Purpose: We performed this study as a comprehensive evaluation of variables reported to affect lower pole stone clearance after shock wave lithotripsy using artificial neural network analysis. Materials and Methods: The radiographic images and treatment records of 680 patients with lower pole renal calculi treated with primary shock wave lithotripsy using the Wolf Piezolith 2500 (Wolf, Knittlingen, Germany) lithotriptor were retrospectively evaluated by applying artificial neural network analysis. Successful stone clearance was defined as absent fragments of any size detected on plain x-ray with tomography and/or excretory pyelography performed 6 months after treatment. Prognostic variables included patient characteristics, laboratory values, stone characteristics and the spatial anatomy of the lower pole, as defined by infundibular length, diameter, caliceal pelvic height, 2 measurements of the lower infundibulopelvic and infundibuloureteropelvic angles as well as the pattern of dynamic urinary transport. Results: Artificial neural network analysis had 92% accuracy for correctly predicting lower pole stone clearance. The pattern of dynamic urinary transport represented the most influential predictor of stone clearance, followed by a measure of the infundibuloureteropelvic angle, body mass index, caliceal pelvic height and stone size. Anatomical measurements of lower pole anatomy and classification of the type of urinary transport were well reproducible with low intra-observer and interobserver variability (correlation coefficient α >0.8). Conclusions: In a comprehensive analysis of variables reported to influence lower pole stone clearance artificial neural network analysis predicted stone clearance with a high degree of accuracy. The relative importance of dynamic urinary transport in lower pole stones and the usefulness of artificial neural network analysis to predict shock wave lithotripsy outcomes in individuals must be confirmed in a prospective trial.
- Kidney calculi
- Neural networks (computer)