Granted Patent B1

US 12,596,703 · App. 18/963,603 · Granted Apr 7, 2026

Inventors: Blake Browder (Dallas, TX); Joy Figarsky (Little Rock, AR)

Assignee: BH Operations, LLC

G06F16/2365G06F16/254G06Q40/0841

Abstract

Claims (64)

1 . An apparatus for adaptive data conversion, the apparatus comprising:

at least a processor; and

a memory communicatively connected to the at least a processor, wherein the memory contains instructions configuring the at least a processor to:

access an interactive data structure comprising a plurality of input elements in a plurality of first formats, wherein the plurality of input elements comprises user-selected output data associated with each of the plurality of input elements, wherein at least one input element of the plurality of input elements comprises an image of a billable charge from a medical facility and information on the medical facility;

identify an input attribute of each of the plurality of input elements utilizing an attribute machine-learning model which comprises:

receiving attribute training data, wherein the attribute training data correlates a plurality of exemplary input element data to a plurality of exemplary input attribute data;

training the attribute machine-learning model using the attribute training data; and

identifying the input attribute using the trained attribute machine-learning model, wherein the input attribute for the at least one input element comprises an image-based attribute identifying a diagnostic category of the billable charge;

convert each of the plurality of input elements in the plurality of first formats to a second format as a function of the input attribute, wherein converting each of the plurality of input elements comprises:

selecting one conversion pathway from a plurality of conversion pathways for each of the plurality of input elements as a function of the input attribute and the plurality of first formats, wherein the conversion pathway comprises a format specific conversion sequence; and

converting the plurality of input elements using the selected conversion pathway, wherein the format specific conversion sequence comprises an optical character recognition (OCR) to convert at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into machine-encoded text, wherein converting at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into the machine-encoded text comprises converting an image of text in at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into the machine-encoded text and further comprises:

pre-processing image components of the image of text by de-skewing at least one of the image components by applying a transform operation to the at least one of the image components; and

implementing an OCR algorithm comprising a matrix matching process by comparing pixels of at least one of the pre-processed images components to pixels of a stored glyph on a pixel-by-pixel basis;

generate output data as a function of the converted plurality of input elements, including at least the text-converted image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements converted into the machine-encoded text by the OCR, in the second format as converted by at least the OCR, wherein generating the output data comprises:

generating output training data, wherein the output training data comprises historical output data;

training an output machine-learning model using the output training data; and

generating the output data using the trained output machine-learning model, wherein the output data comprises a monetary value that represents an amount payable by an insurance company to the medical facility based on the billable charge from the medical facility and the identification of the diagnostic category of the billable charge; and

update the interactive data structure as a function of the output data.

2 . The apparatus of claim 1 , wherein the format specific conversion sequence comprises a plurality of data converting machine-learning models, wherein a second data converting machine-learning model receives an output of a first data converting machine-learning model as an input.

3 . The apparatus of claim 1 , wherein the format specific conversion sequence comprises the OCR and a large language model, wherein the large language model receives an output of the OCR as an input.

4 . The apparatus of claim 1 , wherein the format specific conversion sequence comprises an encoder.

5 . The apparatus of claim 1 , wherein generating the output training data comprises:

generating cohort training data, wherein the cohort training data comprises exemplary input elements correlated to exemplary output cohorts;

training a cohort classifier using the cohort training data; and

classifying the plurality of input elements to one or more output cohorts using the trained cohort classifier.

6 . The apparatus of claim 5 , wherein generating the output training data comprises:

generating classification training data, wherein the classification training data comprises exemplary output cohorts correlated to exemplary training data;

training a training data classifier using the classification training data; and

updating the output training data as a function of the one or more output cohorts using the training data classifier.

7 . The apparatus of claim 1 , wherein generating the output data comprises generating a user interface displaying the output data on a user device.

