System and methods for estimation of blood flow characteristics using reduced order model and machine learning
Systems and methods are disclosed for determining blood flow characteristics of a patient. One method includes: receiving, in an electronic storage medium, patient-specific image data of at least a portion of vasculature of the patient having geometric features at one or more points; generating a patient-specific reduced order model from the received image data, the patient-specific reduced order model comprising estimates of impedance values and a simplification of the geometric features at the one or more points of the vasculature of the patient; creating a feature vector comprising the estimates of impedance values and geometric features for each of the one or more points of the patient-specific reduced order model; and determining blood flow characteristics at the one or more points of the patient-specific reduced order model using a machine learning algorithm trained to predict blood flow characteristics based on the created feature vectors at the one or more points.
1. A computer-implemented method of determining blood flow characteristics of a patient, the method comprising:
receiving, in an electronic storage medium, a patient-specific anatomical model of at least a portion of the patient's vasculature including geometric features at one or more points;
truncating the patient-specific anatomical model at locations of the patient-specific anatomical model for which blood flow characteristics are to be determined;
applying boundary conditions to the truncated patient-specific anatomical model, the boundary conditions associated with hemodynamics at an inflow of blood flow, outflow of blood flow, and a vessel wall of the truncated patient-specific anatomic model;
using the boundary conditions to estimate a first value for the blood flow characteristic at one or more locations of the truncated patient-specific anatomic model;
generating a reduced-order model that is representative of the truncated patient-specific anatomic model, the reduced-order model including one or more points having an impedance value based on the estimated first value of the blood flow characteristic at the one or more locations of the truncated patient-specific anatomic model;
updating the reduced-order model by employing a machine-learning algorithm that is trained, based on errors determined between at least one impedance value of at least one training reduced-order model and at least one corresponding impedance value determined by computational fluid dynamics, to reduce error in one or more impedance values of an input reduced-order model; and
using the updated reduced-order model to determine a second value for the blood flow characteristic at the one or more locations of the truncated patient-specific anatomic model.
2. The computer-implemented method of claim 1 , wherein the machine learning algorithm includes one or more of: a support vector machine (SVM), a multi-layer perceptron (MLP), a multivariate regression (MVR), a neural network, a tree-based classifier, or a weighted linear or logistic regression.
3. The computer-implemented method of claim 1 , wherein the blood flow characteristics include one or more of: a blood pressure, a fractional flow reserve (FFR), a blood flow rate or a flow velocity, a velocity or pressure field, a hemodynamic force, and an organ and/or tissue perfusion characteristic.
4. The computer-implemented method of claim 1 , wherein the reduced order model includes estimates of third values that simplify geometric features at the one or more points of the vasculature of the patient.
5. The computer-implemented method of claim 1 , wherein the impedance value includes one or more of a resistance value, a capacitance value, or inductance value.
6. The computer-implemented method of claim 1 , further comprising:
splitting the truncated patient-specific anatomical model into a plurality of regions of the patient's vasculature,
wherein the reduced-order model includes a respective reduced-order model for each of the plurality of regions of the patient's vasculature.
7. The computer-implemented method of claim 6 , wherein the splitting is based on the estimated first values of the blood flow characteristics.
8. The computer-implemented method of claim 7 , wherein the splitting is configured to distinguish between one or more healthy regions of the patient's vasculature and one or more unhealthy regions of the patient's vasculature.
9. The computer-implemented method of claim 8 , wherein the one or more unhealthy regions of the patient's vasculature include one or more of an ostial bifurcation, a non-ostial bifurcation, a stenosed region, or an expansion region.
