Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke

Hulin Kuang; Wu Qiu; Anna M. Boers; Scott Brown; Keith Muir; Charles B.L.M. Majoie; Diederik W.J. Dippel; Phil White; Jonathan Epstein; Peter J. Mitchell; Antoni Dávalos; Serge Bracard; Bruce Campbell; Jeffrey L. Saver; Tudor G. Jovin; Marta Rubiera; Alexander V. Khaw; Jai J. Shankar; Enrico Fainardi; Michael D. Hill; Andrew M. Demchuk; Mayank Goyal; Bijoy K. Menon

doi:10.1161/STROKEAHA.120.030092

Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke

Hulin Kuang
, Wu Qiu^*
, Anna M. Boers
, Scott Brown
, Keith Muir
, Charles B.L.M. Majoie
, Diederik W.J. Dippel
, Phil White
, Jonathan Epstein
, Peter J. Mitchell
, Antoni Dávalos
, Serge Bracard
, Bruce Campbell
, Jeffrey L. Saver
, Tudor G. Jovin
, Marta Rubiera
, Alexander V. Khaw
, Jai J. Shankar
, Enrico Fainardi
, Michael D. Hill

Andrew M. Demchuk, Mayank Goyal, Bijoy K. Menon^*

^*Corresponding author for this work

University of Calgary
University of Calgary
Amsterdam University Medical Centers
Altair Biostatistics
University of Glasgow
Erasmus MC
Newcastle University
CHU de Nancy
University of Melbourne
Departament de Fisica de la Universitat Autonoma de Barcelona
Université de Lorraine
David Geffen School of Medicine at UCLA
Cooper Neurological Institute
Vall d'Hebron Hospital Universitari
Western University
University of Manitoba
University of Florence

Research output: Contribution to journal › Article › peer-review

32 Scopus citations

Abstract

Background and Purpose: Prediction of infarct extent among patients with acute ischemic stroke using computed tomography perfusion is defined by predefined discrete computed tomography perfusion thresholds. Our objective is to develop a threshold-free computed tomography perfusion-based machine learning (ML) model to predict follow-up infarct in patients with acute ischemic stroke. Methods: Sixty-eight patients from the PRoveIT study (Measuring Collaterals With Multi-Phase CT Angiography in Patients With Ischemic Stroke) were used to derive a ML model using random forest to predict follow-up infarction voxel by voxel, and 137 patients from the HERMES study (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) were used to test the derived ML model. Average map, T_max, cerebral blood flow, cerebral blood volume, and time variables including stroke onset-to-imaging and imaging-to-reperfusion time, were used as features to train the ML model. Spatial and volumetric agreement between the ML model predicted follow-up infarct and actual follow-up infarct were assessed. Relative cerebral blood flow <0.3 threshold using RAPID software and time-dependent T_maxthresholds were compared with the ML model. Results: In the test cohort (137 patients), median follow-up infarct volume predicted by the ML model was 30.9 mL (interquartile range, 16.4-54.3 mL), compared with a median 29.6 mL (interquartile range, 11.1-70.9 mL) of actual follow-up infarct volume. The Pearson correlation coefficient between 2 measurements was 0.80 (95% CI, 0.74-0.86, P<0.001) while the volumetric difference was -3.2 mL (interquartile range, -16.7 to 6.1 mL). Volumetric difference with the ML model was smaller versus the relative cerebral blood flow <0.3 threshold and the time-dependent T_maxthreshold (P<0.001). Conclusions: A ML using computed tomography perfusion data and time estimates follow-up infarction in patients with acute ischemic stroke better than current methods.

Original language	English
Pages (from-to)	223-231
Number of pages	9
Journal	Stroke
Volume	52
Issue number	1
DOIs	https://doi.org/10.1161/STROKEAHA.120.030092
State	Published - 1 Jan 2021
Externally published	Yes

Keywords

acute ischemic stroke
computed tomographic perfusion
infarction
machine learning

Access to Document

10.1161/STROKEAHA.120.030092

Cite this

Kuang, H., Qiu, W., Boers, A. M., Brown, S., Muir, K., Majoie, C. B. L. M., Dippel, D. W. J., White, P., Epstein, J., Mitchell, P. J., Dávalos, A., Bracard, S., Campbell, B., Saver, J. L., Jovin, T. G., Rubiera, M., Khaw, A. V., Shankar, J. J., Fainardi, E., ... Menon, B. K. (2021). Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke. Stroke, 52(1), 223-231. https://doi.org/10.1161/STROKEAHA.120.030092