8 . The apparatus of claim 7 , wherein generating the output training data comprises:

receiving output feedback for the output data; and

updating the output training data as a function of the output feedback by adding correlations between exemplary input elements and exemplary output data to the historical output data.

9 . A method for adaptive data conversion, the method comprising:

accessing, using at least a processor, an interactive data structure comprising a plurality of input elements in a plurality of first formats, wherein the plurality of input elements comprises user-selected output data associated with each of the plurality of input elements, wherein at least one input element of the plurality of input elements comprises an image of a billable charge from a medical facility and information on the medical facility;

identifying, using the at least a processor, an input attribute of each of the plurality of input elements utilizing an attribute machine-learning model which comprises:

receiving attribute training data, wherein the attribute training data correlates a plurality of exemplary input element data to a plurality of exemplary input attribute data;

training the attribute machine-learning model using the attribute training data; and

identifying the input attribute using the trained attribute machine-learning model, wherein the input attribute for the at least one input element comprises an image-based attribute identifying a diagnostic category of the billable charge;

converting, using the at least a processor, each of the plurality of input elements in the plurality of first formats to a second format as a function of the input attribute, wherein converting each of the plurality of input elements comprises:

selecting one conversion pathway from a plurality of conversion pathways for each of the plurality of input elements as a function of the input attribute and the plurality of first formats, wherein the conversion pathway comprises a format specific conversion sequence; and

converting the plurality of input elements using the selected conversion pathway, wherein the format specific conversion sequence comprises an optical character recognition (OCR) to convert at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into machine-encoded text, wherein converting at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into the machine-encoded text comprises converting an image of text in at least the image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements into the machine-encoded text and further comprises:

pre-processing image components of the image of text by de-skewing at least one of the image components by applying a transform operation to the at least one of the image components; and

implementing an OCR algorithm comprising a matrix matching process by comparing pixels of at least one of the pre-processed images components to pixels of a stored glyph on a pixel-by-pixel basis;

generating, using the at least a processor, output data as a function of the converted plurality of input elements, including at least the text-converted image of the billable charge from the medical facility and the information on the medical facility of the plurality of input elements converted into the machine-encoded text by the OCR, in the second format as converted by at least the OCR, wherein generating the output data comprises:

generating output training data, wherein the output training data comprises historical output data;

training an output machine-learning model using the output training data; and

generating the output data using the trained output machine-learning model, wherein the output data comprises a monetary value that represents an amount payable by an insurance company to the medical facility based on the billable charge from the medical facility and the identification of the diagnostic category of the billable charge; and

updating, using the at least a processor, the interactive data structure as a function of the output data.

10 . The method of claim 9 , wherein the format specific conversion sequence comprises a plurality of data converting machine-learning models, wherein a second data converting machine-learning model receives an output of a first data converting machine-learning model as an input.

11 . The method of claim 9 , wherein the format specific conversion sequence comprises the OCR and a large language model, wherein the large language model receives an output of the OCR as an input.

12 . The method of claim 9 , wherein the format specific conversion sequence comprises an encoder.

13 . The method of claim 9 , wherein generating the output training data comprises:

generating cohort training data, wherein the cohort training data comprises exemplary input elements correlated to exemplary output cohorts;

training a cohort classifier using the cohort training data; and

classifying the plurality of input elements to one or more output cohorts using the trained cohort classifier.

14 . The method of claim 13 , wherein generating the output training data comprises:

generating classification training data, wherein the classification training data comprises exemplary output cohorts correlated to exemplary training data;

training a training data classifier using the classification training data; and

updating the output training data as a function of the one or more output cohorts using the training data classifier.

15 . The method of claim 9 , wherein generating the output data comprises generating a user interface displaying the output data on a user device.

16 . The method of claim 15 , wherein generating the output training data comprises:

receiving output feedback for the output data; and

updating the output training data as a function of the output feedback by adding correlations between exemplary input elements and exemplary output data to the historical output data.

Assignments (2)

References Cited (15)