AI-Driven Correction of Artifacts in Intrarenal Pressure Monitoring with LithoVue™ Elite

Overview

This study quantified artifact prevalence in intrarenal pressure (IRP) measurements during retrograde intrarenal surgery (RIRS) using the LithoVue™ Elite ureteroscope and developed an AI-based model to detect and correct these artifacts. The AI system demonstrated robust performance in identifying and correcting artifactual spikes, improving the accuracy of IRP monitoring critical for patient safety.

Background

Retrograde intrarenal surgery (RIRS) is a key minimally invasive treatment for urolithiasis but carries a risk of postoperative infectious complications, including sepsis in about 5% of cases. Elevated intrarenal pressure (IRP) during RIRS is a modifiable risk factor linked to these complications. The LithoVue™ Elite ureteroscope enables real-time IRP monitoring; however, measurements are frequently distorted by artifacts caused by sensor contact with renal or ureteral walls. Accurate IRP monitoring is essential to balance irrigation, suction, and temperature during surgery and to minimize complications.

Data Highlights

Key Findings

• Artifacts in IRP data are frequent and primarily caused by physical contact of the pressure sensor with pelvicalyceal or ureteral walls.
• Manual labeling of artifacts was performed using synchronized endoscopic video and IRP waveform data by expert urologists.
• The AI-based model was trained on 20 cases and externally validated on 7 independent cases, demonstrating robust artifact detection and correction.
• Thirty-two features capturing waveform morphology, temporal dynamics, and frequency characteristics were critical for model performance.
• Use of synthetic minority oversampling (SMOTE) and additional oversampling of high-pressure artifact segments addressed class imbalance effectively.
• Post-processing techniques enhanced temporal consistency and reduced false positive artifact predictions.

Clinical Implications

The AI-driven artifact correction system improves the reliability of real-time IRP monitoring during RIRS, enabling better intraoperative decision-making to maintain IRP below the safe threshold of 30 mmHg. This can potentially reduce the risk of postoperative infectious complications such as sepsis. Integration of this system into clinical workflows may streamline IRP data interpretation and enhance patient safety.

Conclusion

This study successfully developed and validated an AI-based model that detects and corrects artifacts in IRP measurements obtained via the LithoVue™ Elite ureteroscope, enhancing the accuracy of intrarenal pressure monitoring during RIRS. This advancement supports safer surgical practice by enabling precise IRP control.

References

1. EAU Endourology and AUSET Collaborative Study 2024 -- Development of an AI-driven system for correcting artifacts in intrarenal pressure measurements utilizing the LithoVue™ Elite ureteroscope