10. A computer-implemented method of training a machine-learning algorithm to reduce error in an input reduced order model representative of at least a portion of a patient's vasculature, the method comprising:
for a plurality of individuals:
receiving, in an electronic storage medium, an individual-specific anatomical model of at least a portion of the individual's vasculature including geometric features at one or more points;
truncating the individual-specific anatomical model at locations of the individual-specific anatomical model for which blood flow characteristics are to be determined;
applying boundary conditions to the truncated individual-specific anatomical model, the boundary conditions associated with hemodynamics at an inflow of blood flow, outflow of blood flow, and a vessel wall of the truncated individual-specific anatomic model;
using the boundary conditions to estimate a first value for the blood flow characteristic at one or more locations of the truncated individual-specific anatomic model;
generating a reduced-order model that is representative of the truncated individual-specific anatomic model, the reduced-order model including one or more points having a first impedance value based on the estimated first value of the blood flow characteristic at the one or more locations of the truncated individual-specific anatomic model;
determining, via computational fluid dynamics, a second impedance value for the one or more locations of the truncated individual-specific anatomic model based on the estimated first values of the blood flow characteristics; and
determining an error value between each first impedance value and the corresponding second impedance value; and
training the machine-learning algorithm, using the determined errors from the plurality of individuals, to reduce error in one or more impedance values of an input reduced-order model.
11. The computer-implemented method of claim 10 , wherein the machine learning algorithm includes one or more of: a support vector machine (SVM), a multi-layer perceptron (MLP), a multivariate regression (MVR), a neural network, a tree-based classifier, or a weighted linear or logistic regression.
12. The computer-implemented method of claim 10 , wherein the blood flow characteristics include one or more of: a blood pressure, a fractional flow reserve (FFR), a blood flow rate or a flow velocity, a velocity or pressure field, a hemodynamic force, and an organ and/or tissue perfusion characteristic.
13. The computer-implemented method of claim 10 , wherein the reduced order model includes estimates of third values that simplify geometric features at the one or more points of the vasculature of the patient.
14. The computer-implemented method of claim 10 , wherein the impedance value includes one or more of a resistance value, a capacitance value, or inductance value.
15. The computer-implemented method of claim 10 , further comprising, for each of the plurality of individual:
splitting the truncated patient-specific anatomical model into a plurality of regions of the patient's vasculature;
wherein:
the reduced-order model includes a respective reduced-order model for each of the plurality of regions of the patient's vasculature; and
the splitting is based on the estimated first values of the blood flow characteristics.
16. The computer-implemented method of claim 15 , wherein:
the splitting is configured to distinguish between one or more healthy regions of the patient's vasculature and one or more unhealthy regions of the patient's vasculature; and
the one or more unhealthy regions of the patient's vasculature include one or more of an ostial bifurcation, a non-ostial bifurcation, a stenosed region, or an expansion region.
17. A system for determining blood flow characteristics of a patient, the system comprising:
at least one memory storing instructions and a machine learning algorithm trained, based on errors determined between at least one impedance value of at least one training reduced-order model and at least one corresponding impedance value determined by computational fluid dynamics, to reduce error in one or more impedance values of an input reduced-order model; and
at least one processor operatively connected to the memory and configured to execute the instructions to perform operations, including:
receiving, in the memory, a patient-specific anatomical model of at least a portion of the patient's vasculature including geometric features at one or more points;
truncating the patient-specific anatomical model at locations of the patient-specific anatomical model for which blood flow characteristics are to be determined;
applying boundary conditions to the truncated patient-specific anatomical model, the boundary conditions associated with hemodynamics at an inflow of blood flow, outflow of blood flow, and a vessel wall of the truncated patient-specific anatomic model;
using the boundary conditions to estimate a first value for the blood flow characteristic at one or more locations of the truncated patient-specific anatomic model;
generating a reduced-order model that is representative of the truncated patient-specific anatomic model, the reduced-order model including one or more points having an impedance value based on the estimated first value of the blood flow characteristic at the one or more locations of the truncated patient-specific anatomic model;
updating the reduced-order model by employing the machine-learning algorithm; and
using the updated reduced-order model to determine a second value for the blood flow characteristic at the one or more locations of the truncated patient-specific anatomic model.
18. The system of claim 17 , wherein the operations further include:
splitting the truncated patient-specific anatomical model into a plurality of regions of the patient's vasculature;
wherein the reduced-order model includes a respective reduced-order model for each of the plurality of regions of the patient's vasculature.
19. The system of claim 18 , wherein the splitting is based on the estimated first values of the blood flow characteristics.
20. The system of claim 19 , wherein the splitting is configured to distinguish between one or more healthy regions of the patient's vasculature and one or more unhealthy regions of the patient's vasculature.