Computer vision-based claims processing

Abstract

Systems and methods are described for automatically evaluating a claim submitted to an insurance carrier. Claim information and at least one image associated with the claim may be received, where the image has been submitted to a carrier as supporting evidence of a service performed by a submitter of the claim. The system may provide image data and other claim information from the submitted claim as input to machine learning models configured to identify whether the image data, such as a radiograph, supports the other data in the claim submission, such as a treatment code.

1 - 19 . (canceled) 
     
     
         20 . A computer-implemented method comprising:
 as implemented by one or more computing devices configured with specific executable instructions,
 obtaining, for a first insurance claim, claim information and at least a first radiograph image associated with the first insurance claim, wherein the first radiograph image comprises supporting evidence of a service indicated in the claim information as having been performed by a healthcare provider, wherein the claim information includes a first treatment code associated with the service; 
 initiating execution of one or more machine learning models, wherein input features provided to at least one of the one or more machine learning models are based at least in part on image data of the first radiograph image, wherein at least one of the one or more machine learning models are trained to detect one or more dental pathologies, anatomies, restorations or anomalies present in the image data; 
 determining machine learning results based on output of the one or more machine learning models, wherein the machine learning results are based at least in part on one or more dental pathologies, anatomies, conditions, restorations or anomalies detected to be present in the first radiograph image; 
 retrieving stored information that maps or associates at least a first condition to the first treatment code, wherein the stored information represents claim criteria for claim submissions that include the first treatment code; 
 determining whether the claim information and the machine learning results meet the claim criteria for claim submissions that include the first treatment code, wherein determining whether the machine learning results meet the claim criteria is based at least in part on a determination of whether the first condition indicated in the stored information is consistent with the machine learning results; and 
 based at least in part on the determining whether the claim information and the machine learning results meet the claim criteria, generating at least one of (a) an approval or denial of the first insurance claim, (b) a recommendation to approve or deny the first insurance claim, or (c) user interface data that enables a user to review whether to approve or deny the first insurance claim. 
   
     
     
         21 . The computer-implemented method of  claim 20 , wherein the stored information representing the claim criteria includes both (a) a first rule that applies to results of an automated image analysis of a radiograph and (b) a second rule that applies to patient history data. 
     
     
         22 . The computer-implemented method of  claim 20  further comprising extracting the first image and a plurality of additional images associated with the first insurance claim from a composite image submitted by the healthcare provider as supporting evidence for the first insurance claim. 
     
     
         23 . The computer-implemented method of  claim 20 , wherein the determination of whether the first condition indicated in the stored information is consistent with the machine learning results comprises a determination that the first condition was detected to be present in the first radiograph image by at least one of the one or more machine learning models. 
     
     
         24 . The computer-implemented method of  claim 20 , wherein the one or more machine learning models include at least two different convolutional neural networks that are each trained to identify a different dental pathology. 
     
     
         25 . The computer-implemented method of  claim 20  further comprising:
 generating a confidence score associated with the first insurance claim, wherein the confidence score represents a confidence that the first insurance claim should be approved; 
 determining that the confidence score meets an approval threshold; and 
 based on the confidence score meeting the approval threshold, sending, via an application programming interface (API) to a computing system associated with a first carrier, an indication or recommendation that the first insurance claim be approved by the first carrier. 
 
     
     
         26 . The computer-implemented method of  claim 20 , wherein the stored information that maps or associates at least the first condition to the first treatment code comprises a rule set. 
     
     
         27 . The computer-implemented method of  claim 20 , wherein the stored information that maps or associates at least the first condition to the first treatment code comprises a lookup table. 
     
     
         28 . The computer-implemented method of  claim 20  further comprising:
 generating a confidence score associated with the first insurance claim, wherein the confidence score represents a confidence that the first insurance claim should be approved; 
 determining that the confidence score meets a denial threshold; and 
 based on the confidence score meeting the denial threshold, sending, via an application programming interface (API) to a computing system associated with a first carrier, an indication or recommendation that the first insurance claim be denied by the first carrier. 
 
