Abstract: PO1155
Simplifying the 28-Zone Lung Ultrasound Protocol
Session Information
- Hemodialysis and Frequent Dialysis - 3
October 22, 2020 | Location: On-Demand
Abstract Time: 10:00 AM - 12:00 PM
Category: Dialysis
- 701 Dialysis: Hemodialysis and Frequent Dialysis
Authors
- Sonko, Momodou Lamin, Penn Medicine, Philadelphia, Pennsylvania, United States
- Kuznetsov, Ivan, Penn Medicine, Philadelphia, Pennsylvania, United States
- Arnold, Thomas C., Penn Medicine, Philadelphia, Pennsylvania, United States
- Baston, Cameron, Penn Medicine, Philadelphia, Pennsylvania, United States
- Reisinger, Nathaniel C., Penn Medicine, Philadelphia, Pennsylvania, United States
Background
Lung ultrasound (LUS) using a 28-zone quantitative B-line score (BLS) is a reliable marker of fluid overload (FO) among patients with end-stage kidney disease (ESKD) on hemodialysis (HD), outperforming physical exam and correlating well with cardiovascular outcomes. A trial comparing BLS-guided dry weight probing to usual care showed improved blood pressure and echocardiographic parameters. However, 28-zone BLS study is criticized as impractical for clinical practice. Using a machine learning algorithm we determined whether accurate assessment of FO can be determined using just 4, 6, and 8 scanning zones.
Methods
We analyzed an existing dataset of 28-zone BLS scores obtained from 100 HD patients presenting to acute care at our center for a total of 2800 scored LUS clips. Using linear correlation and discriminant analysis, we fit models that allowed us to approximate the 28-zone BLS based on 4, 6, or 8 zone protocols. We next applied linear discriminant analysis to study whether we could predict FO severity (low: BLS <15, moderate: BLS 15 to 30; high BLS>30) based on the limited zones. Finally, we tested whether we could achieve better diagnostic performance with subsets of scan-zones that had not previously been reported. Final outcome measures were reported as correlation coefficients and Cohen’s kappa.
Results
We found that the BLS of the 4, 6, and 8-zone scan correlated strongly and linearly with the BLS of the full 28-zone scan with Pearson correlations of 0.95, 0.92, and 0.92, respectively. In determining FO severity based on the limited scanning zones, the model produced resultant Cohen’s Kappa values of 0.74, 0.76, and 0.71 for the 4, 6, and 8-zone scans, respectively. We identified an undescribed 4-zone scan that produced a Kappa of 0.82. We found that equal linear weighting of all zones gave the best accuracy.
Conclusion
We found that 4, 6, and 8 zone BLS scores perform similarly to the 28-zone BLS. We identified a subset of 4 zones that gave better accuracy than existing 4, 6, or 8-zone scans. These findings support that a limited number of scanning zones can be used to reliably determine FO. Further work is needed on a larger dataset to validate these findings and to explore the physiological mechanism to support the novel 4-zone scan.