Blood flow quantification in dialysis access using digital subtraction angiography: A retrospective study

Highlights

• Dialysis access blood flow can be quantified using contrast-injected fistulagrams obtained during vascular access interventions.

• The integration of blood flow measurement function in the current angiographic system permits the endpoint determination of vascular access interventions.

• The cross-correlation algorithm with gamma variate smoothing function has the potential for precise blood flow computation (r = 0.92).

Abstract

Background and objective

Vascular access is the “lifeline” of end-stage renal disease patients, which is surgically constructed to remove blood-waste and return artificially filtered blood into circulation. The arteriovenous shunting causes an abrupt change in blood flow and results in increased fluidic stress, which predisposes to access stenosis and thrombosis. While access flow is crucial to evaluate interventional endpoint, application to measure flow using digital angiogram is not yet available. The goal of this study was to determine the feasibility of flow quantification in dialysis access using a software tool and to guide the design of an imaging protocol.

Methods

173 digital subtraction angiographic (DSA) images were retrospectively analyzed to evaluate access flow in a custom-programming environment. Four bolus transit time algorithms and a distance calculation method were assessed for flow computation. Gamma variate function was applied to remove secondary flow and intensity outliers in the bolus time-intensity curves and evaluated for enhancement in computational accuracy. The percent deviations of flow rates computed from dilution of iodinated radio-contrast material were compared with in situ catheter-based flow measurement.

Results

Among the implemented bolus transit time algorithms, quantification error (mean ± standard error) of cross-correlation algorithm without and with gamma variate curve fitting was 35 ± 1% and 22 ± 1%, respectively. All other algorithms had quantification error >27%. The bias and limits of agreement of the cross-correlation algorithm with gamma variate curve fit was −94 ml/min and [−353, 165] mL/min, respectively.

Conclusions

The cross-correlation algorithm with gamma variate curve fit had the best accuracy and reproducibility for image-based blood flow computation. To further enhance accuracy, images may need to be acquired with a dedicated injection protocol with predetermined parameters such as the duration, rate and mode of bolus injection, and the acquisition frame rate.

Introduction

The global burden of kidney disease has continued to rise over the years predisposing the population with increased risks of kidney failure [1], [2], [3]. Patients diagnosed with end-stage renal disease (ESRD) have both kidneys nonfunctional and, therefore, require a kidney transplant or dialysis to survive [4]. This is further exacerbated by the lack of available transplant organs and the inability to maintain a patent and complication-free vascular access for dialysis. Because of the arteriovenous shunting, there is an abrupt change in local hemodynamics that invokes a complex pathophysiological process of intimal remodeling (to undertake neo-fluidic stress) predisposing to access stenosis and thrombosis [5], [6], [7]. Because of the change in luminal diameter either from underlying pathology or corrective procedures, blood flow quantification (access flow) in dialysis grafts/fistulae is the most preferred and reliable method to evaluate access patency and assess the outcomes of various endovascular procedures [8,9]. In particular, access flow of at least 600 ml/min in arteriovenous (AV) prosthetic grafts or at least 400–500 ml/min in AV autogenous fistula is considered the bare minimum threshold for patent access [10]. These thresholds have been used to determine endpoints during dialysis access interventions.

Access-related complications due to thrombosis and stenosis are routinely corrected by thrombectomy, percutaneous transluminal angioplasty (PTA), or in combination with stents/stent-grafts [11]. Access flow is sensitive to the change in luminal radius resulting from vascular obstruction (Hagen–Poiseuille law: flow rate ∝ radius⁴) [12]. The restoration of the blocked lumen from PTA or thrombectomy results in an instant increase in blood flow. The final flow rate obtained after intervention allows radio-surgeons to predict the time interval the patient would most likely not need re-angioplasty. For instance, when restored flow rate in grafts is greater than 1000 mL/min, it requires fewer repeat interventions, has longer graft survival, and less likely to thrombose within the first six months of angioplasty [13]. Similarly, fistulae with postoperative AF above 300 mL/min are more likely to remain patent for over a year [14]. The inability to restore the desired flow after corrective angioplasty helps in timely referral and/or collaboration with vascular surgeons for potential access salvage interventions.

While digital angiography is used to visualize vasculature and perform access-related interventions, measurement of access flow based on the contrast-injected angiographic images is currently unavailable in the interventional radiology suites [15]. In the United States, catheter-based thermodilution technology (non-imaging) is primarily used to measure access flow by rapidly injecting saline bolus into the vascular access and detecting the change in blood-saline temperature downstream of the site of injection [16,17]. Though this method is quick and simple to operate, it incurs an additional cost to the examination and the accuracy can be influenced by the operator's experience and the rate of bolus injection. Duplex imaging can technically be used for blood flow measurement, however, the quality of imaging (low signal to noise ratio) or equipment, the requirement to maintain a certain angle of insonation (≤60°), and operator dependent judgment limits its use and reliable interpretation [18], [19], [20]. On the other hand, ultrasound is more often used to examine thrombosed or non-thrombosed access prior to vascular access intervention and to establish/ guide the initial entry point of medical devices into the vasculature.

Blood flow measurement based on digital imaging has been well investigated [15,[21], [22], [23], [24], [25], [26], [27]]. A review by Shpilfoygel et al. systematically outlines the various approaches available for blood flow computation [28]; however, the application to the dialysis realm is yet to be thoroughly explored. In particular, the prospect of access flow measurement discussing the optimum approach, accuracy, and limitations of various algorithms, and other technicalities associated with the imaging protocol and radiation hazards is due to attract research attention in this direction. It is hypothesized that the access flow can be mathematically computed from digital subtraction angiographic (DSA) images obtained during fistulography. The ability to measure access flow using the current angiographic system adds an additional value to the imaging armamentarium and enables the possibility to enhance flow quantification accuracy by optimizing the variables of acquisition. Previously, we had demonstrated the feasibility of access flow computation in a small set of DSA images obtained from clinical examinations [29]. However, the accuracy was inadequate and primarily affected by secondary flow and intensity fluctuations from incomplete contrast-blood mixing inherent in angiograms. In this study, image preprocessing and gamma variate curve fitting model has been studied to correct bolus time-intensity curves, and its effect on flow quantification accuracy has been evaluated in an expanded image dataset. The accuracy and functioning of the catheter-based flow meter has also been verified against an ultrasonic (US) flow module.