Prediction of Fuhrman nuclear grade for clear cell renal carcinoma by a multi-information fusion model that incorporates CT-based features of tumor and serum tumor associated material

Machine Learning-Enabled Fuhrman Grade in Clear-cell Renal Carcinoma Prediction Using Two-dimensional Ultrasound Images

OBJECTIVE

Accurate assessment of Fuhrman grade is crucial for optimal clinical management and personalized treatment strategies in patients with clear cell renal cell carcinoma (CCRCC). In this study, we developed a predictive model using ultrasound (US) images to accurately predict the Fuhrman grade.

METHODS

Between March 2013 and July 2023, a retrospective analysis was conducted on the US imaging and clinical data of 235 patients with pathologically confirmed CCRCC, including 67 with Fuhrman grades Ⅲ and Ⅳ. This study included 201 patients from Hospital A who were divided into training set (n = 161) and an internal validation set (n = 40) in an 8:2 ratio. Additionally, 34 patients from Hospital B were included for external validation. US images were delineated using ITK software, and radiomics features were extracted using PyRadiomics software. Subsequently, separate models for clinical factors, radiomics features, and their combinations were constructed. The model's performance was assessed by calculating the area under the receiver operating characteristic curve (AUC), calibration curve and decision curve analysis (DCA).

RESULTS

In total, 235 patients diagnosed with CCRCC, comprising 168 low-grade and 67 high-grade tumors, were included in this study. A comparison of the predictive performances of different models revealed that the logistic regression model exhibited relatively good stability and robustness. The AUC of the combined model for the training, internal validation and external validation sets were 0.871, 0.785 and 0.826, respectively, which were higher than those of the clinical and imaging histology models. Furthermore, the calibration curve demonstrated excellent concordance between the predicted Fuhrman grade probability of CCRCC using the combined model and the observed values in both the training and validation sets. Additionally, within the threshold range of 0-0.93, the combined model demonstrated substantial clinical utility, as evidenced by DCA.

CONCLUSION

The application of US radiomics techniques enabled objective prediction of Fuhrman grading in patients with CCRCC. Nevertheless, certain clinical indicators remain indispensable, underscoring the pressing need for their integrated use in clinical practice.

ADVANCES IN KNOWLEDGE

Previous studies have predominantly focused on using computed tomography or magnetic resonance imaging modalities to predict the Fuhrman grade of CCRCC. Our findings demonstrate that a prediction model based on US images is more cost-effective, easily accessible and exhibits commendable performance. Consequently, this study offers a promising approach to maximizing the use of US examinations in future research.

Deep learning using contrast-enhanced ultrasound images to predict the nuclear grade of clear cell renal cell carcinoma

PURPOSE

To assess the effectiveness of a deep learning model using contrastenhanced ultrasound (CEUS) images in distinguishing between low-grade (grade I and II) and high-grade (grade III and IV) clear cell renal cell carcinoma (ccRCC).

METHODS

A retrospective study was conducted using CEUS images of 177 Fuhrmangraded ccRCCs (93 low-grade and 84 high-grade) from May 2017 to December 2020. A total of 6412 CEUS images were captured from the videos and normalized for subsequent analysis. A deep learning model using the RepVGG architecture was proposed to differentiate between low-grade and high-grade ccRCC. The model's performance was evaluated based on sensitivity, specificity, positive predictive value, negative predictive value and area under the receiver operating characteristic curve (AUC). Class activation mapping (CAM) was used to visualize the specific areas that contribute to the model's predictions.

RESULTS

For discriminating high-grade ccRCC from low-grade, the deep learning model achieved a sensitivity of 74.8%, specificity of 79.1%, accuracy of 77.0%, and an AUC of 0.852 in the test set.

CONCLUSION

The deep learning model based on CEUS images can accurately differentiate between low-grade and high-grade ccRCC in a non-invasive manner.

Association between CT-based adipose variables, preoperative blood biochemical indicators and pathological T stage of clear cell renal cell carcinoma

Background

Clear cell renal cell carcinoma (ccRCC) is corelated with tumor-associated material (TAM), coagulation system and adipocyte tissue, but the relationships between them have been inconsistent. Our study aimed to explore the cut-off intervals of variables that are non-linearly related to ccRCC pathological T stage for providing clues to understand these discrepancies, and to effectively preoperative risk stratification.

Methods

This retrospective analysis included 218 ccRCC patients with a clear pathological T stage between January 1st, 2014, and November 30th, 2021. The patients were categorized into two cohorts based on their pathological T stage: low T stage (T1 and T2) and high T stage (T3 and T4). Abdominal and perirenal fat variables were measured based on preoperative CT images. Blood biochemical indexes from the last time before surgery were also collected. The generalized sum model was used to identify cut-off intervals for nonlinear variables.

Results

In specific intervals, fibrinogen levels (FIB) (2.63-4.06 g/L) and platelet (PLT) counts (>200.34 × 109/L) were significantly positively correlated with T stage, while PLT counts (<200.34 × 109/L) were significantly negatively correlated with T stage. Additionally, tumor-associated material exhibited varying degrees of positive correlation with T stage at different cut-off intervals (cut-off value: 90.556 U/mL).

Conclusion

Preoperative PLT, FIB and TAM are nonlinearly related to pathological T stage. This study is the first to provide specific cut-off intervals for preoperative variables that are nonlinearly related to ccRCC T stage. These intervals can aid in the risk stratification of ccRCC patients before surgery, allowing for developing a more personalized treatment planning.