Patent Application
20240277423
During catheter-based treatments, surgeons are hampered by the limited sensory awareness of the microcatheters as they move through small, fragile, mobile vessels. While navigating a long (more than 1 meter) catheter, surgeons experience diminished haptic feedback. To navigate catheters, surgeons currently rely on static images of vessels superimposed on live fluoroscopy that shows the radio-opaque catheter tip. These perception deficits increase cognitive burden and diminish patient outcomes.
To address the aforementioned needs, this disclosure provides a novel approach for indirect endovascular sensing that equips surgeons with two unprecedented capabilities: 1) estimate the distance between the catheter tip and a target (clot/aneurysm wall) and 2) evaluate the quality of the engagement of the catheter tip with a clot. The disclosed technology offers a low-cost rapidly deployable sensory solution, which is compatible with existing catheter technology and alleviates the above-described perception barriers in endovascular procedures such as ischemic stroke treatment and aneurysm embolization.
The present disclosure provides a new low-cost sensory solution that addresses needs in domains (e.g., neuroendovascular stroke treatment) where existing technologies such as ultrasound or integrated distal contact/force sensory solutions (e.g., strain gauges, fiber Bragg grating) stand in contrast to the need to achieve a design solution with a large working bore and a diameter smaller than 2 mm.
The disclosure describes several methods and embodiments for achieving sensory information for catheter-based interventions in general and specifically for endovascular ischemic stroke and aneurysm embolization procedures. The disclosure includes four key elements to its sensory function:
Potential areas of application include:
In one embodiment, the disclosure provides a method for determining the distance between a tip of a catheter and a target. The method includes applying flow and pressure excitation to a proximal end of the catheter, measuring a pressure change, with a pressure sensor in fluid communication with the catheter, at the proximal end of the catheter, the pressure change based on a cycle of applied proximal pressure excitation and advancement of the catheter toward the target, applying a machine learning model to the measured pressure change to determine the distance between the tip of the catheter and the target, and providing the distance between the tip of the catheter and the target to a user.
In another embodiment, the disclosure provides a method of classifying quality of engagement between a tip of a catheter and a target. The method includes applying flow and pressure excitation to a proximal end of the catheter, measuring a pressure signal, with a pressure sensor in fluid communication with the catheter, at the proximal end of the catheter, the pressure change based on a cycle of applied pressure excitation and advancement of the catheter toward the target, defining feature vectors for the pressure signal, computing changes in one of the feature vectors within two cycles of incremental axial motion of the catheter, applying a classification algorithm to the pressure signal to determine whether the tip of the catheter is in contact with the target, if the tip of the catheter is in contact with the target, applying vacuum excitation to promote catheter engagement with the target, and applying a regression method to predict quality of engagement of the tip of the catheter with the target.
Before any embodiments of the invention are explained in detail, it is to be understood that the invention is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the following drawings. The invention is capable of other embodiments and of being practiced or of being carried out in various ways.
Stroke affects almost 795,000 patients per year and results in the fifth highest number of deaths in the U.S. (140,000/year) according to the Centers for Disease Control and Prevention, of which 87% are ischemic in nature. About 10% of these strokes (80,000/year) are due to large vessel occlusions (LVO) (blockage of one of the major arteries of the brain).
In 2015, MT became the treatment standard for acute LVO patients after several clinical trials showed its benefits. Patients treated with MT achieved recanalization rates up to 88% and had reduced disability rates at 90 days. The advantage of MT over intravenous thrombolysis with r-tPA alone (previous standard of care) was strong enough to justify ceasing patient enrollment in some of these clinical trials due to ethical concerns for the patients in the control group.
During MT procedures, neurointerventionalists inject a contrast agent through a guide catheter to obtain a digital subtraction angiogram (DSA) of the target vasculature. These images serve as a roadmap to navigate specialized thrombectomy catheters over a microcatheter and microwire to the LVO.
