Attention focusing for multiple patients monitoring

1. A system for monitoring a room of a patient, comprising:

a server;

a monitoring device placed in the room of the patient,

wherein the monitoring device is configured to:

obtain an image stream of at least a part of the room, wherein the image stream comprises a first image and a second image that is subsequent to the first image in the image stream;

use a machine learning based classification to identify first states found in the first image;

use the machine learning based classification to identify second states found in the second image;

compare the first states to the second states to identify a first active state, wherein the first active state corresponds to at least one difference between the first states and the second states; and

in response to identifying the first active state, transmit a first notification to the server, wherein the first notification comprises the second image; and

wherein the server is configured to:

responsive to receiving the first notification, transmit the second image to a user device; and

the user device configured to:

display, on the user device, a plurality of images wherein each image of the plurality of images corresponds to a different room;

update, in response to receiving the second image from the server, the plurality of images so that the second image replaces at least one image of the plurality of images;

receive, from a user, a request to display the image stream on the user device;

transmit, to the server, a request to provide the image stream to the user device; and

replace, in response to receiving the image stream from the server, the second image with the image stream.

2. The system of claim 1 ,

wherein the first notification comprises the first active state, and

wherein the server is further configured to:

transmit, to the user device, the first active state.

3. The system of claim 1 ,

wherein the image stream further comprises a third image and a fourth image, and

wherein the monitoring device is further configured to:

identify a second active state in the third image;

record a time of identifying the second active state; and

responsive to identifying the second active state in the fourth image and the second active state persisting for at least a threshold duration of time, transmit a second notification to the server, wherein the second notification comprises an indication of the second active state.

4. The system of claim 1 ,

wherein the image stream further comprises a third image subsequent to the second image;

wherein the monitoring device is further configured to:

identify third states based on the third image;

responsive to determining that the third states do not differ from the second states, not transmitting the third image to the server; and

responsive to determining that the third states differ from the second states, transmitting a second notification to the server, wherein the second notification comprises the third image.

5. A method for monitoring a room of a patient, comprising:

obtaining, by a monitoring device, an image stream of at least a part of the room of the patient;

obtaining, by the monitoring device, first states of the part of the room of the patient, wherein the first states are determined using a machine learning based classification applied to a first image of the image stream;

obtaining, by the monitoring device, second states of the part of the room of the patient, wherein the second states are determined using the machine learning based classification applied to a second image of the image stream;

responsive to identifying, by the monitoring device, a state difference between the first states and the second states, transmitting a first notification to a server, wherein the first notification comprises the second image and the state difference;

obtaining, by the monitoring device and from a third image of the image stream, third states of the part of the room of the patient;

responsive to determining that the third states include a monitored condition, recording a time associated with the monitored condition;

obtaining, by the monitoring device and from a fourth image of the image stream, fourth states of the part of the room of the patient; and

responsive to determining that the fourth states include the monitored condition and that the monitored condition persisted for a threshold duration of time, transmitting a second notification to the server, wherein the second notification comprises an indication of the monitored condition and the threshold duration of time of the monitored condition.

6. The method of claim 5 , wherein the machine learning based classification used to determine the first states and the second states employs a multi-label image classification model.

7. The method of claim 6 , wherein the first states and the second states each comprises at least one of an activity of the patient and a state of the patient.

8. The method of claim 7 , wherein the state of the patient comprises respective states indicating whether the patient is sitting, whether the patient is lying down, whether the patient is getting out of a bed, whether the patient is standing, whether the patient is walking, and whether the patient is on a floor.

9. The method of claim 7 , wherein the activity of the patient comprises respective states indicating whether the patient eating, whether the patient is drinking, and whether the patient is waving.

10. The method of claim 5 , further comprising:

obtaining, by the monitoring device and from a fifth image of the image stream, fifth states of the part of the room of the patient, wherein the fifth image is subsequent to the fourth image in the image stream; and

responsive to the fifth states not including the monitored condition, resetting the time associated with the monitored condition.

11. An apparatus for monitoring a monitored environment, comprising:

a camera;

a memory; and

a processor configured to execute instructions stored in the memory to:

obtain, from the camera, an image stream of at least a part of the monitored environment;

apply a multi-label image classification to a first image of the image stream to obtain first classification labels;

apply the multi-label image classification to a second image of the image stream to obtain second classification labels;

identify state differences by comparing the first classification labels to the second classification labels, wherein to identify the state differences comprises to:

obtain, from the first image of the image stream, first states of the monitored environment;

responsive to determining that the first states include a monitored condition, record a time associated with the monitored condition; and

obtain, from the second image of the image stream, second states of the monitored environment; and

responsive to determining that the second states include the monitored condition and that the monitored condition persisted for a threshold duration of time, transmit a notification to a server, wherein the notification comprises an indication of the monitored condition and the threshold duration of time of the monitored condition.

12. The apparatus of claim 11 , wherein the monitored environment is a room of a patient.

13. The apparatus of claim 12 , wherein the monitored condition indicates whether the patient has not moved in at least the threshold duration of time.

14. The apparatus of claim 12 , wherein the monitored condition indicates whether the patient has not been detected in an image in at least the threshold duration of time.

15. The system of claim 1 , wherein the user device is further configured to display the second image highlighted with respect to at least one other image of the plurality of images.

16. The system of claim 15 , wherein the user device is further configured to display the second image with an indication of the first active state overlaying the second image.