Dictation selection

• US 8,032,372 B1
• Filed: 09/13/2005
• Issued: 10/04/2011
• Est. Priority Date: 09/13/2005
• Status: Active Grant

• Current Assignee: DeliverHealth Solutions LLC
• Original Assignee: eScription, Inc. (Microsoft Corporation)
• Inventors: Zimmerman, Roger Scott, Zavaliagkos, George
• Primary Examiner(s): Dorvil; Richemond
• Assistant Examiner(s): Colucci; Michael
• Application Number:
  US11/225,794
• Time in Patent Office: 2,212 Days
• Field of Search: 704/235, 704/275, 704/232, 704/1, 704/256, 704/270, 704/260, 704/257, 704/251, 704/200, 704/2, 382/310, 705/34, 705/3, 381/98
• US Class Current: 704/235
• CPC Class Codes: G06F 40/44 Statistical methods, e.g. probability models
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G16H 10/20 for electronic clinical trials or questionnaires
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 15/00 ICT specially adapted for medical reports, e.g. generation or transmission thereof

Abstract

A computer program product for computing a correction rate predictor for medical record dictations, the computer program product residing on a computer-readable medium includes computer-readable instructions for causing a computer to obtain a draft medical transcription of at least a portion of a dictation, the dictation being from medical personnel and concerning a patient, determine features of the dictation to produce a feature set comprising a combination of features of the dictation, the features being relevant to a quantity of transcription errors in the transcription, analyze the feature set to compute a predicted correction rate associated with the dictation and use the predicted correction rate to determine whether to provide at least a portion of the transcription to a transcriptionist.

Claims

1. A non-transitory computer-readable medium encoded with a computer program product for computing a correction rate predictor for medical record dictations, the computer program product comprising computer-readable instructions for causing a computer to:

initiate automatic speech recognition processing of a dictation to initiate producing a draft medical transcription of at least a portion of the dictation;

determine features of at least a portion of the dictation to produce a feature set comprising a combination of features of the dictation, the features being relevant to a quantity of expected transcription errors in the transcription of the at least a portion of the dictation;

analyze the feature set to compute a predicted correction rate associated with the at least a portion of the dictation;

use the predicted correction rate to determine whether to abandon the automatic speech recognition processing prior to automatic speech recognition processing of all of the dictation;

use the predicted correction rate to determine whether to provide at least a portion of the draft medical transcription of the at least a portion of the dictation to a transcriptionist; and

provide the at least a portion of the draft medical transcription to the transcriptionist if the predicted correction rate is below a threshold.

2. The computer-readable medium of claim 1 wherein the instructions further cause a computer to:

compute an actual correction rate by comparing the draft medical transcription associated with the dictation and a corresponding edited transcription, and counting a number of changes to produce the edited transcription from the draft medical transcription; and

compute a correction rate prediction error by comparing the predicted correction rate to the actual correction rate.

3. The computer-readable medium of claim 2 wherein the instructions further cause a computer to modify the correction rate predictor to reduce the prediction error between the predicted correction rate and the actual correction rate.

4. The computer-readable medium of claim 2 wherein instructions for causing a computer to obtain a draft medical transcription of a dictation cause the computer to produce a draft transcription only if a predicted correction rate associated with the dictation is below a threshold value.

5. The computer-readable medium of claim 4 wherein instructions further cause the computer to compute the threshold value by analyzing a behavior of the correction rate predictor on a subset of dictations that are stored in the database.

6. The computer-readable medium of claim 1 wherein the features include an element selected from the group consisting of:

a background noise measure, signal-to-noise ratio, average spectrum, and overall audio quality measure.

7. The computer-readable medium of claim 1 wherein the instructions for causing the computer to determine features cause the computer to determine features for the entire dictation.

8. The computer-readable medium of claim 1 wherein the instructions for causing the computer to store the predicted correction rate cause the computer to retrieve the correction rate predictor in accordance with at least one of a worktype, a speaker, a speaker-worktype pair, or a specialty associated with the dictation.

