Deep learning algorithms for detection of critical findings in head CT scans

a retrospective study

Sasank Chilamkurthy, Rohit Ghosh, Swetha Tanamala, Mustafa Biviji, Norbert G Campeau, Vasantha Kumar Venugopal, Vidur Mahajan, Pooja Rao, Prashant Warier

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

BACKGROUND: Non-contrast head CT scan is the current standard for initial imaging of patients with head trauma or stroke symptoms. We aimed to develop and validate a set of deep learning algorithms for automated detection of the following key findings from these scans: intracranial haemorrhage and its types (ie, intraparenchymal, intraventricular, subdural, extradural, and subarachnoid); calvarial fractures; midline shift; and mass effect. METHODS: We retrospectively collected a dataset containing 313 318 head CT scans together with their clinical reports from around 20 centres in India between Jan 1, 2011, and June 1, 2017. A randomly selected part of this dataset (Qure25k dataset) was used for validation and the rest was used to develop algorithms. An additional validation dataset (CQ500 dataset) was collected in two batches from centres that were different from those used for the development and Qure25k datasets. We excluded postoperative scans and scans of patients younger than 7 years. The original clinical radiology report and consensus of three independent radiologists were considered as gold standard for the Qure25k and CQ500 datasets, respectively. Areas under the receiver operating characteristic curves (AUCs) were primarily used to assess the algorithms. FINDINGS: The Qure25k dataset contained 21 095 scans (mean age 43 years; 9030 [43%] female patients), and the CQ500 dataset consisted of 214 scans in the first batch (mean age 43 years; 94 [44%] female patients) and 277 scans in the second batch (mean age 52 years; 84 [30%] female patients). On the Qure25k dataset, the algorithms achieved an AUC of 0·92 (95% CI 0·91-0·93) for detecting intracranial haemorrhage (0·90 [0·89-0·91] for intraparenchymal, 0·96 [0·94-0·97] for intraventricular, 0·92 [0·90-0·93] for subdural, 0·93 [0·91-0·95] for extradural, and 0·90 [0·89-0·92] for subarachnoid). On the CQ500 dataset, AUC was 0·94 (0·92-0·97) for intracranial haemorrhage (0·95 [0·93-0·98], 0·93 [0·87-1·00], 0·95 [0·91-0·99], 0·97 [0·91-1·00], and 0·96 [0·92-0·99], respectively). AUCs on the Qure25k dataset were 0·92 (0·91-0·94) for calvarial fractures, 0·93 (0·91-0·94) for midline shift, and 0·86 (0·85-0·87) for mass effect, while AUCs on the CQ500 dataset were 0·96 (0·92-1·00), 0·97 (0·94-1·00), and 0·92 (0·89-0·95), respectively. INTERPRETATION: Our results show that deep learning algorithms can accurately identify head CT scan abnormalities requiring urgent attention, opening up the possibility to use these algorithms to automate the triage process.Qure.ai.

Original languageEnglish (US)
Pages (from-to)2388-2396
Number of pages9
JournalLancet (London, England)
Volume392
Issue number10162
DOIs
StatePublished - Dec 1 2018

Fingerprint

Retrospective Studies
Head
Learning
Area Under Curve
Intracranial Hemorrhages
Datasets
Triage
Craniocerebral Trauma
Radiology
ROC Curve
India
Consensus
Stroke

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Chilamkurthy, S., Ghosh, R., Tanamala, S., Biviji, M., Campeau, N. G., Venugopal, V. K., ... Warier, P. (2018). Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study. Lancet (London, England), 392(10162), 2388-2396. https://doi.org/10.1016/S0140-6736(18)31645-3

Deep learning algorithms for detection of critical findings in head CT scans : a retrospective study. / Chilamkurthy, Sasank; Ghosh, Rohit; Tanamala, Swetha; Biviji, Mustafa; Campeau, Norbert G; Venugopal, Vasantha Kumar; Mahajan, Vidur; Rao, Pooja; Warier, Prashant.

In: Lancet (London, England), Vol. 392, No. 10162, 01.12.2018, p. 2388-2396.

