A warning system for urolithiasis via retrograde intrarenal surgery using machine learning: an experimental study.

Area of Science:

• Biomedical Engineering
• Surgical Technology
• Urology

Background:

• Retrograde intrarenal surgery (RIRS) for urolithiasis carries a risk of kidney and ureter damage from laser exposure.
• A hypothesis suggests varying shockwave intensities are delivered based on laser interaction with stone versus tissue.

Purpose of the Study:

• To develop a real-time warning system to prevent or minimize laser-induced tissue damage during RIRS.
• To differentiate between laser-stone and laser-tissue interactions during surgery.

Main Methods:

• Simulated a surgical environment using a raw chicken model with artificial stones.
• Developed an acceleration measurement system to record power signal data.
• Utilized Fast Fourier Transform (FFT) for signal analysis and Discrete Wavelet Transform (DWT) for feature extraction.
• Employed a random forest classification algorithm to identify laser-tissue interface states.

Main Results:

• FFT analysis revealed distinct spectral differences below 2500 Hz for idle, stone, and tissue states.
• The DWT and random forest model achieved a maximum average test accuracy of over 95%.
• Testing with dummy data confirmed the system's accuracy and lack of bias (33.33% accuracy).

Conclusions:

• The developed monitoring system accurately identifies laser exposure states within 0.5 seconds.
• This system can generate laser irradiance status, potentially minimizing surgeon error in RIRS procedures.
• The technology shows promise for enhancing patient safety during urolithiasis treatment.