Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke

Manon L. Tolhuisen; Jan W. Hoving; Miou S. Koopman; Manon Kappelhof; Henk van Voorst; Agnetha E. Bruggeman; Adam M. Demchuck; Diederik W.J. Dippel; Bart J. Emmer; Serge Bracard; Francis Guillemin; Robert J. van Oostenbrugge; Peter J. Mitchell; Wim H. van Zwam; Michael D. Hill; Yvo B.W.E.M. Roos; Tudor G. Jovin; Olvert A. Berkhemer; Bruce C.V. Campbell; Jeffrey Saver; Phil White; Keith W. Muir; Mayank Goyal; Henk A. Marquering; Charles B. Majoie; Matthan W.A. Caan; CLEAN NO IV and HERMES investigators on behalf of the

doi:10.3390/diagnostics12081786

Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke

Manon L. Tolhuisen^*
, Jan W. Hoving
, Miou S. Koopman
, Manon Kappelhof
, Henk van Voorst
, Agnetha E. Bruggeman
, Adam M. Demchuck
, Diederik W.J. Dippel
, Bart J. Emmer
, Serge Bracard
, Francis Guillemin
, Robert J. van Oostenbrugge
, Peter J. Mitchell
, Wim H. van Zwam
, Michael D. Hill
, Yvo B.W.E.M. Roos
, Tudor G. Jovin
, Olvert A. Berkhemer
, Bruce C.V. Campbell
, Jeffrey Saver

Phil White, Keith W. Muir, Mayank Goyal, Henk A. Marquering, Charles B. Majoie, Matthan W.A. Caan^*, CLEAN NO IV and HERMES investigators on behalf of the

^*Corresponding author for this work

Amsterdam UMC
Amsterdam University Medical Centers
University of Calgary
Erasmus MC
Université de Lorraine
Maastricht University Medical Center
Maastricht University
Royal Melbourne Hospital
University of Calgary
University of Pittsburgh
University of Melbourne
David Geffen School of Medicine at UCLA
Newcastle University
Newcastle upon Tyne Hospitals
University of Glasgow

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

18 Scopus citations

Abstract

Infarct volume (FIV) on follow-up diffusion-weighted imaging (FU-DWI) is only moderately associated with functional outcome in acute ischemic stroke patients. However, FU-DWI may contain other imaging biomarkers that could aid in improving outcome prediction models for acute ischemic stroke. We included FU-DWI data from the HERMES, ISLES, and MR CLEAN-NO IV databases. Lesions were segmented using a deep learning model trained on the HERMES and ISLES datasets. We assessed the performance of three classifiers in predicting functional independence for the MR CLEAN-NO IV trial cohort based on: (1) FIV alone, (2) the most important features obtained from a trained convolutional autoencoder (CAE), and (3) radiomics. Furthermore, we investigated feature importance in the radiomic-feature-based model. For outcome prediction, we included 206 patients: 144 scans were included in the training set, 21 in the validation set, and 41 in the test set. The classifiers that included the CAE and the radiomic features showed AUC values of 0.88 and 0.81, respectively, while the model based on FIV had an AUC of 0.79. This difference was not found to be statistically significant. Feature importance results showed that lesion intensity heterogeneity received more weight than lesion volume in outcome prediction. This study suggests that predictions of functional outcome should not be based on FIV alone and that FU-DWI images capture additional prognostic information.

Original language	English
Article number	1786
Journal	Diagnostics
Volume	12
Issue number	8
DOIs	https://doi.org/10.3390/diagnostics12081786
State	Published - Aug 2022
Externally published	Yes

Keywords

acute ischemic stroke
follow-up DWI
functional independence
infarct core image features
infarct core segmentation
support vector machine