Research output: Contribution to journalArticle

Chilamkurthy, S, Ghosh, R, Tanamala, S, Biviji, M, Campeau, NG, Venugopal, VK, Mahajan, V, Rao, P & Warier, P 2018, 'Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study', Lancet (London, England), vol. 392, no. 10162, pp. 2388-2396. https://doi.org/10.1016/S0140-6736(18)31645-3
Chilamkurthy, Sasank ; Ghosh, Rohit ; Tanamala, Swetha ; Biviji, Mustafa ; Campeau, Norbert G ; Venugopal, Vasantha Kumar ; Mahajan, Vidur ; Rao, Pooja ; Warier, Prashant. / Deep learning algorithms for detection of critical findings in head CT scans : a retrospective study. In: Lancet (London, England). 2018 ; Vol. 392, No. 10162. pp. 2388-2396.
@article{468050771afa4631ac5bfbee4c6a6be0,
title = "Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study",
abstract = "BACKGROUND: Non-contrast head CT scan is the current standard for initial imaging of patients with head trauma or stroke symptoms. We aimed to develop and validate a set of deep learning algorithms for automated detection of the following key findings from these scans: intracranial haemorrhage and its types (ie, intraparenchymal, intraventricular, subdural, extradural, and subarachnoid); calvarial fractures; midline shift; and mass effect. METHODS: We retrospectively collected a dataset containing 313 318 head CT scans together with their clinical reports from around 20 centres in India between Jan 1, 2011, and June 1, 2017. A randomly selected part of this dataset (Qure25k dataset) was used for validation and the rest was used to develop algorithms. An additional validation dataset (CQ500 dataset) was collected in two batches from centres that were different from those used for the development and Qure25k datasets. We excluded postoperative scans and scans of patients younger than 7 years. The original clinical radiology report and consensus of three independent radiologists were considered as gold standard for the Qure25k and CQ500 datasets, respectively. Areas under the receiver operating characteristic curves (AUCs) were primarily used to assess the algorithms. FINDINGS: The Qure25k dataset contained 21 095 scans (mean age 43 years; 9030 [43{\%}] female patients), and the CQ500 dataset consisted of 214 scans in the first batch (mean age 43 years; 94 [44{\%}] female patients) and 277 scans in the second batch (mean age 52 years; 84 [30{\%}] female patients). On the Qure25k dataset, the algorithms achieved an AUC of 0·92 (95{\%} CI 0·91-0·93) for detecting intracranial haemorrhage (0·90 [0·89-0·91] for intraparenchymal, 0·96 [0·94-0·97] for intraventricular, 0·92 [0·90-0·93] for subdural, 0·93 [0·91-0·95] for extradural, and 0·90 [0·89-0·92] for subarachnoid). On the CQ500 dataset, AUC was 0·94 (0·92-0·97) for intracranial haemorrhage (0·95 [0·93-0·98], 0·93 [0·87-1·00], 0·95 [0·91-0·99], 0·97 [0·91-1·00], and 0·96 [0·92-0·99], respectively). AUCs on the Qure25k dataset were 0·92 (0·91-0·94) for calvarial fractures, 0·93 (0·91-0·94) for midline shift, and 0·86 (0·85-0·87) for mass effect, while AUCs on the CQ500 dataset were 0·96 (0·92-1·00), 0·97 (0·94-1·00), and 0·92 (0·89-0·95), respectively. INTERPRETATION: Our results show that deep learning algorithms can accurately identify head CT scan abnormalities requiring urgent attention, opening up the possibility to use these algorithms to automate the triage process.Qure.ai.",
author = "Sasank Chilamkurthy and Rohit Ghosh and Swetha Tanamala and Mustafa Biviji and Campeau, {Norbert G} and Venugopal, {Vasantha Kumar} and Vidur Mahajan and Pooja Rao and Prashant Warier",
year = "2018",
month = "12",
day = "1",
doi = "10.1016/S0140-6736(18)31645-3",
language = "English (US)",
volume = "392",
pages = "2388--2396",
journal = "The Lancet",
issn = "0140-6736",
publisher = "Elsevier Limited",
number = "10162",

}

TY - JOUR

T1 - Deep learning algorithms for detection of critical findings in head CT scans

