Predictive accuracy of computer-generated C-index nephrometry scores compared with human-generated scores in predicting oncologic and perioperative outcomes.

Abstract

623 Background: The Centrality index (C-index) score is a validated nephrometry scoring system that requires precise measurements and mathematical calculations of cross sectional imaging. Like other nephrometry scores, its implementation has been slowed by required time investment and interobserver variability. We sought to automate this score on preoperative computerized tomography scans by developing an artificial intelligence-generated C-index score. We then aimed to evaluate its ability to predict meaningful oncologic and perioperative outcomes as compared to human-generated C-index nephrometry scores. Methods: 300 patients with preoperative computerized tomography with early arterial contrast phase were identified from a cohort of 544 consecutive patients undergoing surgical extirpation for suspected renal cancer. A deep neural network approach was used to automatically segment kidneys and tumors, and then programed to generate the measurements and calculate C-index score. Human C-index scores were independently calculated by medical personnel blinded to AI-scores. AI- and Human score agreement was assessed using bivariate linear regression correlation and their predictive abilities for both oncologic and perioperative outcomes were assessed using logistic regression and compared with receiver operating characteristic (ROC) curve analyses with measurements of areas under the curve (AUC). Results: Median age was 60 years (IQE 51–68), and 40% were female. Median tumor size was 4.2 cm and 91.3% had malignant tumors. 27% were high stage, 37% high grade, and 63% underwent partial nephrectomy. There was significant agreement between Human scores and AI-scores on linear regression analysis (R2 = 0.738, p <0.0001). Both AI- and Human generated C-index scores similarly predicted meaningful oncologic outcomes, with lower levels of either C-index score associated with increased risk of malignant histology (H-score p = 0.018, AI score p =0.014) high-grade disease (both p <0.0001), and high stage disease (both p <0.0001). Lower levels of either AI or human generated C-index scores also predicted a radical nephrectomy rather than partial nephrectomy surgical approach (both p <0.0001). AUC measurements (Table) were similar but consistently superior for AI generated C-index scores. Conclusions: Fully automated AI-generated C-index scores are comparable to human-generated C-index scores and predict a wide variety of meaningful patient-centered outcomes. Once validated in additional populations, our results suggest that our AI generated C-index could be delivered automatically from a preoperative CT scan to a clinician and patient at the point of care to aid in decision making. [Table: see text]