Tumor characterization and outcome prediction through quantitative measurements of tumor-associated vasculature
The present disclosure relates to a method. The method may be performed by accessing data derived from one or more routine clinical medical imaging scans including a lesion in which the lesion and associated vasculature are segmented in a three-dimensional segmentation. At least two features are extracted from the three-dimensional segmentation of the associated vasculature. The at least two features include at least one feature indicative of a morphology of the associated vasculature or a portion thereof, and at least one feature indicative of a function of the associated vasculature or a portion thereof. The at least two features, and/or one or more statistics of the at least two features, are provided to a machine learning model trained to make a prediction concerning the lesion. The prediction concerning the lesion is received from the machine learning model.
1. A method, comprising:
accessing data derived from one or more routine clinical medical imaging scans including a lesion in which the lesion and associated vasculature are segmented in a three-dimensional segmentation;
extracting at least two features, the at least two features including at least one feature indicative of a morphology of the associated vasculature or a portion thereof, and at least one feature indicative of a function of the associated vasculature or a portion thereof, the at least one feature indicative of the morphology extracted from the three-dimensional segmentation of the associated vasculature, and the at least one feature indicative of the function being a pharmacokinetic measurement extracted from a region of tissue in the one or more routine clinical imaging scans that is perfused by the associated vasculature;
providing the at least two features, and/or one or more statistics of the at least two features, to a machine learning model trained to make a prediction concerning the lesion; and
receiving, from the machine learning model, the prediction concerning the lesion.
2. The method of claim 1 , wherein the prediction concerning the lesion comprises one of a predicted response to a treatment for the lesion or a prognosis for the lesion.
3. The method of claim 2 , wherein the prediction concerning the lesion concerns response to neoadjuvant chemotherapy for the lesion.
4. The method of claim 1 , wherein the at least one feature indicative of the morphology of the associated vasculature comprises one or more of a torsion per branch of a plurality of branches of the associated vasculature, a curvature standard deviation per branch of the plurality of branches, a mean curvature per branch, a maximum curvature per branch of the plurality of branches, a curvature skewness per branch of the plurality of branches, a curvature kurtosis per branch of the plurality of branches, a global vascular curvature, the torsion across the plurality of branches, a vessel volume, a vessel volume normalized to a volume of a region of interest comprising the lesion, a vessel volume normalized to a volume of the lesion, a total vessel length, a number of branches of the plurality of branches that enter the lesion, or a percentage of branches of the plurality of branches that enter the lesion.
5. The method of claim 1 , wherein the pharmacokinetic measurement comprises a measurement comprises a measurement of temporal contrast enhancement changes.
6. The method of claim 5 , wherein the pharmacokinetic measurement of temporal contrast enhancement changes comprises signal enhancement ratio, time to peak enhancement, rate of uptake, or rate of washout.
7. The method of claim 1 , wherein the machine learning model is one of, or an ensemble of two or more of, a logistic regression model, a Cox regression model, a Least Absolute Shrinkage and Selection Operator (LASSO) regression model, a naïve Bayes classifier, a support vector machine (SVM) with a linear kernel, a SVM with a radial basis function (RBF) kernel, a linear discriminant analysis (LDA) classifier, a quadratic discriminant analysis (QDA) classifier, a logistic regression classifier, a decision tree, a random forest, a diagonal LDA, a diagonal QDA, a neural network, an AdaBoost algorithm, an elastic net, a Gaussian process classification, or a nearest neighbors classification.
8. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform the method of claim 1 .
9. An apparatus, comprising:
memory configured to store data derived from one or more routine clinical medical imaging scans including a lesion, wherein the lesion and associated vasculature are segmented in a three-dimensional segmentation; and
one or more processors configured to:
access the data derived from one or more routine clinical medical imaging scans including the lesion in which the lesion and the associated vasculature are segmented in the three-dimensional segmentation,
extract at least two features, the at least two features including at least one feature indicative of a morphology of the associated vasculature or a portion thereof, and at least one feature indicative of a function of the associated vasculature or a portion thereof, the at least one feature indicative of the morphology extracted from the three-dimensional segmentation of the associated vasculature, and the at least one feature indicative of the function being a pharmacokinetic measurement extracted from a region of tissue in the one or more routine clinical imaging scans that is perfused by the associated vasculature, and
provide the at least two features, and/or one or more statistics of the at least two features, to a machine learning model trained to make a prediction concerning the lesion.
10. The apparatus of claim 9 , wherein the at least one feature indicative of the morphology includes a feature describing a vessel orientation.
11. The apparatus of claim 9 , wherein the at least one feature indicative of the function includes a pharmacokinetic measurement of temporal contrast enhancement changes.
12. The apparatus of claim 9 , wherein the prediction relates to a response to neoadjuvant chemotherapy.
13. A method, comprising:
accessing data derived from one or more routine clinical medical imaging scans including a lesion in which the lesion and associated vasculature are segmented in a three-dimensional segmentation;
extracting at least one feature indicative of a spatial organization of the associated vasculature, the at least one feature taken from a projection onto a plane of at least a portion of the three-dimensional segmentation of the associated vasculature;
providing the at least one feature and/or at least one statistic of the at least one feature to a trained machine learning model; and
receiving, from the trained machine learning model, a prediction concerning the lesion.
14. The method of claim 13 , wherein the prediction concerning the lesion comprises one of a predicted response to a treatment for the lesion or a prognosis for the lesion.
15. The method of claim 14 , wherein the prediction concerns response to neoadjuvant chemotherapy for the lesion.
16. The method of claim 13 , wherein the at least one feature indicative of the spatial organization of the associated vasculature comprise one or more of a vessel orientation along a XY projection image, a vessel orientation along a XZ projection image, a vessel orientation along a YZ projection image, a vessel orientation along a rotation-elevation projection image, a vessel orientation along a distance-rotation projection image, or a vessel orientation along a distance-elevation projection image.
17. The method of claim 13 , further comprising:
extracting at least one feature indicative of a morphology of the associated vasculature; and
providing the at least one feature indicative of the spatial organization and the at least one feature indicative of a morphology and/or statistics thereof to the trained machine learning model.
18. The method of claim 17 , further comprising:
identifying branches of the associated vasculature in the three-dimensional segmentation;
wherein the at least one feature indicative of morphology is based, at least in part, on the branches of the associated vasculature.
19. The method of claim 17 , wherein the at least one feature indicative of the morphology comprises one or more of a torsion per branch of a plurality of branches of the associated vasculature, a curvature standard deviation per branch of the plurality of branches, a mean curvature per branch, a maximum curvature per branch of the plurality of branches, a curvature skewness per branch of the plurality of branches, a curvature kurtosis per branch of the plurality of branches, a global vascular curvature, the torsion across the plurality of branches, a vessel volume, a vessel volume normalized to a volume of a region of interest comprising the lesion, a vessel volume normalized to a volume of the lesion, a total vessel length, a number of branches of the plurality of branches that enter the lesion, or a percentage of branches of the plurality of branches that enter the lesion.
20. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform the method of claim 13 .