T2 - a retrospective study

AU - Chilamkurthy, Sasank

AU - Ghosh, Rohit

AU - Tanamala, Swetha

AU - Biviji, Mustafa

AU - Campeau, Norbert G

AU - Venugopal, Vasantha Kumar

AU - Mahajan, Vidur

AU - Rao, Pooja

AU - Warier, Prashant

PY - 2018/12/1

Y1 - 2018/12/1

N2 - BACKGROUND: Non-contrast head CT scan is the current standard for initial imaging of patients with head trauma or stroke symptoms. We aimed to develop and validate a set of deep learning algorithms for automated detection of the following key findings from these scans: intracranial haemorrhage and its types (ie, intraparenchymal, intraventricular, subdural, extradural, and subarachnoid); calvarial fractures; midline shift; and mass effect. METHODS: We retrospectively collected a dataset containing 313 318 head CT scans together with their clinical reports from around 20 centres in India between Jan 1, 2011, and June 1, 2017. A randomly selected part of this dataset (Qure25k dataset) was used for validation and the rest was used to develop algorithms. An additional validation dataset (CQ500 dataset) was collected in two batches from centres that were different from those used for the development and Qure25k datasets. We excluded postoperative scans and scans of patients younger than 7 years. The original clinical radiology report and consensus of three independent radiologists were considered as gold standard for the Qure25k and CQ500 datasets, respectively. Areas under the receiver operating characteristic curves (AUCs) were primarily used to assess the algorithms. FINDINGS: The Qure25k dataset contained 21 095 scans (mean age 43 years; 9030 [43%] female patients), and the CQ500 dataset consisted of 214 scans in the first batch (mean age 43 years; 94 [44%] female patients) and 277 scans in the second batch (mean age 52 years; 84 [30%] female patients). On the Qure25k dataset, the algorithms achieved an AUC of 0·92 (95% CI 0·91-0·93) for detecting intracranial haemorrhage (0·90 [0·89-0·91] for intraparenchymal, 0·96 [0·94-0·97] for intraventricular, 0·92 [0·90-0·93] for subdural, 0·93 [0·91-0·95] for extradural, and 0·90 [0·89-0·92] for subarachnoid). On the CQ500 dataset, AUC was 0·94 (0·92-0·97) for intracranial haemorrhage (0·95 [0·93-0·98], 0·93 [0·87-1·00], 0·95 [0·91-0·99], 0·97 [0·91-1·00], and 0·96 [0·92-0·99], respectively). AUCs on the Qure25k dataset were 0·92 (0·91-0·94) for calvarial fractures, 0·93 (0·91-0·94) for midline shift, and 0·86 (0·85-0·87) for mass effect, while AUCs on the CQ500 dataset were 0·96 (0·92-1·00), 0·97 (0·94-1·00), and 0·92 (0·89-0·95), respectively. INTERPRETATION: Our results show that deep learning algorithms can accurately identify head CT scan abnormalities requiring urgent attention, opening up the possibility to use these algorithms to automate the triage process.Qure.ai.

AB - BACKGROUND: Non-contrast head CT scan is the current standard for initial imaging of patients with head trauma or stroke symptoms. We aimed to develop and validate a set of deep learning algorithms for automated detection of the following key findings from these scans: intracranial haemorrhage and its types (ie, intraparenchymal, intraventricular, subdural, extradural, and subarachnoid); calvarial fractures; midline shift; and mass effect. METHODS: We retrospectively collected a dataset containing 313 318 head CT scans together with their clinical reports from around 20 centres in India between Jan 1, 2011, and June 1, 2017. A randomly selected part of this dataset (Qure25k dataset) was used for validation and the rest was used to develop algorithms. An additional validation dataset (CQ500 dataset) was collected in two batches from centres that were different from those used for the development and Qure25k datasets. We excluded postoperative scans and scans of patients younger than 7 years. The original clinical radiology report and consensus of three independent radiologists were considered as gold standard for the Qure25k and CQ500 datasets, respectively. Areas under the receiver operating characteristic curves (AUCs) were primarily used to assess the algorithms. FINDINGS: The Qure25k dataset contained 21 095 scans (mean age 43 years; 9030 [43%] female patients), and the CQ500 dataset consisted of 214 scans in the first batch (mean age 43 years; 94 [44%] female patients) and 277 scans in the second batch (mean age 52 years; 84 [30%] female patients). On the Qure25k dataset, the algorithms achieved an AUC of 0·92 (95% CI 0·91-0·93) for detecting intracranial haemorrhage (0·90 [0·89-0·91] for intraparenchymal, 0·96 [0·94-0·97] for intraventricular, 0·92 [0·90-0·93] for subdural, 0·93 [0·91-0·95] for extradural, and 0·90 [0·89-0·92] for subarachnoid). On the CQ500 dataset, AUC was 0·94 (0·92-0·97) for intracranial haemorrhage (0·95 [0·93-0·98], 0·93 [0·87-1·00], 0·95 [0·91-0·99], 0·97 [0·91-1·00], and 0·96 [0·92-0·99], respectively). AUCs on the Qure25k dataset were 0·92 (0·91-0·94) for calvarial fractures, 0·93 (0·91-0·94) for midline shift, and 0·86 (0·85-0·87) for mass effect, while AUCs on the CQ500 dataset were 0·96 (0·92-1·00), 0·97 (0·94-1·00), and 0·92 (0·89-0·95), respectively. INTERPRETATION: Our results show that deep learning algorithms can accurately identify head CT scan abnormalities requiring urgent attention, opening up the possibility to use these algorithms to automate the triage process.Qure.ai.

UR - http://www.scopus.com/inward/record.url?scp=85058466258&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058466258&partnerID=8YFLogxK

U2 - 10.1016/S0140-6736(18)31645-3

DO - 10.1016/S0140-6736(18)31645-3

M3 - Article

VL - 392

SP - 2388

EP - 2396

JO - The Lancet

JF - The Lancet

SN - 0140-6736

IS - 10162

ER -