Presented by: Connie Chen BS
Indiana University School of Medicine

Introduction

Surgical decision making for kidney stones is often performed by visual estimation of stone burden, but as the quantity and size of stones increases, estimation becomes challenging as there is no standardized method for obtaining this measurement. Advances in artificial intelligence (AI) have allowed for automated calculation of stone volume; however, no study has evaluated how AI models compare with standard estimates or if true stone volume can predict outcomes after mini-PCNL. We aimed to compare stone volume with other methods of measuring stone burden in their predictive value for outcomes after mini-PCNL.

Materials

Patients undergoing unilateral mini-PCNL at our institution between 2019 and 2021 were included in this retrospective analysis. Stone burden was assessed through three ways: cumulative stone diameter, maximum diameter of largest stone, and total stone volume measured on pre-operative CT scan using a web-based AI segmentation tool. Mini-PCNL was defined as access sheath ≤ 22 French in size. A patient was stone-free if there were no fragments > 3mm on post-operative CT scan. Primary outcome was stone-free status (SFS) at discharge. Secondary outcome was operative time. Pearson correlation was used to determine if stone volume correlates with cumulative stone diameter and/or maximum diameter. Logistic regression and linear regression were used to compare the predictive value for SFS and operative time.

Results

,

A total number of 57 CT scans were analyzed. Pearson correlation revealed that stone volume had a positive correlation with cumulative stone diameter (r = 0.99, p < 0.001) and maximum diameter (r = 0.97, p < 0.001). Logistic regression revealed that cumulative stone diameter (OR 0.966, 95% CI 0.909-1.028), maximum diameter (OR 1.014, 95% CI 0.887-1.158), and stone volume (OR 1.351, 95% CI 0.483-3.777) did not predict SFS. Linear regression revealed that cumulative stone diameter was a predictor of operative time (estimate 1.917, 95% CI 0.859-2.975, p < 0.001). Maximum diameter (estimate 1.479, 95% CI -0.810-3.768) and stone volume (estimate 14.365, 95% CI -0.054-28.784) did not predict operative time.

Conclusion

In a cohort of mini-PCNL patients, operative time was best predicted by cumulative stone diameter. Both total stone diameter and maximum stone diameter correlate very well with stone volume. Further refinement of AI software measuring large stone burden volume and operative measures is needed to determine predictive value of stone-free rate and operative time in more complex stone procedures.

Funding

None

Co-Authors

Austen Slade, MD, MBA
Indiana University School of Medicine

Elisa Sarmiento, MSPH
Indiana University School of Medicine

Thomas Max Shelton, MD
Indiana University School of Medicine

RJ Caras, DO
Indiana University School of Medicine

Marcelino Rivera, MD
Indiana University School of Medicine