Method and system for the computer-assisted implementation of radiology recommendations

1. A method for computer-assisted implementation of a set of follow-up recommendations associated with a set of patients, the method comprising:

for each patient of the set of patients:

receiving a radiology report associated with the patient;

automatically processing the radiology report, comprising:

determining a set of incidental findings in the radiology report, wherein determining the set of incidental findings comprises:

processing the radiology report with a first set of trained neural network models to detect a set of findings, the first set of trained neural network models comprising a first multi-transformer model configured to implement parallelization;

processing the set of findings with a first set of rule-based logic to determine a first set of features, wherein the first set of rule-based logic is configured to determine a context associated with the detected set of findings, wherein determining the context comprises checking for inclusion of negation words proximal to the set of findings;

determining the set of incidental findings based on the first set of features;

determining a set of follow-up recommendations associated with the set of incidental findings in the radiology report, wherein determining the set of follow-up recommendations comprises:

processing the radiology report with a second set of trained neural network models to locate a set of recommendation candidate sentences, the second set of trained neural network models comprising a second multi-transformer model configured to implement parallelization;

processing each of the set of recommendation candidate sentences with a second set of rule-based logic to determine:

 a second set of features for the set of recommendation candidate sentences; and

 a level of completion for each of the set of recommendation candidate sentences;

determining the set of follow-up recommendations based on at least one of the second set of features and the levels of completion;

for the set of patients, initiating a set of follow-up actions in response to automatically processing the radiology report, wherein automatically initiating the set of follow-up actions comprises, automatically:

scheduling a set of follow-up imaging processes for a first subset of the set of patients;

messaging a set of primary care physicians associated with a second subset of the set of patients; and

upon detecting that each patient of a third subset of the set of patients does not have a primary care physician, initiating the assignment of a second set of primary care physicians to the third subset of patients; and

automatically populating a set of worklists based on the set of follow-up actions.

2. The method of claim 1 , further comprising comparing the set of incidental findings with the set of follow-up recommendations to check for a set of missing follow-up recommendations.

3. The method of claim 2 , wherein automatically populating the set of worklists comprises:

in response to determining that there are no missing follow-up recommendations for a patient of the set of patients, triggering an assignment of the patient to a first worklist of the set of worklists;

in response to determining at least one of: the presence of at least one missing follow-up recommendation and a level of completion below a threshold associated with a follow-up recommendation, performing at least one of:

triggering an assignment of the patient to a second worklist; and

automatically adding supplemental text to the radiology report.

4. The method of claim 1 , wherein processing the findings section further comprises evaluating a decision tree based on the first set of features, wherein determining the set of incidental findings comprises categorizing a portion of the set of findings as the set of incidental findings based on evaluating the decision tree.

5. The method of claim 4 , wherein categorizing the portion further comprises assigning a particular type of incidental finding to the findings text.

6. The method of claim 1 , wherein the first set of features comprises a size measurement.

7. The method of claim 1 , wherein the set of findings is detected based at least in part on a predetermined set of pathology keywords.

8. The method of claim 1 , wherein automatically organizing each of the first and second worklists comprises sorting the patients based on a progress metric associated with the follow-up recommendation.

9. The method of claim 8 , wherein the set of patients is further sorted based on a type associated with each of the set of incidental findings.

10. The method of claim 9 , wherein the set of patients is further sorted based on risk metric calculated for the patient, wherein the risk metric is calculated based on at least one of: demographic information associated with the patient and a medical history associated with the patient.

11. The method of claim 1 , wherein the method is performed absent of radiologist input.

12. The method of claim 1 , further comprising, in response to detecting that a follow-up recommendation of the set of follow-up recommendations has a completion level below a predetermined threshold, automatically adding supplemental text to the radiology report.

13. The method of claim 12 , further comprising determining the supplemental text, wherein determining the supplemental text comprises referencing a lookup table.

14. The method of claim 13 , wherein the lookup table is determined based on a set of radiology consensus guidelines.

15. The method of claim 1 , wherein each of the first and second set of trained neural network models comprises a neural network.

16. The method of claim 1 , wherein determining the set of findings comprises locating a predetermined findings section of the radiology report.

17. The method of claim 1 , wherein the first and second multi-transformer models are not required to process data in the radiology report in order.