9. The computer-readable medium of claim 1, wherein the instructions further cause the computer to use the predicted correction rate to determine whether to provide at least a portion of the draft medical transcription to a transcriptionist by comparing the predicted correction rate to a threshold and determining that the at least a portion of the draft medical transcription is to be provided to a transcriptionist when the predicted correction rate is below the threshold.

10. The computer-readable medium of claim 1, the computer-readable instructions further causing the computer to combine the features into a classifier according to an artificial neural network.

11. The computer-readable medium of claim 3, wherein the computer-readable instructions further cause the computer to weigh more heavily errors from recent transcriptions than from earlier transcriptions.

12. The computer-readable medium of claim 1, wherein the computer-readable instructions further cause the computer to associate a pseudo correction rate with the at least a portion of the dictation for a draft transcript of the at least a portion of the dictation that was not initially edited.

13. The computer-readable medium of claim 1, wherein at least one feature is specific to a selected word in a raw word recognition output.

14. The computer-readable medium of claim 1, wherein the features are structured in a decision tree such that features first analyzed in the tree comprise features that are less computationally expensive to compute than later features in the tree.

15. An automatic transcription system for processing a medical dictation, the system comprising:

at least one processor programmed to;

initiate automatic speech recognition processing of a dictation to initiate producing a draft transcription of at least a portion of the dictation;

compute a correction rate prediction for the at least a portion of the dictation by analyzing a plurality of features of the dictation, the correction rate prediction indicating a quality of the at least a portion of the dictation for production of the draft transcription;

use the correction rate prediction to determine whether to abandon the automatic speech recognition processing prior to automatic speech recognition processing of all of the at least a portion of the dictation; and

produce the draft transcription for editing if and only if the correction rate prediction for the at least a portion of the dictation meets a criterion.

16. The automatic transcription system of claim 15, wherein the at least one processor is further programmed to compare the correction rate prediction to a threshold value to determine whether to produce the draft transcription for the dictation.

17. The automatic transcription system of claim 15, wherein the at least one processor is further programmed to produce a draft transcription for a portion of a dictation based on the correction rate prediction for a portion of the dictation.

18. The automatic transcription system of claim 17, wherein the at least one processor is further programmed to store the draft transcription and the correction rate prediction for the dictation in a database.

19. The automatic transcription system of claim 18 wherein the correction rate prediction stored in the database identifies whether a specific user utilizes the draft transcription.

20. The automatic transcription system of claim 15 wherein the plurality of features comprises a background noise measure, signal-to-noise ratio, average spectrum, and/or overall audio quality measure.

21. A method comprising acts of:

(A) initiating automatic speech recognition processing of a medical dictation to initiate producing a draft medical transcription of at least a portion of the dictation;

(B) determining, by at least one processor, features of at least a portion of the medical dictation to produce a feature set comprising a combination of features of the medical dictation, the features being relevant to a quantity of expected transcription errors in the transcription of the at least a portion of the medical dictation;

(C) analyzing, by the at least one processor, at least part of the feature set to compute a predicted correction rate associated with the at least a portion of the medical dictation;

(D) comparing, by the at least one processor, the predicted correction rate to a threshold;

(E) abandoning the automatic speech recognition processing prior to automatic speech recognition processing of all of the medical dictation when the predicted correction rate meets or exceeds the threshold;

(F) providing, by the at least one processor, at least a portion of the draft medical transcription to a transcriptionist when the predicted correction rate is below the threshold; and

(G) refraining, by the at least one processor, from providing the at least a portion of the draft medical transcription to the transcriptionist for editing when the predicted correction rate meets or exceeds the threshold.

22. The method of claim 21, further comprising an act of:

(H) instructing the transcriptionist to transcribe the at least a portion of the medical dictation without editing the draft medical transcription when the predicted correction rate is at or above the threshold.

23. The method of claim 21 wherein the features include an element selected from the group consisting of:

a background noise measure, signal-to-noise ratio, average spectrum, and overall audio quality measure.

24. The method of claim 21, wherein the act (A) further comprises obtaining the draft medical transcription from an automatic speech recognition system.