Method and system for computer-aided triage of stroke

1. A method for detecting a suspected hemorrhage, the method comprising:

receiving a set of images associated with a patient;

processing the set of images with a set of trained models to identify a particular region from the set of images, wherein producing the set of trained models comprises:

performing a first training process, comprising training the set of trained models based on a first set of images labeled as positive with respect to a hemorrhage and a second set of images labeled as negative with respect to a hemorrhage;

using the set of trained models to produce a set of outputs of the first training process;

identifying that the set of outputs of the first training process comprises a set of false positive hemorrhage determinations; and

updating the set of trained models by performing a second training process, comprising training the set of trained models based on the set of outputs of the first training process, the set of outputs comprising the identified set of false positive hemorrhage determinations;

in response to detecting the suspected hemorrhage based on processing the set of images, triggering an action, the action operable to improve an efficiency of triage of the patient.

2. The method of claim 1 , wherein the action is triggered at a user device associated with a specialist.

3. The method of claim 2 , wherein the user device is a personal mobile user device of the specialist.

4. The method of claim 2 , wherein the action triggers assignment of treatment of the patient to the specialist.

5. The method of claim 4 , wherein the action comprises a notification at the user device.

6. The method of claim 1 , wherein receiving the set of images, processing the set of images, and triggering the action are performed during triage of the patient.

7. The method of claim 6 , wherein the patient initially arrives at a first point of care.

8. The method of claim 7 , wherein the specialist is associated with a second point of care, the second point of care remote from the first point of care.

9. The method of claim 1 , wherein the second training process is performed after the first training process.

10. A system for detecting a suspected hemorrhage, the system comprising:

a set of trained models, wherein the set of trained models is produced with:

a first training process, wherein in the first training process, the set of trained models is trained based on a first set of images labeled as positive with respect to a hemorrhage and a second set of images labeled as negative with respect to a hemorrhage;

a false positive identification process, wherein in the false positive identification process, the set of trained models are used to produce a set of outputs of the first training process, and a set of false positive hemorrhage determinations are identified within the set of outputs of the first training process; and

a second training process, wherein in the second training process, the set of trained models is trained based on the identified set of false positive hemorrhage determinations;

a computing subsystem, wherein the computing subsystem:

receives a third set of images associated with a patient; and

processes the third set of images with the set of trained models to identify a particular region from the third set of images; and

a client application executable on a user device of a recipient, wherein the client application:

triggers an action in response to detecting the suspected hemorrhage based on processing the third set of images, the action operable to improve an efficiency of triage of the patient.

11. The system of claim 10 , wherein the particular region comprises a segmented hyperdense region in the third set of images, wherein detecting the suspected hemorrhage comprises determining that a volume of the segmented hyperdense region exceeds a predetermined threshold.

12. The system of claim 10 , wherein the set of false positive hemorrhage determinations is produced as a set of outputs of the first training process.

13. The system of claim 10 , wherein the second training process is performed after the first training process.

14. A method for detecting a suspected hemorrhage, the method comprising:

training a set of models to produce a trained set of models, comprising training the set of models based on:

a first set of images having a first set of labels, the first set of labels indicating a presence of a hemorrhage;

a second set of images having a second set of labels, the second set of labels indicating an absence of a hemorrhage; and

a third set of images having a third set of labels, the third set of labels indicating a false positive detection of a hemorrhage by a trained model;

receiving a fourth set of images associated with a patient;

processing the fourth set of images with the trained set of models; and

triggering an action in response detecting the suspected hemorrhage based on processing the fourth set of images, the action operable to improve an efficiency of triage of the patient.

15. The method of claim 14 , wherein the fourth set of images is imaged at a first point of care, and wherein the action comprises an initiation of a transfer of the patient to a second point of care, the second point of care remote from the first point of care.

16. The method of claim 15 , wherein the action is triggered automatically.

17. The method of claim 14 , wherein the action triggers transmission of a notification to a specialist located at a 2 nd point of care remote from a 1 st point of care at which the patient is located, wherein an input from the specialist in response to the notification triggers an assignment of treatment of the patient to the specialist.

18. The method of claim 14 , wherein the third set of images is a subset of the first set of images, and wherein the third set of labels replaces the first set of labels originally determined for the subset of the first set of images.

19. The method of claim 18 , wherein training the set of models comprises a first training process, wherein in the first training process, the set of models is trained based on the first and second sets of images, wherein in the second training process, the set of models is trained based on the third set of images, and wherein the second training process is performed after the first training process.

20. The method of claim 14 , wherein processing the fourth set of images comprises determining a severity score associated with the suspected hemorrhage, wherein the action is determined at least in part based on the severity score.

21. The method of claim 1 , further comprising, prior to processing the set of images, evaluating one or more exclusion criteria based on the set of images, comprising eliminating one or more categories of false positives associated with hemorrhage determinations based on the set of images.