/li\u003e\n\u003cli\u003eChauhan SS, Dixit R, Chowdhury V, Khurana N. Perfusion CT in the Evaluation of Renal Cell Carcinoma. JCDR.2022:13(1):2046-2056.\u003c/li\u003e\n\u003cli\u003eDeng Y, Soule E, Samuel A, Shah S, Cui E, Asare-Sawiri M, et al. CT texture analysis in the differentiation of major renal cell carcinoma subtypes and correlation with Fuhrman grade. Eur Radiol. 2019 Dec;29(12):6922\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eChen M, Yin F, Yu Y, Zhang H, Wen G. CT-based multi-phase Radiomic models for differentiating clear cell renal cell carcinoma. Cancer Imaging. 2021 Dec;21(1):42.\u003c/li\u003e\n\u003cli\u003eScrima AT, Lubner MG, Abel EJ, Havighurst TC, Shapiro DD, Huang W, et al. Texture analysis of small renal cell carcinomas at MDCT for predicting relevant histologic and protein biomarkers. Abdom Radiol. 2019 Jun;44(6):1999\u0026ndash;2008.\u003c/li\u003e\n\u003cli\u003eLubner MG, Stabo N, Abel EJ, Del Rio AM, Pickhardt PJ. CT Textural Analysis of Large Primary Renal Cell Carcinomas: Pretreatment Tumor Heterogeneity Correlates With Histologic Findings and Clinical Outcomes. American Journal of Roentgenology. 2016 Jul;207(1):96\u0026ndash;105.\u003c/li\u003e\n\u003cli\u003eGupta A, Garg S, Yadav N, Dhanakshirur RR, Jain K, Nayyar R, Kaushal S, Das CJ. Predicting Renal Cell Carcinoma Subtypes and Fuhrman Grading Using Multiphasic CT-Based Texture Analysis and Machine Learning Techniques. Indian Journal of Radiology and Imaging. 2024 Dec 11.\u003c/li\u003e\n\u003cli\u003eBudai BK, Stollmayer R, R\u0026oacute;nasz\u0026eacute;ki AD, K\u0026ouml;rmendy B, Zsombor Z, Palot\u0026aacute;s L, et al. Radiomics analysis of contrast-enhanced CT scans can distinguish between clear cell and non-clear cell renal cell carcinoma in different imaging protocols. Front Med. 2022 Oct 13;9:974485.\u003c/li\u003e\n\u003cli\u003eGupta A, Dhanakshirur RR, Jain K, Garg S, Yadav N, Seth A, Das CJ. Deep Learning for Detecting and Subtyping Renal Cell Carcinoma on Contrast-Enhanced CT Scans Using 2D Neural Network with Feature Consistency Techniques. Indian Journal of Radiology and Imaging. 2024 Dec 11.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1 : Demographic details of the patients\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"442\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 218px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCharacteristics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 224px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePatients , n= 66\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 218px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge in years, median ( range)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 224px;\"\u003e\n \u003cp\u003e\u0026nbsp;68 ( 43-79)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 218px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; -Male\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; -Female\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 224px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e40 ( 60%)\u003c/p\u003e\n \u003cp\u003e26 (40%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 218px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eClear cell RCC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 224px;\"\u003e\n \u003cp\u003e52 (78.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 218px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNon clear cell RCC\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Papillary\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Chromophobe\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 224px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e10 (15.1%)\u003c/p\u003e\n \u003cp\u003e4 (6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2: Inter-reader variability of various PCT and DECT parameters\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"\" width=\"565\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 184px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eParameters\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePearson Correlation\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(r)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpearman Correlation (\u0026rho;)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChi Square Test\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eP-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerfusion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBV\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.935\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.946\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e370.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e\u0026lt;0.0005\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBF\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.837\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.922\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e721.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e\u0026lt;0.0005\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMTT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.422\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e88.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e0.629\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTTP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.311\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.373\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e73.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e0.937\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDECT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.894\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.894\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e7.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e\u0026lt;0.0005\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.889\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.862\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e173.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 105px;\"\u003e\n \u003cp\u003e\u0026lt;0.0005\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 3 Comparison of features between clear cell renal cell carcinoma (ccRCC) and non-clear cell renal cell carcinoma (non-ccRCC) across the three imaging modalities:\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"\" width=\"601\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eStudy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFeature\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"bottom\" style=\"width: 137px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eccRCC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"bottom\" style=\"width: 131px;\"\u003e\n \u003cp\u003e\u003cstrong\u003enon-ccRCC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1 Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eStd. dev.\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eStd. dev.\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"6\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCT\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTexture Analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 66)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eMean\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e167.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e24.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e142.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e27.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.018\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.766\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eVariance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e1440.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e663.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e505.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e235.