The need for safe and fast treatment stands in direct contrast to the fact that surgeons currently rely on subjective measures (e.g., lack of blood return in the aspiration catheter) to infer engagement with the thrombus. These perception barriers contribute to a low success rate of 25% in achieving complete reperfusion from a single thrombectomy attempt-known as “first pass effect”-which is associated with significant improvement in clinical outcome. Each failed clot retrieval attempt reduces a patient's chances of recovery due to continued brain oxygen deprivation.
Patient outcomes following endovascular treatment for acute ischemic stroke depend heavily on the ischemic time (time from onset of blockage to recanalization) and on the extent of recanalization of the cerebral vasculature. In a recent study by Kunz et al., based on data from the HERMES meta-analysis, it was shown that each 10 minute delay in ischemic time correlated with 40 days of added patient disability and $10,000 of added economic burden.
“First pass effect” is the achievement of complete revascularization from a single thrombectomy device pass. It has been associated with reduced ischemic time, lower mortality, reduced disability rate, and fewer procedural adverse events. However, with current thrombectomy techniques, first pass recanalization is achieved in only 25% of cases. Clot retrieval is abandoned after several time-consuming failed attempts when further benefits of restoring blood flow are outweighed by the surgical risk and the cumulative ischemic burden already suffered by the patient.
First-pass failure is partly due to sensory deficiencies hampering the surgeons' perception. These perception barriers are: 1) lack of simultaneous visualization of clot location and catheter tip, and 2) lack of sensory feedback about the level of engagement between the catheter tip and the clot. Surgeons currently rely on subjective measures (e.g., lack of blood return in the aspiration catheter) to infer engagement with the thrombus. After aspiration for 2-5 minutes with the clot engaged, the surgeon removes the catheter. If unsuccessful, the process of catheter preparation, navigation, and suction must be repeated.
In the United States, about 6.5 million people, are thought to harbor an unruptured intracranial aneurysm. Of those, approximately 30,000 people annually suffer from aneurysm rupture and 15% die before reaching a hospital. Worldwide, there are 500,000 annual deaths from ruptured aneurysms, with 50% of the victims younger than 50.
Endovascular coil embolization of ruptured and unruptured aneurysms is a common treatment in which the goal is to exclude the aneurysm from circulation, often by filling it with coils. Even when performed by experts, intraprocedural aneurysmal rupture (IAR) is a common and feared complication of endovascular treatment. This occurs in 1-11% of coiling procedures due to perforation of the aneurysm by the microwire, microcatheter, or coil. IAR causes a 4-fold increased risk of death or disability, consistent with 39% morbidity and 33% mortality rates. Extensive expertise and high clinical volume in endovascular aneurysm treatment has not been shown to affect the rate of IAR. Risk of mortality from IAR ranges as high as 40%. IAR occurs most commonly during treatment of ruptured aneurysms, aneurysms in the anterior communicating artery (likely from vessel tortuosity), and small aneurysms. These aneurysms require gentle placement of a microcatheter within the aneurysm sac in a position that allows for deployment of the coils (see
During this process, the microcatheter is pushed over a microguidewire under live fluoroscopy, while using a static magnified roadmap of the intracranial vasculature to assist in navigation. This requires the patient to remain still and the operator to consider how the vasculature may shift as the catheters are manipulated within mobile vessels. With forward motion of the microcatheter, compressive elastic energy builds. As the microwire is removed, axial “spring-back” of the catheter tip can cause catheter malposition or aneurysm perforation.
Currently, perception of the microcatheter placement within the aneurysm is limited to the views captured by biplane angiography relative to the radio-opaque marker at the catheter tip. Resistance is inferred by observing the behavior of the wire or coil during its deployment and noting retraction of the microcatheter during forward pressure on the coil. When a balloon is simultaneously used to protect the parent vessel from which the aneurysm arises (balloon-assisted coil embolization), the operator's awareness of resistance is further limited because the microcatheter tip used to coil the aneurysm becomes fixed in place by the inflated balloon. The disclosure provides a sensory modality to enable the surgeon to measure the microcatheter's distance from the aneurysmal wall and ensure optimal positioning. This provides the interventionalist with an added sensory cue that would help avoid the often-fatal IAR.
