Literature DB >> 31317003

Detecting insertion, substitution, and deletion errors in radiology reports using neural sequence-to-sequence models.

John Zech1, Jessica Forde2, Joseph J Titano1, Deepak Kaji3, Anthony Costa3, Eric Karl Oermann3.   

Abstract

BACKGROUND: Errors in grammar, spelling, and usage in radiology reports are common. To automatically detect inappropriate insertions, deletions, and substitutions of words in radiology reports, we proposed using a neural sequence-to-sequence (seq2seq) model.
METHODS: Head CT and chest radiograph reports from Mount Sinai Hospital (MSH) (n=61,722 and 818,978, respectively), Mount Sinai Queens (MSQ) (n=30,145 and 194,309, respectively) and MIMIC-III (n=32,259 and 54,685) were converted into sentences. Insertions, substitutions, and deletions of words were randomly introduced. Seq2seq models were trained using corrupted sentences as input to predict original uncorrupted sentences. Three models were trained using head CTs from MSH, chest radiographs from MSH, and head CTs from all three collections. Model performance was assessed across different sites and modalities. A sample of original, uncorrupted sentences were manually reviewed for any error in syntax, usage, or spelling to estimate real-world proofreading performance of the algorithm.
RESULTS: Seq2seq detected 90.3% and 88.2% of corrupted sentences with 97.7% and 98.8% specificity in same-site, same-modality test sets for head CTs and chest radiographs, respectively. Manual review of original, uncorrupted same-site same-modality head CT sentences demonstrated seq2seq positive predictive value (PPV) 0.393 (157/400; 95% CI, 0.346-0.441) and negative predictive value (NPV) 0.986 (789/800; 95% CI, 0.976-0.992) for detecting sentences containing real-world errors, with estimated sensitivity of 0.389 (95% CI, 0.267-0.542) and specificity 0.986 (95% CI, 0.985-0.987) over n=86,211 uncorrupted training examples.
CONCLUSIONS: Seq2seq models can be highly effective at detecting erroneous insertions, deletions, and substitutions of words in radiology reports. To achieve high performance, these models require site- and modality-specific training examples. Incorporating additional targeted training data could further improve performance in detecting real-world errors in reports.

Entities:  

Keywords:  Radiology; artificial intelligence; machine learning; natural language processing; neural networks (computer)

Year:  2019        PMID: 31317003      PMCID: PMC6603352          DOI: 10.21037/atm.2018.08.11

Source DB:  PubMed          Journal:  Ann Transl Med        ISSN: 2305-5839


  20 in total

1.  Prepopulated radiology report templates: a prospective analysis of error rate and turnaround time.

Authors:  C M Hawkins; S Hall; J Hardin; S Salisbury; A J Towbin
Journal:  J Digit Imaging       Date:  2012-08       Impact factor: 4.056

2.  Error rates in breast imaging reports: comparison of automatic speech recognition and dictation transcription.

Authors:  Sarah Basma; Bridgette Lord; Lindsay M Jacks; Mohamed Rizk; Anabel M Scaranelo
Journal:  AJR Am J Roentgenol       Date:  2011-10       Impact factor: 3.959

3.  Non-clinical errors using voice recognition dictation software for radiology reports: a retrospective audit.

Authors:  Chian A Chang; Rodney Strahan; Damien Jolley
Journal:  J Digit Imaging       Date:  2011-08       Impact factor: 4.056

4.  Voice recognition versus transcriptionist: error rates and productivity in MRI reporting.

Authors:  Rodney H Strahan; Michal E Schneider-Kolsky
Journal:  J Med Imaging Radiat Oncol       Date:  2010-10       Impact factor: 1.735

5.  Voice recognition dictation: radiologist as transcriptionist.

Authors:  John A Pezzullo; Glenn A Tung; Jeffrey M Rogg; Lawrence M Davis; Jeffrey M Brody; William W Mayo-Smith
Journal:  J Digit Imaging       Date:  2008-12       Impact factor: 4.056

6.  Frequency and spectrum of errors in final radiology reports generated with automatic speech recognition technology.

Authors:  Leslie E Quint; Douglas J Quint; James D Myles
Journal:  J Am Coll Radiol       Date:  2008-12       Impact factor: 5.532

7.  Journal Club: Voice recognition dictation: analysis of report volume and use of the send-to-editor function.

Authors:  Dennis R Williams; Sheila K Kori; Brenda Williams; Sandra J Sackrison; Henryk M Kowalski; Michael G McLaughlin; Brian S Kuszyk
Journal:  AJR Am J Roentgenol       Date:  2013-11       Impact factor: 3.959

8.  Implementation of speech recognition in a community-based radiology practice: effect on report turnaround times.

Authors:  Luciano M Prevedello; Stephen Ledbetter; Cameron Farkas; Ramin Khorasani
Journal:  J Am Coll Radiol       Date:  2013-10-23       Impact factor: 5.532

9.  Liability for typographical errors.

Authors:  Leonard Berlin
Journal:  AJR Am J Roentgenol       Date:  2011-02       Impact factor: 3.959

10.  The "p word".

Authors:  Paul H Ellenbogen
Journal:  J Am Coll Radiol       Date:  2012-09       Impact factor: 5.532

View more
  4 in total

1.  Automated Misspelling Detection and Correction in Persian Clinical Text.

Authors:  Azita Yazdani; Marjan Ghazisaeedi; Nasrin Ahmadinejad; Masoumeh Giti; Habibe Amjadi; Azin Nahvijou
Journal:  J Digit Imaging       Date:  2020-06       Impact factor: 4.056

2.  Application of a Domain-specific BERT for Detection of Speech Recognition Errors in Radiology Reports.

Authors:  Gunvant R Chaudhari; Tengxiao Liu; Timothy L Chen; Gabby B Joseph; Maya Vella; Yoo Jin Lee; Thienkhai H Vu; Youngho Seo; Andreas M Rauschecker; Charles E McCulloch; Jae Ho Sohn
Journal:  Radiol Artif Intell       Date:  2022-05-25

3.  Using BERT Models to Label Radiology Reports.

Authors:  John R Zech
Journal:  Radiol Artif Intell       Date:  2022-07-27

4.  A systematic review of natural language processing applied to radiology reports.

Authors:  Arlene Casey; Emma Davidson; Michael Poon; Hang Dong; Daniel Duma; Andreas Grivas; Claire Grover; Víctor Suárez-Paniagua; Richard Tobin; William Whiteley; Honghan Wu; Beatrice Alex
Journal:  BMC Med Inform Decis Mak       Date:  2021-06-03       Impact factor: 2.796

  4 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.