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.710\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eEnergy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e1.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e1.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e1.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.972\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.169\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eEntropy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e7.04\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.298\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e6.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.295\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.842\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eSmoothness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.986\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e0.982\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.482\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.597\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eML Model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.937\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.165\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e0.417\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.243\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.934\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"6\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCT\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePerfusion\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 66)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eBF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e883\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e780\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e532.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e430.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.04\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.766\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eBV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e118\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e85.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e70.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e60.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.728\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eMTT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e11.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e3.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e12.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e4.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.548\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eTTP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e11.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e2.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e13.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e3.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.643\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eMIP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e158\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e38.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e33.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.848\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eML Model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.910\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDECT\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 66)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eIR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e65.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e23.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e35.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e17.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.910\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eIC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e2.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e1.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.894\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 104px;\"\u003e\n \u003cp\u003eML Model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.935\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOverall ML Model\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 61px;\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 70px;\"\u003e\n \u003cp\u003e0.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 75px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.954\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5952087/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5952087/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo evaluate and compare the diagnostic performance of CT texture analysis (CTTA), perfusion CT (PCT), and dual-energy CT (DECT) in distinguishing between clear-cell renal cell carcinoma (ccRCC) and non-ccRCC.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis retrospective study included 66 patients with RCC (52 ccRCC and 14 non-ccRCC) who underwent DECT and PCT imaging before surgery between January 2017 and December 2022. Two independent radiologists measured DECT parameters (iodine concentration and iodine ratio) and PCT parameters (blood flow, blood volume, mean transit time, and time to peak) using circular ROIs placed on tumors. For CTTA, the largest tumor cross-section in the corticomedullary phase was manually annotated using the \"labelme\" tool, and texture features were extracted with Python libraries including \"scipy\" and \"numpy.\" Multivariate logistic regression analysis was performed to assess the ability of PCT, DECT, and CTTA models to predict tumor subtypes.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAll three imaging modalities demonstrated high diagnostic accuracy, with F1 scores of 0.9107, 0.9358, and 0.9348 for PCT, DECT, and CTTA, respectively. Inter-reader agreement for PCT and DECT parameters was excellent (Pearson correlation\u0026thinsp;\u0026gt;\u0026thinsp;0.85). None of the three models were significantly different (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). While each modality could effectively differentiate between ccRCC and non-ccRCC, higher iodine ratio (IR) on DECT and increased entropy on CTTA were independent predictors of ccRCC, with F1 scores of 0.9345 and 0.9272, respectively (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The combined ML model integrating DECT, PCT, and CTTA parameters yielded the highest diagnostic accuracy, with an F1 score of 0.954.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe diagnostic accuracy of PCT, DECT, and CTTA in distinguishing between ccRCC and non-ccRCC tumors was equivalent and high. However, among these three methods, only IR on DECT and entropy on CTTA were identified as independent predictors of the RCC subtype; hence, these two quantitative markers may be more applicable in clinical practice.\u003c/p\u003e\u003ch2\u003eClinical relevance:\u003c/h2\u003e \u003cp\u003eAccurate, non-invasive biomarkers are essential to differentiate RCC subtypes, aiding in prognosis and guiding targeted therapies, particularly in ccRCC, where treatment options differ significantly.\u003c/p\u003e","manuscriptTitle":"Quantitative CT-based biomarkers for predicting Renal cell carcinoma subtypes: a comparison of Dual-Energy CT, Perfusion CT, and CT texture parameters","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-12 12:24:19","doi":"10.21203/rs.3.rs-5952087/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3dd26fb1-739b-468a-a0a4-c9575fcd0d29","owner":[],"postedDate":"February 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-02-12T12:24:26+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-12 12:24:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5952087","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5952087","identity":"rs-5952087","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}