Sensing of catheter tip forces has received extensive attention in the literature and point force sensing exists commercially for larger cardiac catheters. However, no tip contact or proximity sensing solutions currently can be used at the scale needed for intracranial procedures.
Ultrasonic sensing for flow velocimetry has been developed and explored for the past two decades. The presence of the ultrasonic piezoelectric elements at the catheter tip prevents miniaturization while retaining a hollow and flexible tip. Such catheters have been used for cardiac applications. Miniature IVUS ablation and imaging catheters (e.g., Boston Scientific's Ultra-ICE™) lack the necessary tip flexibility and a working bore.
The disclosure is focused on indirect measures for estimating target (e.g., a clot) engagement and proximity to the target (e.g., a clot or aneurysm wall. Within the area of fluid mechanics, Womersley's seminal work on pulsatile flow in a straight pipe has been followed by numerous works focused on pulsatile flow in soft arteries and works on pressure wave propagation in pipes and arteries. Also, many works focused on flow modeling within the cerebral network.
The flow and pressure (e.g., vacuum) excitation-based sensing technique disclosed herein presents a scenario that is different from either the natural or the engineered situations in previous publications. Unlike the natural arterial blood flow, in the present problem the pulsatile jet produced by the syringe through the catheter tip is reminiscent of a zero-net-mass-flux oscillatory jet, or synthetic jet, in many flow control applications. However, the present jet is confined by a flexible vessel (the artery) and is thus unlike most of the synthetic jets. The presence of the blood clot or the aneurysm adds complexities to the flow problem, which does not lend itself to a simple analytical solution. Although there have been many studies of the blood flow through an aneurysm, including fluid-structure interaction (FSI), they primarily focused on blood flow in the natural direction, rather than the opposite direction created by vacuum here. Therefore, disclosed herein is a computational fluid dynamics (CFD) model to solve the fluid-structure interaction caused by catheter excitation, and from the CFD results, machine learning is employed to generate convenient function expressions of the pressure change in relation to the catheter-clot/aneurysm distance that can later be used for the sensing purpose.
As noted above, there exist current limitations of the state-of-the-art in perception for stroke and aneurysm embolization: (1) Surgeons currently lack the ability to sense the position of a catheter tip relative to the flexible and delicate vasculature. This results in reliance on static overlay images of vascular anatomy which does not reflect the true state of the vasculature and contributes to a heavy cognitive burden and a high rate of aneurysm ruptures. (2) In stroke, surgeons must resort to subjective measures to guess when a clot is sufficiently engaged for retrieval. There have been no prior works on quantifying the quality of catheter tip engagement with a clot. This limitation contributes to a low first-pass effect and poor clinical outcomes. (3) None of the catheters or prior works address sensing at the scale needed for neurointervention while preserving a working bore and a very flexible tip as required for safe intracranial navigation.
The disclosure offers a transformative low-cost sensory technology that can help overcome much of the aforementioned scientific gaps and clinical needs. With the above-mentioned technological and scientific advancements falling short of providing sensory feedback, a new approach is needed. A pragmatic approach is disclosed to achieve situational awareness that can help neuroendovascular surgeons. Although previous works on pulsed flow provide a theoretical foundation to understanding the mechanics of the problem, they cannot be directly applied here since a pulsed flow in a very small volume containing the catheter and the portion of artery up to the clot and its connecting network is considered. Furthermore, most works focus on modeling the cardiac arterial network and there are limited works on modeling pulsed flow within the cerebral vascular network and even fewer works considering the presence of a catheter. Therefore, disclosed herein are CFD models that also consider the fluid-structure interaction and to augment these models by casting the problem of distance measurement to a target as an identification problem and by leveraging machine learning regression methods.
