Abstract:Abstract Background and Aims Vascular access dysfunction is one of the leading causes of morbidity and a major contributor to healthcare costs in hemodialysis (HD) patients. Inexpensive, non-invasive tools for routine assessment of vascular access function are needed. Hemodynamically relevant stenoses in arteriovenous fistulas (AVF) lead to a reduction in access flow rate (Qa) and changes in blood flow patterns in the AVF that may be picked up by palpation and auscultation. We hypothesized that these changes in blood flow patterns can not only be felt and heard but also seen, i.e., that they may be detectable in video recordings done with commercially available smartphones after digital motion augmentation. Methods We studied HD patients with AVF dysfunction requiring balloon angioplasty and/or stenting. One-minute video recordings of the skin above the AVF and Qa measurements were conducted before and after the endovascular intervention. Videos were recorded with an iPhone 6S (Apple Inc., Cupertino, CA, USA). Qa was measured by HVT100 endovascular flowmeter (Transonic Systems Inc., Ithaca, NY, USA). Significant access stenosis was defined as a >50% reduction of luminal diameter. Degree of stenosis was assessed by angiography. Frame-to-frame pixel changes in video images were amplified using “Eulerian Video Magnification” (Massachusetts Institute of Technology, MA, USA; http://people.csail.mit.edu/mrub/evm/#code). The time-domain data were then transformed into the frequency-domain signals. Fifty random 10-second segments were sampled per one-minute video, and the frequency with the lowest magnitude (Fmin) was determined in each sample (example shown in Fig. 1). The average Fmin was then assessed for its association with the degree of AVF stenosis. Results Ninety subjects were studied (Table 1). AVF interventions were successful in all patients. Post-intervention Qa (1638 ± 714 ml/min) was on average 1.23-fold higher than pre-intervention Qa (1373 ± 684 ml/min; P<0.01, paired t-test). Subjects were grouped by degree of stenosis, and the number of subjects in each category is shown in Fig. 1B. Higher degrees of stenosis were associated with greater increases in Qa from before to after the intervention (P<0.01, one-way ANOVA; Fig. 1C). Interestingly, the degree of AVF stenosis was also positively related with the change in Fmin from before to after the intervention (P=0.08, one-way ANOVA; Fig. 1D). Conclusion Simple smartphone video recordings of AVF appear to contain frequency-domain information that correlates with hemodynamic changes caused by AVF stenoses. While the Fmin metric employed in our analysis is not ideal, these results should encourage the quest for other parameters that exhibit higher correlations with vascular access dysfunction. If successful, this would allow commercially-available smartphones to be used as ubiquitous tools for quick, non-invasive, ambulatory surveillance of AVF function, thereby allowing timely referrals and avoidance of emergency interventions.

Quality Evaluation via PPG on the AVFs of Hemodialysis Patients Based on Both Blood Flow Volume and Degree of Stenosis

Machine Learning Classification for Assessing the Degree of Stenosis and Blood Flow Volume at Arteriovenous Fistulas of Hemodialysis Patients Using a New Photoplethysmography Sensor Device

Achieving High Accuracy in Predicting Blood Flow Volume at the Arteriovenous Fistulas of Hemodialysis Patients by Intelligent Quality Assessment on PPGs

Arteriovenous Fistula Flow Dysfunction Surveillance: Early Detection Using Pulse Radar Sensor and Machine Learning Classification

The feasibility study of photoplethysmography features for arteriovenous fistula stenosis detection in hemodialysis patients with statistical approach

The prototype device for non-invasive diagnosis of arteriovenous fistula condition using machine learning methods

Dysfunction Screening in Experimental Arteriovenous Grafts for Hemodialysis Using Fractional-Order Extractor and Color Relation Analysis

Classification of Photoplethysmographic Signal Quality with Deep Convolution Neural Networks for Accurate Measurement of Cardiac Stroke Volume

Photoplethysmogram Signal Quality Assessment Using Support Vector Machine and Multi-Feature Fusion

DeepVAQ : an adaptive deep learning for prediction of vascular access quality in hemodialysis patients

Improving accuracy of vascular access quality classification in hemodialysis patients using deep learning with K highest score feature selection

Acoustic Detection of ArterioVenous Access Stenosis Based on MUSIC Power Spectral Features

Quantification and Visualization of Reliable Hemodynamics Evaluation Based on Non-Contact Arteriovenous Fistula Measurement

Deep learning analysis of blood flow sounds to detect arteriovenous fistula stenosis

P1320ASSESSMENT OF ARTERIOVENOUS FISTULA DYSFUNCTION WITH ACCESS STENOSIS IN HEMODIALYSIS PATIENTS USING SMART-PHONE VIDEOS

Application of Color Doppler Ultrasound for Arteriovenous Fistula in Hemodialysis

The clinical application of ultrasound dilution technology for vascular access flow detection among maintenance hemodialysis patients

Arteriovenous shunt stenosis assessment based on empirical mode decomposition and 1D-convolutional neural network: Clinical trial stage

The application of blood flow sound contrastive learning to predict arteriovenous graft stenosis of patients with hemodialysis

Arteriovenous Fistula Stenosis Classification Method Based on Auxiliary Wave and Transformer

Quality Assessment of Photoplethysmography Signals For Cardiovascular Biomarkers Monitoring Using Wearable Devices