Method and system for computer-aided triage

1. A method for determining a specialist for a patient presenting with a stroke symptom at a first point of care, the method comprising:

receiving, at a remote computing system, a data packet from the first point of care, wherein the data packet comprises a set of computed tomography (CT) images and a set of metadata associated with the set of CT images;

processing the data packet at the remote computing system, comprising:

organizing the set of CT images into a series based on the metadata;

identifying soft matter voxels from the series based on a soft matter mask, the soft matter mask comprising a predetermined Hounsfield Unit (HU) threshold;

registering the soft matter voxels to a set of reference CT images, thereby determining a registered set of voxels;

with a feed-forward deep convolutional network, segmenting a large vessel region in the registered set of voxels;

extracting a centerline of the segmented large vessel region; and

determining a length of the segmented large vessel region based on the centerline; and

comparing the centerline length with a predetermined threshold, wherein the predetermined threshold is greater than a corresponding anatomical length;

when the centerline length is less than the predetermined threshold, determining a specialist based on a lookup table;

transmitting a notification and a second data packet comprising a set of compressed images to a user device associated with the specialist; and

determining a next point of care for the patient based on an input from the specialist after viewing the set of compressed images.

2. The method of claim 1 , wherein the method is performed in parallel with a standard radiology workflow at a first point of care, wherein in the standard radiology workflow, the data packet is concurrently reviewed at a radiologist workstation, and wherein the radiologist determines and notifies the specialist based on a visual assessment of the set of computed tomography (CT) images at the workstation.

3. The method of claim 1 , wherein the set of metadata comprise an identifier associated with the first point of care, wherein the lookup table correlates the specialist with the identifier, and wherein the lookup table is determined based on a location of the first point of care and a location of the specialist.

4. The method of claim 1 , further comprising determining a special condition, wherein when the special condition is determined, the method further comprises transmitting the second data packet and the notification to the specialist.

5. The method of claim 4 , wherein the special condition comprises at least one of a partial occlusion and a fetal origin posterior cerebral artery.

6. The method of claim 5 , wherein determining a partial occlusion comprises:

identifying centerline voxels, lying along the large vessel centerline, within the segmented large vessel region;

detecting the partial occlusion when a HU value of the centerline voxels differs more than a threshold value from HU values of adjacent voxels.

7. The method of claim 5 , wherein determining the fetal origin posterior cerebral artery comprises:

determining an orientation of the centerline; and

detecting the fetal origin posterior cerebral artery when the orientation is more posterior than distal.

8. The method of claim 1 , wherein determining the set of compressed images comprises, at the remote computing system:

determining a portion of the first set of images corresponding to air; and

setting each of a set of voxels corresponding to the portion to a single predetermined value.

9. The method of claim 1 , wherein processing the data packet at the remote computing system is initiated after a predetermined time period has passed after receiving the data packet at the remote computing system.

10. The method of claim 1 , wherein determining the second data packet comprises removing patient information from the set of metadata of the first data packet.

11. A method for computer-aided triage, the method comprising, at a processing system:

receiving a data packet comprising a set of computed tomography (CT) images of a brain of the patient sampled at the first point of care, wherein the data packet is concurrently sent to a standard radiology workflow operating in parallel with the method;

determining a large vessel region from the set of CT images;

extracting a large vessel centerline from the large vessel region;

calculating a centerline length of the large vessel centerline; and

comparing the centerline length with a predetermined threshold;

detecting a large vessel occlusion when the centerline length is less than the predetermined threshold;

in response to large vessel occlusion detection:

presenting a notification on a mobile device associated with a specialist from the standard radiology workflow, the specialist associated with a second point of care;

displaying a compressed version of the set of CT images on the mobile device in response to interaction with the notification; and

selecting one of a set of next points of care for the patient, the set of next points of care comprising at least the first and second points of care, based on an input from the specialist after viewing the compressed CT images.

12. The method of claim 11 , wherein the large vessel region comprises a middle cerebral artery M1 (MCA-M1) region and a terminal internal carotid artery (t-ICA) region.

13. The method of claim 11 , further comprising:

at a medical routing system connected to the CT scanner, intercepting the set of CT images en route to the standard radiology workflow; and

transmitting the set of CT images from the medical routing system to a remote computing system, wherein the processing system is the remote computing system.

14. The method of claim 11 , wherein determining a large vessel region from the set of CT images is initiated after a predetermined time period has passed after receiving the data packet at the processing system.

15. The method of claim 11 , wherein determining the large vessel region comprises:

extracting soft matter regions from the set of CT images;

registering the soft matter regions to a reference series having a target vessel region;

after registration, probabilistically segmenting the soft matter regions falling within the target vessel region; and

applying a binary mask to the probabilistically segmented soft matter regions falling within the target region, wherein voxels having probabilities above a threshold probability are included in the large vessel region.

16. The method of claim 15 , wherein extracting the soft matter regions comprises applying a soft matter mask to the set of CT images, the soft matter mask comprising a predetermined Hounsfield Unit threshold.

17. The method of claim 11 , wherein determining the large vessel region comprises segmenting the large vessel region from the set of CT images with a feed-forward deep convolutional network using three-dimensional convolutions.

18. The method of claim 11 , wherein the predetermined threshold is greater than a standard anatomical length for a large vessel associated with the large vessel region, thereby increasing an average occurrence of false positives in the method.

19. The method of claim 11 , further comprising:

detecting a partial occlusion, comprising:

identifying centerline voxels, lying along the large vessel centerline, within the large vessel region;

detecting the partial occlusion when a HU value of the centerline voxels differs more than a threshold value from HU values of adjacent voxels adjacent the centerline voxels; and

determining the specialist and presenting the notification on the mobile device when the partial occlusion is detected.

20. The method of claim 11 , further comprising:

extending the large vessel centerline based on HU values of end voxels, adjacent centerline ends of the large vessel centerline, to generate an extended centerline, wherein the centerline length is calculated from the extended centerline;

detecting a fetal posterior cerebral artery (fetal PCA), comprising:

determining an orientation of centerline extension; and

detecting the fetal PCA when the orientation is more posterior than distal; and

determining the specialist and presenting the notification on the mobile device when the fetal PCA is detected.