@article{cac4964122504b93a4023073f105856f,

title = "Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke",

abstract = "Background and Purpose: Prediction of infarct extent among patients with acute ischemic stroke using computed tomography perfusion is defined by predefined discrete computed tomography perfusion thresholds. Our objective is to develop a threshold-free computed tomography perfusion-based machine learning (ML) model to predict follow-up infarct in patients with acute ischemic stroke. Methods: Sixty-eight patients from the PRoveIT study (Measuring Collaterals With Multi-Phase CT Angiography in Patients With Ischemic Stroke) were used to derive a ML model using random forest to predict follow-up infarction voxel by voxel, and 137 patients from the HERMES study (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) were used to test the derived ML model. Average map, Tmax, cerebral blood flow, cerebral blood volume, and time variables including stroke onset-to-imaging and imaging-to-reperfusion time, were used as features to train the ML model. Spatial and volumetric agreement between the ML model predicted follow-up infarct and actual follow-up infarct were assessed. Relative cerebral blood flow <0.3 threshold using RAPID software and time-dependent Tmaxthresholds were compared with the ML model. Results: In the test cohort (137 patients), median follow-up infarct volume predicted by the ML model was 30.9 mL (interquartile range, 16.4-54.3 mL), compared with a median 29.6 mL (interquartile range, 11.1-70.9 mL) of actual follow-up infarct volume. The Pearson correlation coefficient between 2 measurements was 0.80 (95\% CI, 0.74-0.86, P<0.001) while the volumetric difference was -3.2 mL (interquartile range, -16.7 to 6.1 mL). Volumetric difference with the ML model was smaller versus the relative cerebral blood flow <0.3 threshold and the time-dependent Tmaxthreshold (P<0.001). Conclusions: A ML using computed tomography perfusion data and time estimates follow-up infarction in patients with acute ischemic stroke better than current methods.",

keywords = "acute ischemic stroke, computed tomographic perfusion, infarction, machine learning",

author = "Hulin Kuang and Wu Qiu and Boers, \{Anna M.\} and Scott Brown and Keith Muir and Majoie, \{Charles B.L.M.\} and Dippel, \{Diederik W.J.\} and Phil White and Jonathan Epstein and Mitchell, \{Peter J.\} and Antoni D{\'a}valos and Serge Bracard and Bruce Campbell and Saver, \{Jeffrey L.\} and Jovin, \{Tudor G.\} and Marta Rubiera and Khaw, \{Alexander V.\} and Shankar, \{Jai J.\} and Enrico Fainardi and Hill, \{Michael D.\} and Demchuk, \{Andrew M.\} and Mayank Goyal and Menon, \{Bijoy K.\}",

year = "2021",

month = jan,

day = "1",

doi = "10.1161/STROKEAHA.120.030092",

language = "Ingl{\'e}s",

volume = "52",

pages = "223--231",

journal = "Stroke",

issn = "0039-2499",

publisher = "Wolters Kluwer Health",

number = "1",

}

Kuang, H, Qiu, W, Boers, AM, Brown, S, Muir, K, Majoie, CBLM, Dippel, DWJ, White, P, Epstein, J, Mitchell, PJ, Dávalos, A, Bracard, S, Campbell, B, Saver, JL, Jovin, TG, Rubiera, M, Khaw, AV, Shankar, JJ, Fainardi, E, Hill, MD, Demchuk, AM, Goyal, M & Menon, BK 2021, 'Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke', Stroke, vol. 52, no. 1, pp. 223-231. https://doi.org/10.1161/STROKEAHA.120.030092

TY - JOUR

T1 - Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke

AU - Kuang, Hulin

AU - Qiu, Wu

AU - Boers, Anna M.

AU - Brown, Scott

AU - Muir, Keith

AU - Majoie, Charles B.L.M.

AU - Dippel, Diederik W.J.

AU - White, Phil

AU - Epstein, Jonathan

AU - Mitchell, Peter J.

AU - Dávalos, Antoni

AU - Bracard, Serge

AU - Campbell, Bruce

AU - Saver, Jeffrey L.

AU - Jovin, Tudor G.

AU - Rubiera, Marta

AU - Khaw, Alexander V.

AU - Shankar, Jai J.

AU - Fainardi, Enrico

AU - Hill, Michael D.

