Kalon Morgan: My name is Kalon Morgan. I'm an incoming fourth-year medical student at Rocky Vista University. I'm also a Leadership and Innovation Fellowship Training Scholar with the Department of Urology at University of California, Irvine. I'm presenting a project titled Automated CT-Based Stone Volume Determination: An Artificial Intelligence Algorithm to Calculate Kidney Stone Volume.

From 219 CT scans, we identified and manually segmented a total of 771 stones. Stone volumes were calculated using 3D software. For comparison, stone volumes were also calculated using the European Association of Urology ellipsoid formula. Based on manually segmented stone volumes, a Convolutional Neural Network Algorithm was trained and programmed to segment stones and to calculate their volume.

Pearson's correlation coefficients were calculated. Dice score, a measure of 3D fit, was calculated comparing the CNN generated volumes and manually segmented volumes. Compared to manually-calculated stone segmentations, the CNN had a Pearson's correlation coefficient of 0.82, versus the European Association ellipsoid formula, which had a Pearson's correlation coefficient of 0.75. The CNN algorithm had a dice score of 0.56.

So, in conclusion, a highly trained, AI deep learning algorithm accurately segmented renal stone volumes, and calculated stone volumes. The algorithm outperformed an ellipsoid formula for determining stone volume. Thank you.