Kidney Week

Abstract: SA-PO0077

Contact-Free Arteriovenous Access Monitoring in Patients on Hemodialysis Using Optical Imaging and Audio Analysis: Multicenter CMAP Interim Results

Session Information

• Artificial Intelligence: Kidney Imaging, Pathology, and Precision Phenotyping
  October 24, 2026 | Location: Exhibit Hall A, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Artificial Intelligence, Digital Health, and Data Science

• 300 Artificial Intelligence, Digital Health, and Data Science

Authors

• Kumbar, Lalathaksha Murthy, Henry Ford Hospital, Detroit, Michigan, United States

Lalathaksha Murthy Kumbar, MD, MBA, FASN

• Drori, Hagay, PatenSee, Tel Aviv-Yafo, Tel Aviv District, Israel

Hagay Drori

• Cohen-Hagai, Keren, Meir Medical Center Department of Nephrology, Kefar Sava, Center District, Israel

Keren Cohen-Hagai, MD

• Kenis, Irina, Tel Aviv University Faculty of Medical and Health Sciences, Tel Aviv-Yafo, Tel Aviv District, Israel

Irina Kenis, MD

• Sohaney, Ryann, Henry Ford Hospital, Detroit, Michigan, United States

Ryann Sohaney, DO, MS

• Soi, Vivek, Henry Ford Hospital, Detroit, Michigan, United States

Vivek Soi, MD

• Soman, Sandeep S., Henry Ford Hospital, Detroit, Michigan, United States

Sandeep S. Soman, MD, FASN

Background

Arteriovenous (AV) access monitoring based on physical examination is subjective, operator-dependent, and lacks longitudinal validity for tracking or prognostication. PatenSee is a non-contact, clinician-free optical system that acquires "look, listen, and feel" parameters for standardized AV access assessment.

Methods

We conducted a prospective, multicenter, multinational, longitudinal study in patients on chronic hemodialysis with AV fistulas or grafts. PatenSee recorded non-contact optical vibrometry signals over the access prior to needle cannulation. Audio ("listen") data were compared with digital stethoscope recordings using band-limited (100–800 Hz) spectral cosine similarity. Each access had ≥6 sessions; the primary analysis used three recordings from sessions 4–6. A session was considered successful if the cosine similarity between PatenSee and the stethoscope met a prespecified agreement threshold of ≥0.75. Exploratory endpoints included feasibility metrics (usable signal rate, scan duration) and longitudinal vibration analysis to track hemodynamic changes associated with access dysfunction.

Results

As of April 2026, 859 PatenSee sessions were acquired in 60 AV accesses over up to 10 months (median 13 sessions per access; range 6–24). Automated quality control excluded 12.8% of recordings, leaving 749 measurements from 54 accesses. For the primary endpoint set (three sessions per patient; n=162 sessions), the mean cosine similarity between PatenSee audio and digital stethoscope recordings was 0.90 (95% CI 0.88–0.91), with a median of 0.90 (95% CI 0.89–0.92) (1 indicates complete similarity and 0 indicates no similarity), surpassing the prespecified feasibility threshold. Analysis of all successful sessions showed similar performance (mean 0.90, median 0.91). The average time required for PatenSee data capture was 61 seconds (95% CI 59–62).

Conclusion

In this prospective multicenter interim analysis, non-contact optical monitoring with PatenSee achieved high agreement with digital stethoscope recordings and met the prespecified feasibility target, while providing rapid, standardized AV access assessment fully embedded in the routine dialysis center workflow. These findings support the potential for scalable, clinician-free AV access monitoring.

Funding

• Commercial Support – PatenSee Inc