Machine learning‑based classification of renal cell carcinoma and upper tract urothelial carcinoma using preoperative hematological markers

Renal tumors are the second most common type of urological cancer. These tumors consist of renal cell carcinoma (RCC; 80‑85%) and upper tract urothelial carcinoma (UTUC; 5‑10%). Accurate preoperative differentiation between these two entities remains challenging despite advances in imaging and endoscopic techniques, often leading to potential misdiagnoses and inappropriate treatment strategies. Therefore, the development of novel diagnostic methods to improve preoperative diagnosis remains an area of notable clinical interest. At present, the potential for using machine learning (ML) based on preoperative hematological markers as non‑invasive tools for cancer diagnosis and prognosis is being investigated. However, the application of ML based on preoperative hematological markers to differentiate RCC from UTUC remains unclear. The present study aimed to develop a classification system to differentiate between RCC and UTUC using ML based on preoperative hematological markers and clinical data from 172 patients with renal cancer, including 115 cases (66.9%) of RCC and 57 cases (33.1%) of UTUC. To address class imbalance, the Synthetic Minority Oversampling Technique was applied within a leakage‑free pipeline restricted to the training data. Feature selection was performed using recursive feature elimination with cross‑validation. A panel of ML algorithms was screened. The six best‑performing models were tuned and evaluated, with performance assessed by accuracy, balanced accuracy, precision, recall, F1 score and the area under the receiver operating characteristic curve (AUC). The Extra Trees classifier showed the most favorable overall performance, with an accuracy of 75.0%, a balanced accuracy of 0.71 and an AUC value of 0.755 (bootstrap AUC, 0.71; 95% CI, 0.59‑0.83). These preliminary findings suggest that ML models based on routinely available preoperative hematological markers may serve as a useful, low‑cost adjunct to support the differentiation of RCC and UTUC.