Assessment of arterial calcifications
AssigneeCase Western Reserve University
InventorsDavid L. Wilson, Di Wen, Brendan L. Eck, Jacob Levi, Yingnan Song
Life of this patent
Estimating the term…
Abstract
Embodiments discussed herein facilitate employing a pretrained model to determine risk(s) of adverse event(s) based on Computed Tomography (CT) image volume(s) of an artery and/or training a model to determine such risk(s). Example embodiments can determine risk based on territory-specific calcium score(s) and/or intensity/morphological/location features discussed herein. Various embodiments can determine risk(s) based on a single CT image volume and/or change(s) in CT image volumes taken over a series of time points.
Claims23
A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
- accessing a Computed Tomography (CT) image volume of at least one artery of a patient;
- identifying one or more calcification candidates based on the CT image volume;
- identifying one or more territories of the at least one artery based on a registration of an atlas to the CT image volume;
- calculating, for each territory of the one or more territories, an associated calcium score for that territory, based on the registration of the atlas to the CT image volume;
- extracting, for each calcification candidate of the one or more calcification candidates, one or more associated features for that calcification candidate;
- providing the one or more associated features for each calcification candidate to a pretrained model, wherein the pretrained model is one of a pretrained Machine Learning model or a pretrained Deep Learning model;
- and
- making a risk assessment based upon the calcium score associated with a territory of the one or more territories, and not upon a whole heart calcium score, and also based upon the one or more associated features.
A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
- accessing, for each of a plurality of time points respectively occurring at different times, an associated Computed Tomography (CT) image volume for that time point, wherein the associated CT image volume for each time point of the plurality of time points is an associated CT image volume of at least one artery of a patient;
- calculating, for each territory of one or more territories of the at least one artery, an associated calcium score for that territory for each time point, based on the associated CT image volume for that time point;
- extracting, from each calcified region of one or more calcified regions of the at least one artery, one or more associated features for that calcified region for each time point, from the associated CT image volume for that time point;
- co-registering the associated CT image volume for each time point of the plurality of time points;
- analyzing, for each calcified region of the one or more calcified regions, changes in the associated features for that calcified region over the plurality of time points;
- and
- predicting, via a pretrained model, an associated risk for each adverse event of one or more adverse events, based on one or more of the associated calcium score for each territory of the one or more territories and for each time point of the plurality of time points, or the changes in the associated features for each calcified region of the one or more calcified regions over the plurality of time points, wherein the pretrained model is one of a pretrained Machine Learning model or a pretrained Deep Learning model.
A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
- accessing a plurality of Computed Tomography (CT) image volumes associated with at least one artery;
- obtaining, for each CT image volume of the plurality of CT image volumes, a clinical outcome of an associated patient for that CT image volume;
- identifying, based on each CT image volume of the plurality of CT image volumes, one or more associated calcification candidate regions in that CT image volume;
- extracting, from each associated calcification candidate region of the one or more associated calcification candidate regions of each CT image volume of the plurality of CT image volumes, one or more associated features for that associated calcification candidate region of that CT image volume;
- and
- for each CT image volume of the plurality of CT image volumes, training a model based on the clinical outcome of the associated patient for that CT image volume and the one or more associated features for each associated calcification candidate region of the one or more associated calcification candidate regions of that CT image volume, wherein the model is one of a Machine Learning model or a Deep Learning model.
Dependencies are inferred from the claim text.