Systems and methods for improved optical character recognition of health records
Systems and methods to improve the optical character recognition of records, and in particular health records, are provided. An image of a medical record is received, and an initial optical image recognition (OCR) on the image is performed to identify text information. The OCR signal quality may be measured, and areas of insufficient OCR signal quality may be isolated. The signal quality is determined by a weighted average of semantic analysis of the resulting text, and/or OCR accuracy measures. The OCR process may be repeated on the isolated regions of lower signal quality, each time using a different OCR transform, until all regions are completed with a desired degree of signal quality (accuracy). All the regions of the document may then be recompiled into a single document for outputting.
1. A computerized method for generating machine readable documents, the method comprising:
receiving an image;
segmenting the image into multiple continuous portions based upon containing similar attributes;
selecting a first optical image recognition (OCR) transform from a plurality of OCR transforms comprising a weighted average of semantic analysis of text information, a calculated figure of merit equation between characters in identified text information and an example font, and maximizing entropy within identified text information area in a linear model, to analyze each of the continuous portions;
for each of the continuous portions, performing an initial OCR on the image to identify text information using the selected first OCR transform;
measuring an OCR signal quality of the initial OCR;
when the OCR signal quality of the initial OCR of a corresponding continuous portion is sufficient, outputting the text information of the initial OCR; and
when the OCR signal quality of the initial OCR is insufficient, the method further comprises:
selecting a second OCR transform from remaining of the plurality of OCR transforms to perform a second OCR to identify a text information on the corresponding continuous portion;
measuring a signal quality of the second OCR on the corresponding continuous portion; and
if the signal quality of the second OCR of the corresponding continuous portion is sufficient, outputting the text information from the second OCR.
2. The method of claim 1 , wherein the OCR signal quality is determined by OCR accuracy measures.
3. The method of claim 1 , further comprising identifying regions of continuous OCR signal quality within the OCR image.
4. The method of claim 1 , further comprising identifying the transform for each region of contiguous signal that has the highest OCR quality signal.
5. The method of claim 4 , further comprising applying the identified transform to each region of continuous signal to generate a plurality of OCR portions.
6. The method of claim 5 , further comprising recompiling the plurality of OCR portions into a single document.
7. The method of claim 6 , further comprising outputting the single document.
8. A computerized system for generating machine readable documents comprising:
an optical character recognition (OCR) server configured to receive an image, segment the image into continuous portions based upon containing similar attributes, and for each of the continuous portions, selecting a first optical image recognition (OCR) transform from a plurality of OCR transforms comprising a weighted average of semantic analysis of text information, a calculated figure of merit equation between characters in identified text information and an example font, and maximizing entropy within identified text information area in a linear model, to analyze each of the continuous portions, performing an initial OCR on the image to identify text information using the selected first OCR transform;
a statistical modeling server configured to measure an OCR signal quality of the initial OCR;
when the OCR signal quality of the initial OCR of a corresponding continuous portion is sufficient the OCR server is further configured to output the text information of the initial OCR; and
when the OCR signal quality of the initial OCR is insufficient, the OCR server is further configured to select a second OCR transform from remaining of the plurality of OCR transforms to perform a second OCR to identify a text information on the corresponding continuous portion, measure a signal quality of the second OCR on the corresponding continuous portion, and if the signal quality of the second OCR of the corresponding continuous portion is sufficient, output the text information from the second OCR; and
the OCR server is further configured to aggregate the outputted text information from each of the continuous portions.
9. The system of claim 8 , wherein the OCR signal quality is determined by OCR accuracy measures.
10. The system of claim 8 , wherein the OCR server is further configured to identify portions of continuous OCR signal quality within the OCR image.
11. The system of claim 8 , wherein the statistical modeling server is further configured to identifying the transform for each portion of continuous signal that has the highest quality signal.
12. The system of claim 11 , wherein the OCR server is further configured to apply the identified transform to each portion of continuous signal to generate a plurality of OCR portions.
13. The system of claim 12 , wherein the OCR server is further configured to recompile the plurality of OCR portions into a single document.
14. The system of claim 13 , wherein the OCR server is further configured to output the single document.
15. The method of claim 1 further comprising when the OCR signal quality of each of the continuous portions is sufficient, combining the continuous portions into a single document for outputting.
16. The method of claim 1 further comprising when the OCR signal quality of the first OCR and the OCR signal quality of the subsequent OCR are insufficient, iteratively applying an additional OCR transform to the portion with insufficient OCR signal quality, wherein the additional OCR transform is different from the first OCR transform and the subsequent OCR transform and measuring the OCR signal quality of the continuous portion until a sufficient OCR signal quality for the continuous portion is reached.
17. The method of claim 1 further comprising processing the image to determine which OCR transform to use as the first OCR transform.
18. The method of claim 1 , wherein the OCR signal quality for a first continuous portion of the continuous portions is sufficient based on the first OCR transform and the OCR signal quality for a second continuous portion of the continuous portions is sufficient based on the subsequent OCR transform, wherein the first continuous portion and the second continuous portion are different.
19. The method of claim 1 , wherein the first OCR transform is at least one of a font matching OCR transform, a cryptogram solving OCR, and a statistical machine translation OCR, wherein the subsequent OCR transform is at least one of the font matching OCR transform, the cryptogram solving OCR, and the statistical machine translation OCR, and wherein the first OCR transform and the subsequent OCR transform are different.