Structured representations for interpretable machine learning applications in medical imaging
Systems and method can be provided to transform input data (e.g., CT imaging data) into structured representations to create interpretable models. Another aspect of the current invention can be generating labels synthetically to apply to real data according to a biologically-based labelling technique to guide the model training with a priori mechanistic knowledge.
1. A method for creating an interpretable model from patient images, the method comprising:
receiving, via a processor, imaging data of a region of interest of a patient;
processing, via the processor, the imaging data to provide spatially-localized or temporally-localized information;
creating, via the processor, a structured representation of the spatially-localized or temporally-localized information that causes at least some of the spatially-localized or temporally-localized information to be indicated as significant spatially-localized or temporally-localized information, wherein creating the structure representation includes:
i) determining a plurality of cross sections at various location in the region of interest;
ii) unwrapping each of the plurality of cross sections into two one dimensional vectors to create a plurality of two one dimensional vectors;
iii) concatenating the plurality of two one dimensional vectors to create the structured representation of the spatially-localized or temporally localized information; and
using, via the processor, the structured representations for the purpose of training to create the interpretable model, wherein the significant spatially-localized or temporally-localized information contributes to the model inference.
2. The method of claim 1 further comprising using the trained interpretable model for inference.
3. The method of claim 1 , wherein the spatially-localized or temporally-localized information includes anatomic information, functional information, tissue characteristics, or any combination thereof.
4. The method of claim 1 , wherein the interpretable model output is a quantitative imaging response variable.
5. The method of claim 1 , wherein the interpretable model output is a measure of cardiovascular disease status.
6. The method of claim 1 , wherein the interpretable model output is a hemodynamic property comprising measure of fractional flow reserve, myocardial blood flow, or any combination thereof.
7. The method of claim 1 , wherein the interpretable model output is an event prediction.
8. The method of claim 7 , wherein the event prediction is an adverse cardiovascular event or an adverse neurovascular event.
9. The method of claim 1 , wherein the interpretable model output is a measure of molecular activity.
10. The method of claim 9 , wherein the molecular activity pertains to gene expression or protein levels.
11. The method of claim 1 , wherein the significant spatially-localized or temporally-localized information is a tubular structure where proximal function of the tubular structure depends at least in part on distal function of the tubular structure.
12. The method of claim 1 , wherein the significant spatially-localized or temporally-localized information couples at least two or more structures in the spatially-localized or temporally-localized information where the interpretable model output is dominated by a subset of those structures.
13. The method of claim 1 , wherein the significant spatially-localized or temporally-localized information is contiguous tissues that manifest differing molecular activity based on biological processes.
14. The method of claim 1 further comprising:
generating, via the processor, interpretable labels for unlabeled data based on a function that incorporates mechanistic rationale; and
wherein creating the interpretable model further comprises using, via the processor, the interpretable labels for the purpose of further training the interpretable model, wherein the mechanistic rationale contributes to the model inference.
15. The method of claim 14 , wherein the interpretable model output is a quantitative imaging response variable.
16. The method of claim 14 , wherein the interpretable model output is a measure of cardiovascular disease status.
17. The method of claim 14 , wherein the interpretable model output is a hemodynamic property comprising measure of fractional flow reserve, myocardial blood flow, or any combination thereof.
18. The method of claim 14 , wherein the interpretable model output is an event prediction.
19. The method of claim 18 , wherein the event prediction is an adverse cardiovascular event or an adverse neurovascular event.
20. The method of claim 19 , wherein the interpretable model output is a measure of molecular activity.
21. The method of claim 20 , wherein the molecular activity pertains to gene expression or protein levels.
22. A system comprising a processor and a non-transient storage medium including processor executable instructions configured to cause the processor to:
receive imaging data of a region of interest of a patient;
process the imaging data to provide spatially-localized or temporally-localized information;
create a structured representation of the spatially-localized or temporally-localized information that causes at least some of the spatially-localized or temporally-localized information to be indicated as significant spatially-localized or temporally-localized information, wherein creating the structure representation includes:
i) determine a plurality of cross sections at various location in the region of interest;
ii) unwrap each of the plurality of cross sections into two one dimensional vectors to create a plurality of two one dimensional vectors;
iii) concatenate the plurality of two one dimensional vectors to create the structured representation of the spatially-localized or temporally localized information; and
use the structured representations for the purpose of training to create an interpretable model, wherein the significant spatially-localized or temporally-localized information contributes to the model inference.
23. The system of claim 22 further comprising using the trained interpretable model for inference.
24. The system of claim 22 , wherein the spatially-localized or temporally-localized information includes anatomic information, functional information, tissue characteristics, or any combination thereof.
25. The system of claim 22 , wherein the interpretable model output is a quantitative imaging response variable.
26. The system of claim 22 wherein the system comprising a processor executable instructions are further configured to cause the processor to:
generate interpretable labels for unlabeled data based on a function that incorporates mechanistic rationale; and
wherein to create the interpretable model, the processor executable instructions are further configured to cause the processor to use the interpretable labels for the purpose of further training the interpretable model, wherein the mechanistic rationale contributes to the model inference.
27. The system of claim 26 , wherein the interpretable model output is a quantitative imaging response variable.
28. The system of claim 26 , wherein the interpretable model output is a measure of cardiovascular disease status.