Data-driven plaque determination in medical imaging
In hemodynamic determination in medical imaging, the classifier is trained from synthetic data rather than relying on training data from other patients. A computer model (in silico) may be perturbed in many different ways to generate many different examples. The flow is calculated for each resulting example. A bench model (in vitro) may similarly be altered in many different ways. The flow is measured for each resulting example. The machine-learnt classifier uses features from medical scan data for a particular patient to estimate the blood flow based on mapping of features to flow learned from the synthetic data. Perturbations or alterations may account for therapy so that the machine-trained classifier may estimate the results of therapeutically altering a patient-specific input feature. Uncertainty may be handled by training the classifier to predict a distribution of possibilities given uncertain input distribution. Combinations of one or more of uncertainty, use of synthetic training data, and therapy prediction may be provided.
1. A method for plaque determination in medical imaging, the method comprising:
acquiring medical scan data representing an anatomical structure of a patient;
extracting a set of features from the medical scan data;
inputting, by a processor, the features to a first machine-trained classifier, the first machine-trained classifier trained to output a value of hemodynamic metric;
inputting the value of the hemodynamic metric and additional features to a second machine-trained classifier; and
outputting, by the processor with application of the first machine-trained classifier to the features, an indicator of a risk of the plaque rupture or a nature of the plaque, the indicator being a function of the value of the hemodynamic metric and being a future evolution of the patient, the future evolution of the patient based on machine training with patient evolution data, the second machine-trained classifier performing the outputting of the indicator in response to the inputting to the second machine-trained classifier.
2. The method of claim 1 wherein outputting comprises outputting the indicator as the risk of the plaque rupture or nature of the plaque as the future evolution of the patient.
3. The method of claim 1 wherein outputting comprises outputting the indicator as the risk of the plaque rupture.
4. The method of claim 1 wherein outputting comprises outputting the indicator as the nature of the plaque.
5. The method of claim 4 wherein outputting comprises outputting where the indicator of the nature of the plaque comprises indication of fibrous tissue, lipid tissue, necrotic tissue, or calcified tissue.
6. The method of claim 1 wherein outputting comprises outputting the indicator as a plaque burden.
7. The method of claim 1 further comprising:
altering at least one of the features, the at least one of the features comprising a plaque-related feature; and
repeating the inputting and outputting with the altered feature.
8. The method of claim 1 wherein inputting comprises inputting an amount of therapy to the second machine-trained classifier, the output of the second machine-trained classifier comprising output by the amount of therapy.
9. The method of claim 1 wherein inputting comprises inputting to the first machine-trained classifier trained from synthetic data, the synthetic data comprising (a) an in vitro model with a ground truth of the hemodynamic metric measured form the in vitro model and/or (b) an in silico model with a ground truth of the hemodynamic metric computed with computation fluid dynamics.
10. The method of claim 9 wherein the synthetic data comprises examples with variation in at least some plaque features.
11. The method of claim 1 wherein extracting the set of the features comprises:
extracting geometrical features of the anatomical structure; and
extracting the features of one or more abnormalities of the anatomical structure.
12. The method of claim 1 wherein extracting the set of the features comprises extracting functional features representing operation of the anatomical structure, wherein the first machine-trained classifier was trained from virtual representations of the operation of the anatomical structure.
13. The method of claim 1 wherein outputting comprises outputting the indicator on a display with an image of the anatomical structure generated from the medical scan data.
14. A method for plaque determination in medical imaging, the method comprising:
generating a plurality of examples of anatomical arrangements;
storing a value for a plaque-related characteristic for each of the examples of the anatomical arrangements;
training, with machine learning, a first classifier for estimating a hemodynamic metric from features extracted from the examples; and
training, with machine learning, using the examples, the hemodynamic metric, and the stored value for the plaque-related characteristic for each of the examples of the anatomical arrangements, a second classifier for predicting the plaque-related characteristic.
15. The method of claim 14 wherein training the second classifier comprises training the second classifier to output the plaque-related characteristic as a likelihood of plaque rupture or a nature of the plaque.
16. The method of claim 14 wherein training the second classifier comprises training the second classifier to output a future evolution of the plaque-related characteristic.
17. A method for plaque determination in medical imaging, the method comprising:
generating a plurality of examples of anatomical arrangements;
storing a value for a plaque-related characteristic for each of the examples of the anatomical arrangements;
training, with machine learning, using the examples and the stored value for the plaque-related characteristic for each of the examples of the anatomical arrangements, a classifier for predicting the plaque-related characteristic;
wherein generating comprises generating with synthetic data not representing any particular patient with perturbing computer modeling, physical modeling, or both in a systematic pattern.