Systems and methods for predicting coronary plaque vulnerability from patient-specific anatomic image data
Systems and methods are disclosed for predicting coronary plaque vulnerability, using a computer system. One method includes acquiring anatomical image data of at least part of the patient's vascular system; performing, using a processor, one or more image characteristics analysis, geometrical analysis, computational fluid dynamics analysis, and structural mechanics analysis on the anatomical image data; predicting, using the processor, a coronary plaque vulnerability present in the patient's vascular system, wherein predicting the coronary plaque vulnerability includes calculating an adverse plaque characteristic based on results of the one or more of image characteristics analysis, geometrical analysis, computational fluid dynamics analysis, and structural mechanics analysis of the anatomical image data; and reporting, using the processor, the calculated adverse plaque characteristic.
1 . A computer-implemented method for evaluating cardiac risk of a patient, comprising:
accessing, via at least one processor, a trained machine-learning model that has been trained, based on (i) individual-specific metrics, from a plurality of individuals, that were determined from geometry reconstructed from Computed Tomography (CT) imaging of anatomy of the plurality of individuals, each individual-specific metric quantifying a respective characteristic of plaque, anatomic geometry, or stenosis in the anatomy of the plurality of individuals, and (ii) cardiac risk data of the plurality of individuals, such that the trained machine-learning model has learned associations between the individual-specific metrics and cardiac risk;
obtaining, via the at least one processor, patient-specific data of the patient based on CT imaging of the anatomy of the patient; and
applying, via the at least one processor, the trained machine-learning model to the patient-specific data to assess one or more of plaque, anatomic geometry, or stenosis status of the patient and generate a prediction of cardiac risk of the patient.
2 . The computer-implemented method of claim 1 , wherein the geometry reconstructed from Computed Tomography (CT) imaging of anatomy of the plurality of individuals includes, in each case, a representation of coronary vasculature, including one or more lumens, plaque, or lumen walls.
3 . The computer-implemented method of claim 1 , wherein the one or more individual-specific metrics quantifies one or more of Atherosclerotic Plaque Characteristics (APCS), plaque burden, a numeric cardiac risk score, napkin ring, necrotic core, lumen narrowing, Minimum Lumen Diameter (MLD), Minimum Lumen Area (MLA), percentage diameter stenosis, percentage area stenosis, epicardial fat volume, myocardium shape, or geometry of one or more of an ascending aorta, a coronary artery tree, myocardium, a valve, or a heart chamber.
4 . The computer-implemented method of claim 1 , further comprising:
generating, via the at least one processor, a patient-specific geometric model of the anatomy of the patient based on the patient-specific data, wherein applying the trained machine-learning model to the patient-specific data includes applying the trained machine-learning model to the patient-specific geometric model.
5 . The computer-implemented method of claim 4 , wherein applying the trained machine-learning model to the one patient-specific geometric model includes:
obtaining one or more locations of plaque in the anatomy of the patient;
generating a respective feature vector for each location of the one or more locations, each feature vector including one or more of patient-specific metrics corresponding to the individual-specific metrics used to train the trained machine-learning model or a numerical description of geometry of the location; and
applying the trained machine-learning model to the respective feature vectors.
6 . The computer-implemented method of claim 1 , wherein the prediction of cardiac risk of the patient includes a plurality of risk predictions at various locations in the anatomy of the patient.
7 . The computer-implemented method of claim 4 , wherein the patient-specific geometric model is represented as a list of points in space corresponding to anatomical geometry.
8 . The computer-implemented method of claim 7 , wherein the list further includes, for each point, an identification of one or more neighboring points.
9 . A computer-implemented method for evaluating cardiac risk of a patient, comprising:
accessing, via at least one processor, a trained machine-learning model that has been trained, based on (i) individual-specific metrics, from a plurality of individuals, that were determined from individual-specific geometric models of coronary calcium in anatomy of the plurality of individuals, the individual-specific geometric models reconstructed from Computed Tomography (CT) imaging of the anatomy of the plurality of individuals, each individual-specific metric quantifying a respective characteristic of the coronary calcium represented in the individual-specific geometric models, and (ii) cardiac risk data of the plurality of individuals, such that the trained machine-learning model has learned associations between the individual-specific metrics and cardiac risk,
obtaining, via the at least one processor, patient-specific data of the patient based on CT imaging of the anatomy of the patient; and
applying, via the at least one processor, the trained machine-learning model to the patient-specific data to assess coronary calcium of the patient and generate a prediction of cardiac risk of the patient.
10 . The computer-implemented method of claim 9 , wherein the individual-specific geometric models include, in each case, a representation of coronary vasculature, including one or more lumens, plaque, or lumen walls.
11 . The computer-implemented method of claim 9 , further comprising:
generating, via the at least one processor, a patient-specific geometric model of the anatomy of the patient based on the patient-specific data, wherein applying the trained machine-learning model to the patient-specific data includes applying the trained machine-learning model to the patient-specific geometric model.
12 . The computer-implemented method of claim 11 , wherein applying the trained machine-learning model to the patient-specific geometric model includes:
obtaining one or more locations of plaque in the anatomy of the patient;
generating a respective feature vector for each location of the one or more locations, each feature vector including one or more of a numerical description of one or more physiological or phenotypic parameters of the patient or a numerical description of geometry of the location; and
applying the trained machine-learning model to the respective feature vectors.
13 . The computer-implemented method of claim 9 , wherein the prediction of cardiac risk of the patient includes a plurality of risk predictions at various locations in the anatomy of the patient.
14 . The computer-implemented method of claim 9 , wherein the patient-specific geometric model is represented as a list of points in space corresponding to anatomical geometry.
15 . The computer-implemented method of claim 14 , wherein the list further includes, for each point, an identification of one or more neighboring points.
16 . A computer-implemented method for evaluating cardiac risk of a patient, comprising:
accessing, via at least one processor, a trained machine-learning model that has been trained, based on (i) individual-specific metrics, from a plurality of individuals, that were determined from individual-specific models of one or more of geometry or coronary calcium in anatomy of the plurality of individuals, the individual-specific models reconstructed from Computed Tomography (CT) imaging of the anatomy of the plurality of individuals, each individual-specific metric quantifying a respective characteristic of one or more of the coronary calcium, plaque, anatomic geometry, or stenosis represented in the individual-specific models, and (ii) cardiac risk data of the plurality of individuals, such that the trained machine-learning model has learned associations between the individual-specific metrics and cardiac risk;
obtaining, via the at least one processor, patient-specific data of the patient based on CT imaging of the anatomy of the patient; and
applying, via the at least one processor, the trained machine-learning model to the patient-specific data to assess one or more of coronary calcium, plaque, anatomic geometry, or stenosis status of the patient and generate a prediction of cardiac risk of the patient.
17 . The computer-implemented method of claim 16 , wherein the individual-specific models include, in each case, a representation of coronary vasculature, including one or more lumens, plaque, or lumen walls.
18 . The computer-implemented method of claim 16 , wherein the prediction of cardiac risk of the patient includes a plurality of risk predictions at various locations in the anatomy of the patient.
19 . The computer-implemented method of claim 16 , further comprising:
generating, via the at least one processor, a patient-specific geometric model of the anatomy of the patient based on the patient-specific data, wherein applying the trained machine-learning model to the patient-specific data includes applying the trained machine-learning model to the patient-specific geometric model, wherein the patient-specific geometric model is represented as a list of points in space corresponding to anatomical geometry.
20 . The computer-implemented method of claim 19 , wherein the list further includes, for each point, an identification of one or more neighboring points.