Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study

Objectives Identifying whether there is a traumatic intracranial bleed (ICB+) on head CT is critical for clinical care and research. Free text CT reports are unstructured and therefore must undergo time-consuming manual review. Existing artificial intelligence classification schemes are not optimise...

Full description

Saved in:
Bibliographic Details
Main Authors: Stephen H Thomas, Jason Pott, Imogen Skene, Ben Bloom, Michael Cheetham, Adrian Haimovich, Raine Astin-Chamberlain, Sophie L Williams, Sandra Langsted
Format: Article
Language:English
Published: BMJ Publishing Group 2025-07-01
Series:BMJ Health & Care Informatics
Online Access:https://informatics.bmj.com/content/32/1/e101433.full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850077557067612160
author Stephen H Thomas
Jason Pott
Imogen Skene
Ben Bloom
Michael Cheetham
Adrian Haimovich
Raine Astin-Chamberlain
Sophie L Williams
Sandra Langsted
author_facet Stephen H Thomas
Jason Pott
Imogen Skene
Ben Bloom
Michael Cheetham
Adrian Haimovich
Raine Astin-Chamberlain
Sophie L Williams
Sandra Langsted
author_sort Stephen H Thomas
collection DOAJ
description Objectives Identifying whether there is a traumatic intracranial bleed (ICB+) on head CT is critical for clinical care and research. Free text CT reports are unstructured and therefore must undergo time-consuming manual review. Existing artificial intelligence classification schemes are not optimised for the emergency department endpoint of classification of ICB+ or ICB−. We sought to assess three methods for classifying CT reports: a text classification (TC) programme, a commercial natural language processing programme (Clinithink) and a generative pretrained transformer large language model (Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting (DECIPHER)-LLM).Methods Primary objective: determine the diagnostic classification performance of the dichotomous categorisation of each of the three approaches.Secondary objective: determine whether the LLM could achieve a substantial reduction in CT report review workload while maintaining 100% sensitivity.Anonymised radiology reports of head CT scans performed for trauma were manually labelled as ICB+/−. Training and validation sets were randomly created to train the TC and natural language processing models. Prompts were written to train the LLM.Results 898 reports were manually labelled. Sensitivity and specificity (95% CI)) of TC, Clinithink and DECIPHER-LLM (with probability of ICB set at 10%) were respectively 87.9% (76.7% to 95.0%) and 98.2% (96.3% to 99.3%), 75.9% (62.8% to 86.1%) and 96.2% (93.8% to 97.8%) and 100% (93.8% to 100%) and 97.4% (95.3% to 98.8%).With DECIPHER-LLM probability of ICB+ threshold of 10% set to identify CT reports requiring manual evaluation, CT reports requiring manual classification reduced by an estimated 385/449 cases (85.7% (95% CI 82.1% to 88.9%)) while maintaining 100% sensitivity.Discussion and conclusion DECIPHER-LLM outperformed other tested free-text classification methods.
format Article
id doaj-art-d9be3cbc7f7c4e4698db14334d707970
institution DOAJ
issn 2632-1009
language English
publishDate 2025-07-01
publisher BMJ Publishing Group
record_format Article
series BMJ Health & Care Informatics
spelling doaj-art-d9be3cbc7f7c4e4698db14334d7079702025-08-20T02:45:46ZengBMJ Publishing GroupBMJ Health & Care Informatics2632-10092025-07-0132110.1136/bmjhci-2025-101433Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER studyStephen H Thomas0Jason Pott1Imogen Skene2Ben Bloom3Michael Cheetham4Adrian Haimovich5Raine Astin-Chamberlain6Sophie L Williams7Sandra Langsted84 Queen Mary University of London Barts and The London School of Medicine and Dentistry, London, UK1 Emergency Department, Barts Health NHS Trust, London, UKQueen Mary University of London, London, UKEmergency Department, Royal London Hospital, Barts Health NHS Trust, London, UKEmergency Department, Barts Health NHS Trust, London, UKDepartment of Emergency Medicine, Harvard Medical School, Boston, Massachusetts, USABarts Health NHS Trust, London, UKBarts Life Sciences, Barts Health NHS Trust, London, UKDepartment of Emergency Medicine, Randers Regional Hospital, Randers, DenmarkObjectives Identifying whether there is a traumatic intracranial bleed (ICB+) on head CT is critical for clinical care and research. Free text CT reports are unstructured and therefore must undergo time-consuming manual review. Existing artificial intelligence classification schemes are not optimised for the emergency department endpoint of classification of ICB+ or ICB−. We sought to assess three methods for classifying CT reports: a text classification (TC) programme, a commercial natural language processing programme (Clinithink) and a generative pretrained transformer large language model (Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting (DECIPHER)-LLM).Methods Primary objective: determine the diagnostic classification performance of the dichotomous categorisation of each of the three approaches.Secondary objective: determine whether the LLM could achieve a substantial reduction in CT report review workload while maintaining 100% sensitivity.Anonymised radiology reports of head CT scans performed for trauma were manually labelled as ICB+/−. Training and validation sets were randomly created to train the TC and natural language processing models. Prompts were written to train the LLM.Results 898 reports were manually labelled. Sensitivity and specificity (95% CI)) of TC, Clinithink and DECIPHER-LLM (with probability of ICB set at 10%) were respectively 87.9% (76.7% to 95.0%) and 98.2% (96.3% to 99.3%), 75.9% (62.8% to 86.1%) and 96.2% (93.8% to 97.8%) and 100% (93.8% to 100%) and 97.4% (95.3% to 98.8%).With DECIPHER-LLM probability of ICB+ threshold of 10% set to identify CT reports requiring manual evaluation, CT reports requiring manual classification reduced by an estimated 385/449 cases (85.7% (95% CI 82.1% to 88.9%)) while maintaining 100% sensitivity.Discussion and conclusion DECIPHER-LLM outperformed other tested free-text classification methods.https://informatics.bmj.com/content/32/1/e101433.full
spellingShingle Stephen H Thomas
Jason Pott
Imogen Skene
Ben Bloom
Michael Cheetham
Adrian Haimovich
Raine Astin-Chamberlain
Sophie L Williams
Sandra Langsted
Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
BMJ Health & Care Informatics
title Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
title_full Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
title_fullStr Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
title_full_unstemmed Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
title_short Digitalizing English-language CT Interpretation for Positive Haemorrhage Evaluation Reporting: the DECIPHER study
title_sort digitalizing english language ct interpretation for positive haemorrhage evaluation reporting the decipher study
url https://informatics.bmj.com/content/32/1/e101433.full
work_keys_str_mv AT stephenhthomas digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT jasonpott digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT imogenskene digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT benbloom digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT michaelcheetham digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT adrianhaimovich digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT raineastinchamberlain digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT sophielwilliams digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy
AT sandralangsted digitalizingenglishlanguagectinterpretationforpositivehaemorrhageevaluationreportingthedecipherstudy