AU - Demchuk, Andrew M.

AU - Goyal, Mayank

AU - Menon, Bijoy K.

PY - 2021/1/1

Y1 - 2021/1/1

N2 - Background and Purpose: Prediction of infarct extent among patients with acute ischemic stroke using computed tomography perfusion is defined by predefined discrete computed tomography perfusion thresholds. Our objective is to develop a threshold-free computed tomography perfusion-based machine learning (ML) model to predict follow-up infarct in patients with acute ischemic stroke. Methods: Sixty-eight patients from the PRoveIT study (Measuring Collaterals With Multi-Phase CT Angiography in Patients With Ischemic Stroke) were used to derive a ML model using random forest to predict follow-up infarction voxel by voxel, and 137 patients from the HERMES study (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) were used to test the derived ML model. Average map, Tmax, cerebral blood flow, cerebral blood volume, and time variables including stroke onset-to-imaging and imaging-to-reperfusion time, were used as features to train the ML model. Spatial and volumetric agreement between the ML model predicted follow-up infarct and actual follow-up infarct were assessed. Relative cerebral blood flow <0.3 threshold using RAPID software and time-dependent Tmaxthresholds were compared with the ML model. Results: In the test cohort (137 patients), median follow-up infarct volume predicted by the ML model was 30.9 mL (interquartile range, 16.4-54.3 mL), compared with a median 29.6 mL (interquartile range, 11.1-70.9 mL) of actual follow-up infarct volume. The Pearson correlation coefficient between 2 measurements was 0.80 (95% CI, 0.74-0.86, P<0.001) while the volumetric difference was -3.2 mL (interquartile range, -16.7 to 6.1 mL). Volumetric difference with the ML model was smaller versus the relative cerebral blood flow <0.3 threshold and the time-dependent Tmaxthreshold (P<0.001). Conclusions: A ML using computed tomography perfusion data and time estimates follow-up infarction in patients with acute ischemic stroke better than current methods.

AB - Background and Purpose: Prediction of infarct extent among patients with acute ischemic stroke using computed tomography perfusion is defined by predefined discrete computed tomography perfusion thresholds. Our objective is to develop a threshold-free computed tomography perfusion-based machine learning (ML) model to predict follow-up infarct in patients with acute ischemic stroke. Methods: Sixty-eight patients from the PRoveIT study (Measuring Collaterals With Multi-Phase CT Angiography in Patients With Ischemic Stroke) were used to derive a ML model using random forest to predict follow-up infarction voxel by voxel, and 137 patients from the HERMES study (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) were used to test the derived ML model. Average map, Tmax, cerebral blood flow, cerebral blood volume, and time variables including stroke onset-to-imaging and imaging-to-reperfusion time, were used as features to train the ML model. Spatial and volumetric agreement between the ML model predicted follow-up infarct and actual follow-up infarct were assessed. Relative cerebral blood flow <0.3 threshold using RAPID software and time-dependent Tmaxthresholds were compared with the ML model. Results: In the test cohort (137 patients), median follow-up infarct volume predicted by the ML model was 30.9 mL (interquartile range, 16.4-54.3 mL), compared with a median 29.6 mL (interquartile range, 11.1-70.9 mL) of actual follow-up infarct volume. The Pearson correlation coefficient between 2 measurements was 0.80 (95% CI, 0.74-0.86, P<0.001) while the volumetric difference was -3.2 mL (interquartile range, -16.7 to 6.1 mL). Volumetric difference with the ML model was smaller versus the relative cerebral blood flow <0.3 threshold and the time-dependent Tmaxthreshold (P<0.001). Conclusions: A ML using computed tomography perfusion data and time estimates follow-up infarction in patients with acute ischemic stroke better than current methods.

KW - acute ischemic stroke

KW - computed tomographic perfusion

KW - infarction

KW - machine learning

UR - https://www.scopus.com/pages/publications/85099022906

U2 - 10.1161/STROKEAHA.120.030092

DO - 10.1161/STROKEAHA.120.030092

M3 - Artículo

C2 - 33280549

AN - SCOPUS:85099022906

SN - 0039-2499

VL - 52

SP - 223

EP - 231

JO - Stroke

JF - Stroke

IS - 1

ER -

Computed Tomography Perfusion-Based Machine Learning Model Better Predicts Follow-Up Infarction in Patients With Acute Ischemic Stroke

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this