Introduction: The prognosis of patients with upper tract urothelial cancer (UTUC) is generally poor. Although neoadjuvant treatment may improve the survival outcomes after surgery, further explorations are required to identify the patients at high risk for recurrence preoperatively. In this study, we investigate the utility of radiomics analysis using computed tomography (CT) for preoperative risk assessment in UTUC. Methods: We analyzed 55 patients with UTUC who underwent radical surgery (approval # M2019-192). All the patients were required to have dynamic CT before surgery, on which the tumor was identified and evaluable. Radiomics analysis was performed on the corticomedullary phase images using LIFEx software. The tumor contour was manually outlined on axial slices, and the volume of interest (VOI) of the tumor was set in each case. Based on tumor segmentation, 124 radiomic features were extracted from each VOI, and important features for predicting metastatic recurrence were identified using Boruta algorithm. Prognostic impact of the radiomic features was assessed using multivariable analysis incorporating preoperative clinical variables. Results: The tumor was located in renal pelvis and ureter in 32 (58%) and 23 (42%) patients, respectively. Clinical T stages were Ta-1/2/=3 in 30 (55%)/15 (27%)/10 (18%) patients, respectively. Hydronephrosis was observed in 15 (27%). On the surgical specimen, 18 cases (33%) had pathological =T3. During the median follow-up of 33 months, 8 patients (15%) experienced metastatic recurrence. The Boruta algorithm identified INTENSITY.BASED_90thPercentile.HU (IB90P) as an important feature for metastatic recurrence (figure). The receiver operating characteristic curve defined the cut-off value of IB90P as 64.5. In multivariable Cox-proportional hazard analysis, IB90P <64.5 (HR=9.9, p=0.002) and hydronephrosis (HR=7.1, p=0.02) were independent risk factors of metastatic recurrence. Two-year metastasis-free survival were 45% and 95% in the patients with IB90P <64.5 and =64.5, respectively (figure). Conclusions: Radiomics analysis using preoperative CT identified an independent risk factor for metastatic recurrence. It may help the preoperative risk stratification and decision making in the management of patients with UTUC. SOURCE OF Funding: None