The experiments also tested feasibility in very soft latex tubing as mock vasculature representing the other end of the spectrum of vessel wall softness. A motion excitation was applied to the piston of the syringe 14 using a slider-crank mechanism or a linear actuator. The catheter tip was placed at several distances from the mock target 26 and a digital vacuum pressure sensor 32 was placed near the syringe 14 to measure the induced cyclic vacuum as a function of syringe stroke. Using a motion excitation of 0.5 mm at 2 Hz, 4 Hz, and 8 Hz, the results obtained are shown in
This approach was validated to create blood clots from swine blood as shown in
A digital subtraction angiogram and SmartMask of the phantom vasculature were obtained by using a contrast agent (e.g., Omni 200) under bi-plane fluoroscopy. Image segmentation algorithms were developed to segment and track a catheter. The segmentation and tracking were achieved at 20 Hz for the vasculature, and 10 Hz for the catheter. Higher segmentation rates (e.g., better than 60 Hz) are achievable when porting Matlab to C++.
The disclosure provides a method for reliably sensing the distance of a catheter tip from a target (e.g., clot/aneurysm wall) and for classifying the quality of engagement with a target (e.g., a clot). The modeling and experimental validation focuses on solving these two problems for targets positioned along the intracranial ICA and MCA as depicted in
The accuracy of target distance/aneurysm wall measurement was validated in ex-vivo and animal models and to establish the utility of such sensory information in a user study. Since the target locations and the vessel diameters vary between subjects, a three-phase approach was defined: Phase 1 (Nominal geometry): focus was on CFD modeling and calibration for a vessel geometry corresponding with the 95% confidence intervals of vessel diameters; Phase 2 (Sampled geometry library): focus was on CFD modeling and calibration based on a library of sample vessel geometries from CT scans; and Phase 3 (Novel geometry): focus was on validation in arbitrary vessel geometries that have not been included in the model calibration set. These phases were completed on ex-vivo models before the planned animal validation.
Vasculature and Clot Models for ex-vivo bench-top studies. In the preliminary stage of the ex-vivo investigation, straight tubing was used as the vascular models. This tubing ranged in hardness from Tygon PVC tubing (ID 3.175 mm) to rubber latex (ID 3.048 mm), and had geometric properties that match the diameter (2-5 mm) of large intracranial vessels (ICA, MCA) and the length from the femoral artery to the site of occlusion in the ICA (about 1 meter). These straight models were used initially to minimize the frictional effects that can mask the traction force experienced by the target.
During Phase 1 and Phase 2 of this study, these straight models were replaced with phantom models matching the geometry of the ICA from CT scans (as in
All of these vessel models were extended to include the aortic arch (and its branches) and the femoral artery. A commercial model (e.g., from Elastrat) may be well-suited for this application as well. It is a transparent soft silicone model that is continuous from femoral artery up to the intracerebral circulation, including thoracic and abdominal aorta, aortic arch, great vessels in the neck, and intracranial ACA and MCA with intact circle of Willis. This model also provides a closed system with a fluid pump to mimic vascular circulation and allows vascular catheter access.
A mock target (e.g., a clot) was lodged at the distal end of the mock vasculature to simulate vessel occlusion. The use of coagulated swine blood as a clot was previously investigated as shown in
CFD Models and Verification/Calibration of These Models. The computational fluid dynamics (CFD) models are a way to generate a large data set exceeding the 30 phantom models upon which experiments were conducted. These models can be run with different geometries and model parameters to generate synthetic data after proper model calibration in Phase 1 and 2. However, these models are not suitable for real-time applications, so we planned to proceed with an intelligent statistical regression model as shown in
Pressure transmission under pulsatile flow excitation in a catheter within a flexible blood vessel. During vacuum excitation, the flow path included the syringe, the interior of the catheter, and the region between the catheter tip and the target, and the gap between the catheter and the vessel wall (
Since the Reynolds number of the flow is relatively small (100 or less), turbulence was not expected. However, vortices near the catheter tip due to jet pulsatility affect the local pressure around the tip. A time-dependent FSI model was developed for the pulsatile jet that included the jet flow, wall compliance of the elastic vessel, and the axial oscillation of the catheter tip. The wall compliance allows the vessel wall to deform inward under the vacuum pressure, thus dynamically changing the fluid volume and forming a flexible reservoir. The catheter also experienced an extension/compression oscillation of around 0.5 mm at the tip due to the shear stresses exerted by the flow on the tube's interior and exterior surfaces (
Model validation. To validate the CFD model, a pressure transducer needle (e.g., From Galetech) will be inserted into the vessel near the catheter tip to monitor the instantaneous pressure P2. For the validation purpose, we will compare the pressure at the base P0, pressure at the tip P2, and the tip oscillation obtained from the numerical simulation with those obtained from the in vitro experiment. The comparison will be done for a large tip-clot distance (˜20 mm), a medium distance (3 to 5 mm), and a close distance (˜1 mm).
