Covariate modulate atlas
The disclosed technology relates generally to medical imaging, and more particularly, some embodiments relate to systems and methods for creating and using a covariate modulated or “dynamic” atlas. Some embodiments of the disclosure provide a method for predicting an alas using General Additive Model (GAM) parameters, wherein the GAM parameters are derived by registering (and optionally segmenting) a plurality of image data sets from a plurality of different subjects to an initial atlas estimate (e.g., a seed atlas), and analyzing the resulting registration, segmentation, and intensity parameters as correlated with input covariates.
1. A method for building an atlas comprising:
receiving, for individual ones of a plurality of subjects, 1) a subject dependent image data set of a target anatomical region from an image data source, and 2) a set of subject dependent covariates from one or both of the image data source or a demographic data source;
receiving an atlas estimate, the atlas estimate including atlas parameters;
determining a set of subject dependent registration parameters by registering the subject dependent image data set to the atlas estimate;
determining a set of general additive model (GAM) parameters based on the registration parameters as a function of the covariates, wherein determining the GAM parameters comprises determining a fit quality for a registration, and responsive to the fit quality for the registration falling outside a predetermined threshold range, discarding the registration parameters, or, responsive to the fit quality falling within the predetermined threshold range, using the registration parameters to determine the GAM parameters; and
responsive to the fit quality falling within the predetermined threshold range, determining an updated atlas estimate using the GAM parameters and the atlas parameters based on a target set of covariates.
2. The method of claim 1 , wherein the subject dependent covariates comprise one or both of demographic parameters or imaging parameters.
3. The method of claim 2 , wherein the demographic parameters comprise one or more of age, gender, ethnicity, genetic factors, medical history, or relevant clinical measures.
4. The method of claim 2 , wherein the imaging parameters comprise one or more of imaging modality parameters, manufacturer's software specifications, or manufacturer's hardware specifications.
5. The method of claim 1 , wherein the atlas parameters comprise one or more of mean signal strength from different imaging modalities, scanner type, field strength, various magnetic properties for MR modality, prior probabilities of tissues and anatomic structures, prior probabilities of neighboring tissues and anatomical structures, shape parameters, or texture parameters.
6. The method of claim 1 , wherein the image data source comprises one or both of an imaging modality source or a picture archive communication system (PACS).
7. The method of claim 1 , further comprising:
receiving a plurality of additional subject dependent imaging data sets and corresponding additional sets of covariates,
registering the additional subject dependent imaging data sets to the atlas,
receiving additional registration parameters, and
determining additional GAM parameters as a function of the corresponding additional sets of covariates and corresponding registration parameters.
8. The method of claim 1 , wherein registering the subject dependent image data set to the atlas estimate comprises an affine rigid body atlas registration process based on one or more of fiducial points, a number of voxels, voxel locations, or image intensities indicated by the image data set.
9. A system configured to build an atlas, the system comprising one or more hardware processors configured by machine-readable instructions to:
receive, for individual ones of a plurality of subjects, 1) a subject dependent image data set of a target anatomical region from an image data source, and 2) a set of subject dependent covariates from one or both of the image data source or a demographic data source;
receive an atlas estimate, the atlas estimate including atlas parameters;
determine a set of subject dependent registration parameters by registering the subject dependent image data set to the atlas estimate;
determine a set of general additive model (GAM) parameters based on the registration parameters and as a function of the covariates, wherein determining the GAM parameters comprises determining a fit quality for a registration, and responsive to the fit quality for the registration falling outside a predetermined threshold range, discarding the registration parameters, or, responsive to the fit quality falling within the predetermined threshold range, using the registration parameters to determine the GAM parameters; and
responsive to the fit quality falling within the predetermined threshold range, determine an updated atlas estimate using the GAM parameters and the atlas parameters based on a target set of covariates.
10. The system of claim 9 , wherein the subject dependent covariates comprise one or both of demographic parameters or imaging parameters.
11. The system of claim 10 , wherein the demographic parameters comprise one or more of age, gender, ethnicity, genetic factors, medical history, or relevant clinical measures.
12. The system of claim 10 , wherein the imaging parameters comprise one or more of imaging modality parameters, manufacturer's software specifications, or manufacturer's hardware specifications.
13. The system of claim 9 , wherein the atlas parameters comprise one or more of mean signal strength from different imaging modalities, scanner type, field strength, various magnetic properties for MR modality, prior probabilities of tissues and anatomic structures, prior probabilities of neighboring tissues and anatomical structures, shape parameters, or texture parameters.
14. The system of claim 9 , wherein the image data source comprises one or more of an imaging modality source or a picture archive communication system (PACS).
15. The system of claim 9 , wherein the one or more hardware processors are further configured to:
receive a plurality of additional subject dependent imaging data sets and corresponding additional sets of covariates;
register the plurality of additional subject dependent imaging data sets to the atlas;
receive additional registration parameters;
determine additional GAM parameters as a function of the additional corresponding sets of covariates and corresponding registration parameters; and
determine the updated atlas estimate using the additional GAM parameters and the atlas parameters based on the target set of covariates.
16. The system of claim 9 , wherein the one or more hardware processors comprise a GAM server, the GAM server comprising a registration engine and a GAM engine, wherein:
the GAM server is configured to receive, for the individual ones of the plurality of subjects, the subject dependent image data set of the target anatomical region from the image data source and the set of subject dependent covariates from one or both of the image data source or the demographic data source;
the GAM server is configured to receive the atlas estimate;
the registration engine is configured to determine the set of subject dependent registration parameters by registering the subject dependent image data set to the atlas estimate;
the GAM engine is configured to determine the set of GAM parameters based on the registration parameters as a function of the covariates; and
the GAM server is configured to determine the updated atlas estimate using the GAM parameters and the atlas parameters based on the target set of covariates.
17. The system of claim 16 , wherein the GAM server further comprises a prediction engine configured to determine the updated atlas based on the subject dependent covariates, the atlas estimate, the registration parameters, and the GAM parameters.
18. The system of claim 9 , wherein the one or more hardware processors are configured such that registering the subject dependent image data set to the atlas estimate comprises an affine rigid body atlas registration process based on one or more of fiducial points, a number of voxels, voxel locations, or image intensities indicated by the image data set.