Presented by: Jan Svihra Jr. MD,PhD
Jessenius faculty of medicine of Comenius University

Introduction

Although several reliable algorithms predicting a shock wave lithotripsy (SWL) success rate exist, a novel aproach of machine learning was seldom used. In this study we developed a clinical nomogram to predict success rate of an SWL of kidney stones using machine learning models.

Materials

A retrospective analysis of SWL in kidney stone patients was performed. Demographic data, stone size, stone area, stone location inside the kidney, stone density, skin to stone distance (SSD), stent presence, hydronephrosis presence and complication rate were analysed. Success of a SWL was defined as a stone-free rate (SFR) with stone fragments smaller than 4 mm. Programming language Python, namely the scikit-learn library, was used for machine learning purposes. The study was approved by the local ethics committee.

Results

,

102 patients (68.6 % males) with a mean stone size of 9.1±3.1 mm and SSD 8.4±1.8 cm were analyzed. Using different machine learning models we have defined a probability of an SWL success. The best probability was found with SSD ≤ 8 cm and stone area ≤ 60 mm² in all locations with exception of a lower kidney pole (Fig. 1).

Conclusion

We have identified parameters to create a machine learning nomogram that predicts success rate of an SWL in patients with kidney stones.

Funding

None

Co-Authors

Peter Svihra, Ing,PhD
CERN

Lukas Bris, MD
University hospital Martin

Igor Sopilko, MD,PhD
University hospital Martin

Jan Luptak, MD,PhD
Jessenius faculty of medicine of Comenius University