Parameter study with the CFD model. After model validation, a systematic study of the governing parameters will be performed, including geometrical parameters such as the catheter diameter d, vessel diameter dc, tip-clot distance, Lc, material parameters such as the dynamic elasticity of the anisotropic vessel wall, and excitation parameters such as the syringe oscillation frequency and magnitude, f and A. These parameters will be grouped into non-dimensional terms, e.g., the jet Reynolds number,
the Strouhal number,
the distance-to-diameter ratio, Lc/d, and the normalized wall elasticity, E/ρu2, where ρ and μ are the blood density and viscosity, respectively, and u is the average jet velocity (˜40 mm/s). These parameters will be varied according to the physiological condition of the artery and the design constraints of the catheter. For the simulation, since the deformation of the vessel wall and displacement of the catheter tip are relatively small, COMSOL Multiphysics or ANSYS FLUENT and their moving-mesh capability will be used to solve the FSI problem.
Pressure transmission and reflection under pulsatile flow excitation near a soft aneurysm wall. It has been known that the presence of an aneurysm may significantly affect the local flow pattern in the blood vessel. However, how the flow pattern, the aneurysm deformation and the pressure drop in the flow as affected under the proposed catheter excitation, are yet unknown. A CFD model will be developed to solve this FSI problem of (
After model validation, a systematic study of the CFD simulations will be performed to study the effects of the parameters including the tip-wall distance, Reynolds number, Strouhal number, vessel/aneurysm wall compliance as defined in the preceding section for clot sensing. The geometry of the MCA bifurcation, including the branch diameters and aneurysm size, will also be varied in the parameter study. The CFD results will be used as the training data in the machine learning package, SINDy, to generate the function expression of the pressure drop around the tip in terms of non-dimensional parameters. In the actual sensing procedure, the size and wall thickness of the aneurysm, as well as the geometry of bifurcation will be obtained from the pre-operative scan. With those data, the function relationship between the base pressure and tip-wall distance will be used to 1) optimize the excitation parameters and 2) perform the aneurysm wall sensing task.
Modeling clot engagement and flow into the catheter after pulsatile vacuum. Once the catheter tip is in engagement with the blood clot, the clot may deform and partially or fully enter the catheter tube to block the flow (
The CFD model will include an FEM model of the blood clot, a simplified leaking channel whose length depends on the entrance length of the clot, as well as the theoretical models of the core flow and the annular flow inside/outside the catheter. The elastic properties of the clot will be taken from J. Y. Chueh, A. K. Wakhloo, G. H. Hendricks, C. F. Silva, J. P. Weaver, and M. J. Gounis, “Mechanical characterization of thromboemboli in acute ischemic stroke and laboratory embolus analogs,” Am. J. Neuroradiol., vol. 32, no. 7, pp. 1237-1244, 2011, doi: 10.3174/ajnr.A2485 and W. Merritt et al., “Quantifying the mechanical and histological properties of thrombus analog made from human blood for the creation of synthetic thrombus for thrombectomy device testing,” J. Neurointerv. Surg., vol. 10, no. 12, pp. 1168-1173, 2018, doi: 10.1136/neurintsurg-2017-013675. During engagement, the blood may experience high shear locally inside a small gap. If that is the case, a non-Newtonian model for the blood will be adopted. For model validation, the pressure at the base, P0, and the clot deformation will be compared with those obtained from the in vitro experiment. After model validation, a systematic study of engagement for the governing parameters including the vessel and catheter diameters, the clot size, and the size of the leaking gap will be performed. The results will be used in machine training to generate a function expression of the extent of deformation in relation to the pressure at the base and the syringe displacement, and this function expression will be used in the catheter procedure to assess engagement quality.
