Automated visual reporting technique for medical imaging processing system

1. A method comprising:

receiving one or more medical images for a patient;

determining a location corresponding to a part of a human body within the received one or more medical images using a natural language processing model (NLP) trained to extract information from a medical report associated with the received one or more medical images, the medical reporting comprising text describing the location;

applying an image recognition model to determine a likelihood that an object present in the received one or more medical images is an abnormality based on the determined location within the received one or more medical images, wherein the image recognition model is trained to detect foreign objects that are not expected to be present at the part of the human body corresponding to the location using a training data set of medical images containing objects;

determining, by the image recognition model, a type of object present in the received one or more medical images;

identifying an abnormality detection model corresponding to the determined type of object, the abnormality detection model trained to detect one or more types of abnormalities for the determined type of object depicted using a training dataset of medical images labeled with the type of object;

applying the abnormality detection model to identify an abnormality in the received one or more medical images, the abnormality detection model determining a likelihood score that the abnormality is present in the received one or more medical images;

responsive to determining that the likelihood score is higher than a threshold value, generating an automated report, the automated report indicating a type of abnormality identified in the received one or more medical images; and

providing the automated report to a viewing user.

2. The method of claim 1 , further comprising:

receiving, from a reviewing user, one or more comments for a medical image of the one or more medical images, each comment corresponding to an observation by the reviewing user of the medical image;

applying the NLP model to the received one or more comments to identify a potential abnormality observed by the reviewing user;

identifying one or more abnormality detection models corresponding to the identified potential abnormality; and

applying the one or more abnormality detection models to the received one or more medical images.

3. The method of claim 1 , wherein the automated report further indicates the determined location in the received one or more medical images corresponding to the identified abnormality.

4. The method of claim 1 , further comprising:

receiving a modification for the automated report from the viewing user; and

retraining the abnormality detection model based on the received modification.

5. The method of claim 1 , further comprising:

receiving, from a reviewing user, one or more comments for each of the received one or more medical images, the one or more comments associated with a portion of the medical image, each comment corresponding to an observation by the reviewing user on a corresponding portion of the medical image;

identifying, from each comment, the portion of the medical image associated with the comment; and

generating an annotated image by overlaying each comment and an indication of the identified portion of the medical image associated with the comment.

6. The method of claim 5 , wherein identifying the portion of the medical image associated with the comment comprises:

identifying, from the one or more comments, a description of the location of the object within the medical image.

7. The method of claim 6 , wherein identifying the description of the of the location within the medical image comprises:

applying the NLP model to the one or more comments for the medical image.

8. The method of claim 5 , wherein identifying the portion of the medical image associated with the comment further comprises:

applying a trained image recognition model to identify one or more objects included in a description of the location of the object within the medical image.

9. The method of claim 8 , further comprising:

training the image recognition model comprising:

receiving a training set, the training set including a plurality of labeled images, each labeled image of the plurality of labeled images including a description of location within the image and an indication of the location corresponding to the description, and

training the image recognition model using a machine learning algorithm and the received training set.

10. The method of claim 1 , wherein training the NLP model further comprises:

receiving a training set, the training set including a plurality of labeled comments, each labeled comment of the plurality of labeled comments indicating a portion of the labeled comment corresponding to a description of a location, and

training the NLP model, using a machine learning algorithm and the received training set, to classify portions of comments.

11. A non-transitory computer readable medium configured to store instructions, the instructions when executed by a processor cause the processor to:

receive one or more medical images for a patient;

determine a location corresponding to a part of a human body within the received one or more medical images using a natural language processing model (NLP) trained to extract information from a medical report associated with the received one or more medical images, the medical reporting comprising text describing the location;

apply an image recognition model to determine a likelihood that an object present in the received one or more medical images is an abnormality based on the determined location within the received one or more medical images, wherein the image recognition model is trained to detect foreign objects that are not expected to be present at the part of the human body corresponding to the location using a training data set of medical images containing objects;

determine, by the image recognition model, a type of object present in the received one or more medical images;

identify an abnormality detection model corresponding to the determined type of object, the abnormality detection model trained to detect one or more types of abnormalities for the determined type of object depicted using a training dataset of medical images labeled with the type of object;

apply the abnormality detection model to identify an abnormality in the received one or more medical images, the abnormality detection model determining a likelihood score that the abnormality is present in the received one or more medical images;

responsive to determining that the likelihood score is higher than a threshold value, generate an automated report, the automated report indicating a type of abnormality identified in the received one or more medical images; and

provide the automated report to a viewing user.

12. The non-transitory computer readable medium of claim 11 , further comprising instructions that cause the processor to:

receive, from a reviewing user, one or more comments for a medical image of the one or more medical images, each comment corresponding to an observation by the reviewing user of the medical image;

apply the NLP model to the received one or more comments to identify a potential abnormality observed by the reviewing user;

identify one or more abnormality detection models corresponding to the identified potential abnormality; and

apply the one or more abnormality detection models to the received one or more medical images.

13. The non-transitory computer readable medium of claim 11 , wherein the automated report further indicates the determined location in the received one or more medical images corresponding to the identified abnormality.

14. The non-transitory computer readable medium of claim 11 , wherein the instructions for further cause the processor to:

receive a modification for the automated report from the viewing user; and

retrain the abnormality detection model based on the received modification.

15. A system comprising:

a processor; and

a non-transitory computer readable medium configured to store instructions, the instructions when executed by a processor cause the processor to:

receive one or more medical images for a patient;

determine a location corresponding to a part of a human body within the received one or more medical images using a natural language processing model (NLP) trained to extract information from a medical report associated with the received one or more medical images, the medical reporting comprising text describing the location;

apply an image recognition model to determine a likelihood that an object present in the received one or more medical images is an abnormality based on the determined location within the received one or more medical images, wherein the image recognition model is trained to detect foreign objects that are not expected to be present at the part of the human body corresponding to the location using a training data set of medical images containing objects;

determine a type of object present in the received one or more medical images;

identify an abnormality detection model corresponding to the determined type of object, the abnormality detection model trained to detect one or more types of abnormalities for the determined type of object depicted using a training dataset of medical images labeled with the type of object;

apply the abnormality detection model to identify an abnormality in the received one or more medical images, the abnormality detection model determining a likelihood score that the abnormality is present in the received one or more medical images;

responsive to determining that the likelihood score is higher than a threshold value, generate an automated report, the automated report indicating a type of abnormality identified in the received one or more medical images; and

provide the automated report to a viewing user.

16. The system of claim 15 , further comprising instructions that cause the processor to:

receive, from a reviewing user, one or more comments for a medical image of the one or more medical images, each comment corresponding to an observation by the reviewing user of the medical image;

apply the NLP model to the received one or more comments to identify a potential abnormality observed by the reviewing user;

identify one or more abnormality detection models corresponding to the identified potential abnormality; and

apply the one or more abnormality detection models to the received one or more medical images.

17. The system of claim 15 , wherein the instructions further cause the processor to:

receive a modification for the automated report from the viewing user; and

retrain the abnormality detection model based on the received modification.