Automatic generation of fundus drawings
Techniques for automating the generation and analysis of fundus drawings are described. Captured images undergo image processing to extract information about image features. Fundus images are generated and recommended labels for the fundus drawing are generated. Fundus drawings can be analyzed and undergo textual processing to extract existing labels. Machine learning models and co-occurrence analysis can be applied to collections of fundus images and drawings to gather information about commonly associated labels, label locations, and user information. The most frequently used labels associated with the image can be identified to improve recommendations and personalize labels.
1. A method implemented by one or more computing devices, comprising:
receiving data indicative of a first medical image;
identifying one or more image features, the one or more image features being indicative of a condition or a label associated with the first medical image;
generating a second medical image based on the one or more image features;
associating a classifier with the one or more image features; and
applying one or more recommendation models to identify, based at least in part on the associated classifier, one or more recommended labels and a label location within the second medical image for each of the one or more recommended labels, one of the one or more recommendation models identifying the label location based at least on performing a co-occurrence analysis to compute a location of the one or more recommended labels in relation to another label.
2. The method of claim 1 , wherein the first medical image is a fundus image and wherein the second medical image is a fundus drawing.
3. The method of claim 1 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the second medical image for each of the one or more recommended labels comprises:
predicting a label location within the second medical image for a recommended label based on a location of a label that is deemed to be similar to the recommended label in one or more other medical images.
4. The method of claim 1 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the second medical image for each of the one or more recommended labels comprises:
designating a particular recommended label as a global label or a local label; and
in response to designating the particular recommended label as a local label, performing a co-occurrence analysis to compute a location of the particular recommended label in relation to another label.
5. The method of claim 1 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the second medical image for each of the one or more recommended labels comprises:
identifying one or more users associated with the associated classifier; and
identifying a recommended label based on labels previously used by the identified one or more users.
6. The method of claim 1 , wherein identifying the one or more image features comprises using an image segmentation technique.
7. The method of claim 1 , wherein identifying the one or more image features comprises using one or more of a scale invariant feature transform, histogram of oriented gradients, learning based analysis, and eigenanalysis.
8. The method of claim 1 , wherein identifying the one or more image features comprises extracting textual features.
9. The method of claim 1 , wherein associating the classifier with the one or more image features comprises:
associating a classifier selected from a set of classifiers to the one or more image features; or
assigning a new classifier to the one or more image features responsive to determining that no relevant classifiers are present in the preexisting set of classifiers.
10. A system, comprising:
one or more processors; and
a memory storing computer readable instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
identifying one or more image features indicative of condition or a label associated with a medical image;
identifying a classifier to be associated with the one or more image features; and
applying one or more recommendation models to identify, based at least in part on the associated classifier, one or more recommended labels and a label location within the medical image for each of the one or more recommended labels, one of the one or more recommendation models identifying the label location based at least on performing a co-occurrence analysis to compute a location of the one or more recommended labels in relation to another label.
11. The system of claim 10 , wherein the medical image is one of a fundus image or a fundus drawing.
12. The system of claim 10 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the medical image for each of the one or more recommended labels comprises:
predicting a label location within the medical image for a recommended label based on a location of a label that is deemed to be similar to the recommended label in one or more other medical images.
13. The system of claim 10 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the medical image for each of the one or more recommended labels comprises:
designating a particular recommended label as a global label or a local label; and
in response to designating the particular recommended label as a local label, performing a co-occurrence analysis to compute a location of the particular recommended label in relation to another label.
14. The system of claim 10 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the one or more recommended labels and the label location within the medical image for each of the one or more recommended labels comprises:
identifying one or more users associated with the associated classifier; and
identifying a recommended label based on labels previously used by the identified one or more users.
15. The system of claim 10 , wherein identifying the one or more image features comprises using an image segmentation technique.
16. The system of claim 10 , wherein identifying the one or more image features comprises using one or more of a scale invariant feature transform, histogram of oriented gradients, learning based analysis, and eigenanalysis.
17. The system of claim 10 , wherein identifying the one or more image features comprises extracting textual features.
18. A method implemented by one or more computing devices, comprising:
identifying one or more image features indicative of a condition or a label associated with a medical image;
associating a classifier with the one or more image features;
identifying one or more recommended labels; and
applying one or more recommendation models to identify, based at least in part on the associated classifier, a label location within the medical image for each of the one or more recommended labels, one of the one or more recommendation models identifying the label location based at least on performing a co-occurrence analysis to compute a location of the one or more recommended labels in relation to another label.
19. The method of claim 18 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the label location within the medical image for each of the one or more recommended labels comprises:
predicting a label location within the medical image for a recommended label based on a location of a label that is deemed to be similar to the recommended label in one or more other medical images.
20. The method of claim 18 , wherein applying the one or more recommendation models to identify, based at least in part on the associated classifier, the label location within the medical image for each of the one or more recommended labels comprises:
designating a particular recommended label as a global label or a local label; and
in response to designating the particular recommended label as a local label, performing a co-occurrence analysis to compute a location of the particular recommended label in relation to another label.