Clot Proximity Sensing Ex-vivo experiments and model update/calibration (Phase 1 and 2). The CFD models will provide the expected pressure reading at the proximal end of the catheter as a function of several parameters including tip proximity to the clot. Oscillatory vacuum aspiration will be used within a range consistent with the reported literature and the catheter will be advanced at a fixed rate using a motor-controlled linear stage. Vacuum levels will be recorded at the pull-phase of the syringe's cyclic motion and the distance from the clot will be obtained from image segmentation and measurement of catheter advancement.
Early experiments in Phase 1 will focus on system parameter identification of the catheter pressure transmission with the aim of maximizing the sensitivity of the change in sensed pressure P0 to a change in the distance of the catheter tip to the clot Lc (
During Phase 1 and Phase 2 experiments, vacuum data and the distance of the catheter tip from the clot, the speed of motion of the syringe piston, the rate of change of the pressure, and the rate of insertion of the catheter will be collected. These sensory variables will serve the purpose of auto-labeling of data sets as belonging to pre-clot engagement versus post-clot engagement phases. The conclusion of Phase 1 experiments will yield a statistical model (distribution) describing the expected change in pressure dP0/dLc and the sensed pressure at a given excitation phase as a function of distance from the clot and this will correspond with the nominal vasculature model with its assumed range of diameter variations. At the end of Phase 2 experiments, similar statistical models will be derived for each patient-specific (CT-based) phantom vasculature.
For Phase 3, these rectified models will be used for each anatomical variation to generate a regression map so we can predict the CFD model output (L_c in
With reference to
MODEL RECTIFICATION. After model reduction, CFD model rectification may still be needed for each of the models in Phase 1 and 2 to capture unmodeled effects. This can be achieved by using a feed-forward term for cancelling unmodeled effects based on a regression model. This can be achieved using support vector regression over the error of the parsimonious CFD model relative to the experimental data.
REGRESSION. To enable rapid computation over the entire range of catheter tip and clot locations along the ICA and MCA, an interpolation map will be created for the expected sensed pressure for known excitation parameters based on the rectified parsimonious CFD model. There are several ways to achieve this, first, a thin-plate spline interpolation may be used, but to preserve the statistical covariance information, support vector regression (SVR) and GPR will be considered.
This regression process will also use information from image segmentation regarding the location of the catheter tip relative to the carotid syphon, which will be available from catheter tip tracking and manual segmentation of the cusp of the carotid syphon at the beginning of the procedure. Using parameters of excitation, specific geometric features of the vessel (e.g., diameters at select arc-lengths), and pressure readings the regressor will estimate the distance Lc.
Aneurysm wall contact A similar approach to the clot distance estimation will be followed, but in these experiments a soft semi-transparent small (3-7 mm) aneurysm model from United Biologics will be used and the size of the aneurysm will be varied. The data collection in the clot engagement case will be replicated, but the focus will be more on detecting catheter tip contact with the aneurysm wall. Aneurysm wall proximity will be determined using a 0.9 mm magnetic tracker marker that will be externally attached to the catheter and also used to digitize the depth at which contact with the aneurysm wall is verified visually.
Engagement quality. For these experiments, a straight blood vessel model will be used to minimize the effects of friction with the catheter. A standard aspiration catheter will be equipped with a clear tip made of clear plastic so it can be observed how the clot behaves after pulsed vacuum is applied when in contact with the catheter. Measurements of the catheter traction force when clot retrieval is attempted will be carried out using a high precision single axis load cell that will hold the mock vasculature carrying the clot. Two clot models will be used to compare the traction force of the catheter with a smooth vs. jagged surface. Clots for these experiment groups will be coagulated in the mock blood vessel with a filling rod having a smooth/jagged distal tip, respectively.