     
     
         29 . A computer-readable, non-transitory storage medium storing computer executable instructions that, when executed by one or more computer systems, configure the one or more computer systems to perform operations comprising:
 obtaining, for a first insurance claim, claim information and at least a first radiograph image associated with the first insurance claim, wherein the first radiograph image has been indicated by a healthcare provider to comprise supporting evidence of a service indicated in the claim information as having been performed by the healthcare provider, wherein the claim information includes a first treatment code associated with the service;   initiating execution of one or more machine learning models, wherein input features provided to at least one of the one or more machine learning models are based at least in part on image data of the first radiograph image, wherein at least one of the one or more machine learning models are trained to detect one or more dental pathologies, anatomies, restorations or anomalies present in the image data;   determining machine learning results based on output of the one or more machine learning models, wherein the machine learning results are based at least in part on one or more dental pathologies, anatomies, conditions, restorations or anomalies detected to be present in the first radiograph image;   retrieving stored information that maps or associates at least a first condition to the first treatment code, wherein the stored information represents claim criteria for claim submissions that include the first treatment code;   determining whether the claim information and the machine learning results meet the claim criteria for claim submissions that include the first treatment code, wherein determining whether the machine learning results meet the claim criteria is based at least in part on a determination of whether the first condition indicated in the stored information is consistent with the machine learning results; and   based at least in part on the determining whether the claim information and the machine learning results meet the claim criteria, generating at least one of (a) a recommendation to approve or deny the first insurance claim, or (b) user interface data that enables a user to review whether to approve or deny the first insurance claim.   
     
     
         30 . The computer-readable, non-transitory storage medium of  claim 29 , wherein the operations further comprise generating a user interface that includes information identifying each of a plurality of insurance claims, wherein information associated with the first insurance claim is presented in the user interface with a visual indication of whether the machine learning results meet the claim criteria. 
     
     
         31 . The computer-readable, non-transitory storage medium of  claim 30 , wherein at least a subset of the plurality of insurance claims are further presented in the user interface with corresponding selectable options for a user to indicate approval or denial of individual insurance claims. 
     
     
         32 . A computer system comprising:
 memory; and   a processor in communication with the memory and configured with processor-executable instructions to perform operations comprising:
 obtaining, for a first insurance claim, claim information and at least a first radiograph image associated with the first insurance claim, wherein the first radiograph image has been submitted as supporting evidence of a service indicated in the claim information as having been performed by a healthcare provider, wherein the claim information includes first treatment identification information identifying the service; 
 initiating execution of one or more machine learning models, wherein input features provided to at least one of the one or more machine learning models are based at least in part on image data of the first radiograph image, wherein at least one of the one or more machine learning models are trained to detect one or more dental pathologies, anatomies, restorations or anomalies present in the image data; 
 determining machine learning results based on output of the one or more machine learning models, wherein the machine learning results are based at least in part on one or more dental pathologies, anatomies, conditions, restorations or anomalies detected to be present in the first radiograph image; 
 retrieving stored information that maps or associates at least a first condition to the first treatment identification information, wherein the stored information represents claim criteria for claim submissions that include the first treatment identification information; 
 determining whether the claim information and the machine learning results meet the claim criteria for claim submissions that include the first treatment identification information, wherein determining whether the machine learning results meet the claim criteria is based at least in part on a determination of whether the first condition indicated in the stored information is also indicated in the machine learning results as being present in the first radiograph image, wherein the first condition is one of a dental pathology, anatomy, restoration or anomaly; and 
 based at least in part on the determining whether the claim information and the machine learning results meet the claim criteria, generating at least one of (a) a recommendation to approve or deny the first insurance claim, or (b) user interface data that enables a user to review whether to approve or deny the first insurance claim. 
   
     
     
         33 . The computer system of  claim 32 , wherein the operations further comprise generating a user interface that enables the user to review whether to approve or deny the first insurance claim. 
     
     
         34 . The computer system of  claim 33 , wherein the user interface includes an option to group insurance claims based on at least one similarity between images associated with insurance claims of a group. 
     
     
         35 . The computer system of  claim 33 , wherein the user interface includes, for each of two or more insurance claims, an associated selectable option for the user to flag an individual claim as warranting investigation. 
     
     
         36 . The computer system of  claim 33 , wherein the user interface includes an option to sort a display order of a plurality of insurance claims in the user interface based on associated confidence scores determined based at least in part on output of at least one of the one or more machine learning models. 
     
     
         37 . The computer system of  claim 32 , wherein the computer system implements an application programming interface (API) configured to receive requests from each of a plurality of carriers, wherein the claim information for the first insurance claim is obtained by the computer system from a computing device associated with a first carrier via the API. 
     
     
         38 . The computer system of  claim 37 , wherein the claim criteria is specific to the first carrier. 
     
     
         39 . The computer system of  claim 32 , wherein the operations further include automatically ranking a plurality of insurance claims for manual review based on associated confidence scores determined based at least in part on output of the one or more machine learning models.