Prediction of risk of post-ablation atrial fibrillation based on radiographic features of pulmonary vein morphology from chest imaging

1. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:

accessing a radiological cardiac image associated with a patient, wherein the radiological cardiac image comprises a plurality of pixels, wherein each pixel of the plurality of pixels has an associated intensity;

obtaining M radiographic features from the radiological cardiac image, wherein the M radiographic features characterize a left atrium size of a heart and a pulmonary vein morphology and wherein M is a positive integer;

extracting, for each radiographic feature of the M radiographic features, an associated value for that radiographic feature based on the radiological cardiac image;

providing to a machine learning classifier, for each radiographic feature of the M radiographic features, the associated value for that radiographic feature;

generating, via the machine learning classifier, a prognosis for atrial fibrillation (AF) after pulmonary vein isolation (PVI), based on a subset of radiographic features of the M radiographic features, wherein the subset of radiographic features is smaller than the M radiographic features, wherein the prognosis comprises a determination of one of recurrence or non-recurrence, and wherein the subset of radiographic features comprise an angle between right pulmonary veins and an angle of left inferior pulmonary vein entry into a left atrium; and

displaying the generated prognosis for AF after PVI.

2. The non-transitory computer-readable medium of claim 1 , wherein the machine learning classifier is a support vector machine (SVM).

3. The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise providing to the machine learning classifier, for each clinical feature of N clinical features, an associated value for that clinical feature, wherein N is a positive integer, and wherein generating the prognosis for AF after PVI is based at least in part on the associated values for each clinical feature of the N clinical features including age, body mass index (BMI), left ventricular ejection fraction (LVEF), history of hypertension, NYHA (New York Heart Association) class of I or greater, and use of apixaban.

4. The non-transitory computer-readable medium of claim 3 , wherein the N clinical features comprise the N most distinguishing clinical features as determined via one of a random forest (RF) algorithm, a t-test, a Wilcoxon rank-sum algorithm, or a minimum redundancy maximum relevance (mRMR) algorithm.

5. The non-transitory computer-readable medium of claim 3 , wherein N is 5.

6. The non-transitory computer-readable medium of claim 1 , wherein the M radiographic features characterize the left atrium size, the pulmonary vein morphology, and an angle of vein entry into a left atrium.

7. The non-transitory computer-readable medium of claim 1 , wherein the subset of radiographic features comprise the most distinguishing radiographic features of the M radiographic features, the most distinguishing radiographic features including an angle between pulmonary veins.

8. The non-transitory computer-readable medium of claim 7 , wherein the subset of radiographic features has 5 radiographic features.

9. The non-transitory computer-readable medium of claim 7 , wherein the M radiographic features comprise one or more of a coronal angle of a left inferior pulmonary vein (LI), an angle between a right superior pulmonary vein (RS) and a right inferior pulmonary vein (RI), a coronal angle of a left superior pulmonary vein (LS), an axial angle of the LI, or a coronal angle of the RI.

10. An apparatus that facilitates generation of a prognosis in connection with a medical condition, the apparatus comprising:

a processor;

a memory configured to store a radiological cardiac image associated with a patient, wherein the radiological cardiac image comprises a plurality of pixels, wherein each pixel of the plurality of pixels has an associated intensity;

an input/output (I/O) interface;

a set of circuits; and

an interface that connects the processor, the memory, the I/O interface, and the set of circuits, the set of circuits comprising:

an image acquisition circuit configured to access the radiological cardiac image;

a feature calculation circuit configured to:

obtain M radiographic features from the radiological cardiac image, wherein the M radiographic features characterize a left atrium size of a heart and a pulmonary vein morphology; and

calculate, for each of the M radiographic features, an associated value of that radiographic feature for the radiological cardiac image, wherein M is a positive integer;

a classifier circuit configured to:

access, for each of the M radiographic features, the associated value of that radiographic feature; and

generate a prognosis for the patient for atrial fibrillation (AF) after pulmonary vein isolation (PVI), based on a subset of radiographic features of the M radiographic features, wherein the subset of radiographic features is smaller than the M radiographic features, wherein the prognosis comprises a determination of one of recurrence or non-recurrence, and wherein the subset of radiographic features comprise an angle between right pulmonary veins and an angle of left inferior pulmonary vein entry into a left atrium; and

a display circuit configured to output the prognosis for the patient associated with the radiological cardiac image.

11. The apparatus of claim 10 , wherein the subset of radiographic features comprise the most distinguishing radiographic features of the M radiographic features, the most distinguishing radiographic features including an angle between pulmonary veins.

12. The apparatus of claim 10 , wherein the classifier circuit is a support vector machine (SVM).

13. The apparatus of claim 10 , wherein the classifier circuit is further configured to access, for each clinical feature of N clinical features, an associated value for that clinical feature, wherein N is a positive integer, and wherein the classifier circuit is configured to generate the prognosis based at least in part on the associated values for each clinical feature of the N clinical features.

14. The apparatus of claim 13 , wherein the N clinical features comprise the N most distinguishing clinical features as determined via one of a random forest (RF) algorithm, a t-test, a Wilcoxon rank-sum algorithm, or a minimum redundancy maximum relevance (mRMR) algorithm.

15. The apparatus of claim 13 , wherein the N clinical features comprise age, body mass index (BMI), left ventricular ejection fraction (LVEF), history of hypertension, NYHA (New York Heart Association) class of I or greater, and use of apixaban.

16. The apparatus of claim 10 , wherein the M radiographic features comprise the M most distinguishing radiographic features as determined via one of a random forest (RF) algorithm, a t-test, a Wilcoxon rank-sum algorithm, or a minimum redundancy maximum relevance (mRMR) algorithm.

17. The apparatus of claim 16 , wherein the M radiographic features comprise one or more of a coronal angle of a left inferior pulmonary vein (LI), an angle between a right superior pulmonary vein (RS) and a right inferior pulmonary vein (RI), a coronal angle of a left superior pulmonary vein (LS), an axial angle of the LI, or a coronal angle of the RI.

18. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:

accessing a radiological cardiac image associated with a heart of a patient;

extracting a plurality of radiographic features from the radiological cardiac image, wherein the plurality of radiographic features characterize a left atrium size of the heart and a pulmonary vein morphology;

providing the plurality of radiographic features to a machine learning classifier, wherein the machine learning classifier is configured to select a subset of the plurality of radiographic features that have a high correlation to recurrence of pulmonary vein isolation (PVI) for the patient; and

generating, via the machine learning classifier, a prognosis for atrial fibrillation (AF) after PVI based on the subset of the plurality of radiographic features, wherein the prognosis comprises a determination of one of recurrence or non-recurrence of AF and wherein the subset of the plurality of radiographic features comprise an angle between right pulmonary veins and an angle of left inferior pulmonary vein entry into a left atrium.

19. The non-transitory computer-readable medium of claim 18 , wherein the subset of the plurality of radiographic features comprise an angle between pulmonary veins.

20. The non-transitory computer-readable medium of claim 18 , wherein the machine learning classifier is a support vector machine (SVM).