These two clot models will be used in the following two-group comparison of pulsatile versus constant vacuum. The first group of experiments will be carried out after applying a fixed vacuum of −27 inHg for three minutes following M. T. Frochler, “Comparison of Vacuum Pressures and Forces Generated by Different Catheters and Pumps for Aspiration Thrombectomy in Acute Ischemic Stroke,” Interv. Neurol., vol. 6, no. 3-4, pp. 199-206, 2017, doi: 10.1159/000475478. The second group will be carried out for the same amount of engagement time, but using the pulsatile vacuum profile that terminates with a fixed vacuum of −27 inHg immediately preceding an attempt of clot retrieval. The catheter will then be withdrawn using a motorized stage while recording the vacuum levels and the traction force experienced by the mock blood vessel supporting the clot. During each experiment, vision data will be used to auto-annotate the data set with a flag indicating success or failure in clot retrieval. This data will be used for the testing of the success of the classification algorithm in predicting clot retrieval. Finally, data of both groups will be compared on the basis of maximal traction force measured and a t-test comparison will be carried out to discern statistically significant differences between the data sets.
The expected outcome of this characterization would be statistical models (distributions) that describe the variability in traction force across different conditions of clot surfaces for both smooth and jagged clot surfaces. Data from the pre- and post-engagement phases will be used to generate a classification outcome predicting whether the clot has been sufficiently engaged prior to a first attempt of clot retrieval.
The challenge in achieving this perception outcome is that, once a clot has engaged the catheter tip, the vacuum measurement alone is not a sufficient indicator for how well the clot has been dislodged and suctioned into the catheter tip. To overcome this challenge, we will first conduct many experiments where sensory data of vacuum, catheter tip motion, distance from clot, and maximal traction force achieved will be collated in a pre-operative statistical model. During a clot retrieval attempt, this model will be fused with the history of intra-operative sensory data to predict the anticipated quality engagement metric. Such information fusion can be achieved using product of distributions. Also, the pre-operative data will be used to train support vector (SV) classifier to predict whether a clot has been sufficiently engaged prior to an attempt of clot retrieval. SV classifiers will be used because they provide a method for classification for cases where the separation hyperplane between the classification groups is nonlinear. The robustness and favorable generalization properties with noisy data, coupled with a compact structure which allows real-time function estimation during motion control, motivate the choice of SV classifiers.
Using experimental data, labeled data sets will be created for training the SV classifiers. The data sets will be used for the jagged and smooth clot groups to investigate feature vector formations that improve classification outcome over the labeled training data sets. The classifiers will be evaluated on artificial data sets generated by including a pool of artificial data sets from the good and poor engagement experiments that were not included in the training phase. The accuracy of the algorithm in predicting good engagement and the correlation between the classifier prediction and the experimental success in clot retrieval will be tested.
Sensory feedback user interface. A graphical user interface (GUI) will be integrated into the proposed experimental setup to provide visual and auditory feedback to neuro-interventionalists about catheter tip proximity to clots and the quality of clot engagement. The GUI will include a visualization of the frontal and lateral views of fluoroscopy imaging of the vasculature. These images will be obtained by transferring the output of the bi-plane fluoroscopy system to the GUI computer via a frame grabber (e.g., Sensoray 2255S). The GUI will enable the surgeon to manually digitize the clot location on the static image (roadmap), which is obtained by applying the “SmartMask” function to the digital subtraction angiography of the vasculature. Once the surgeon suspects proximity to a clot, they will initialize the pre-clot engagement sensing phase and the GUI will display sensed distance and emit a variable pitch sound as a function of proximity to clot interface. After the clot engagement, results of the classifier (good/poor engagement) will be displayed prior to clot retrieval attempt.
Various features of the invention are set forth in the following claims.
This application is a non-provisional of and claims the benefit of U.S. Provisional Patent Application No. 63/210,295, filed on Jun. 14, 2021, the entire contents of which are incorporated herein by reference.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US2022/033502 | 6/14/2022 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63210295 | Jun 2021 | US |