Access to Document

10.3390/diagnostics12081786

Cite this

Tolhuisen, M. L., Hoving, J. W., Koopman, M. S., Kappelhof, M., van Voorst, H., Bruggeman, A. E., Demchuck, A. M., Dippel, D. W. J., Emmer, B. J., Bracard, S., Guillemin, F., van Oostenbrugge, R. J., Mitchell, P. J., van Zwam, W. H., Hill, M. D., Roos, Y. B. W. E. M., Jovin, T. G., Berkhemer, O. A., Campbell, B. C. V., ... on behalf of the, CLEAN. NO. IV. A. HERMES. I. (2022). Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke. Diagnostics, 12(8), Article 1786. https://doi.org/10.3390/diagnostics12081786

@article{b088214115514023a3dc3deb23850f5a,

title = "Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke",

abstract = "Infarct volume (FIV) on follow-up diffusion-weighted imaging (FU-DWI) is only moderately associated with functional outcome in acute ischemic stroke patients. However, FU-DWI may contain other imaging biomarkers that could aid in improving outcome prediction models for acute ischemic stroke. We included FU-DWI data from the HERMES, ISLES, and MR CLEAN-NO IV databases. Lesions were segmented using a deep learning model trained on the HERMES and ISLES datasets. We assessed the performance of three classifiers in predicting functional independence for the MR CLEAN-NO IV trial cohort based on: (1) FIV alone, (2) the most important features obtained from a trained convolutional autoencoder (CAE), and (3) radiomics. Furthermore, we investigated feature importance in the radiomic-feature-based model. For outcome prediction, we included 206 patients: 144 scans were included in the training set, 21 in the validation set, and 41 in the test set. The classifiers that included the CAE and the radiomic features showed AUC values of 0.88 and 0.81, respectively, while the model based on FIV had an AUC of 0.79. This difference was not found to be statistically significant. Feature importance results showed that lesion intensity heterogeneity received more weight than lesion volume in outcome prediction. This study suggests that predictions of functional outcome should not be based on FIV alone and that FU-DWI images capture additional prognostic information.",

keywords = "acute ischemic stroke, follow-up DWI, functional independence, infarct core image features, infarct core segmentation, support vector machine",

author = "Tolhuisen, \{Manon L.\} and Hoving, \{Jan W.\} and Koopman, \{Miou S.\} and Manon Kappelhof and \{van Voorst\}, Henk and Bruggeman, \{Agnetha E.\} and Demchuck, \{Adam M.\} and Dippel, \{Diederik W.J.\} and Emmer, \{Bart J.\} and Serge Bracard and Francis Guillemin and \{van Oostenbrugge\}, \{Robert J.\} and Mitchell, \{Peter J.\} and \{van Zwam\}, \{Wim H.\} and Hill, \{Michael D.\} and Roos, \{Yvo B.W.E.M.\} and Jovin, \{Tudor G.\} and Berkhemer, \{Olvert A.\} and Campbell, \{Bruce C.V.\} and Jeffrey Saver and Phil White and Muir, \{Keith W.\} and Mayank Goyal and Marquering, \{Henk A.\} and Majoie, \{Charles B.\} and Caan, \{Matthan W.A.\} and \{on behalf of the\}, \{CLEAN NO IV and HERMES investigators\}",

note = "Publisher Copyright: {\textcopyright} 2022 by the authors.",

year = "2022",

month = aug,

doi = "10.3390/diagnostics12081786",

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

volume = "12",

journal = "Diagnostics",

issn = "2075-4418",

publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",

number = "8",

}

Tolhuisen, ML, Hoving, JW, Koopman, MS, Kappelhof, M, van Voorst, H, Bruggeman, AE, Demchuck, AM, Dippel, DWJ, Emmer, BJ, Bracard, S, Guillemin, F, van Oostenbrugge, RJ, Mitchell, PJ, van Zwam, WH, Hill, MD, Roos, YBWEM, Jovin, TG, Berkhemer, OA, Campbell, BCV, Saver, J, White, P, Muir, KW, Goyal, M, Marquering, HA, Majoie, CB, Caan, MWA & on behalf of the, CLEANNOIVAHERMESI 2022, 'Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke', Diagnostics, vol. 12, no. 8, 1786. https://doi.org/10.3390/diagnostics12081786

TY - JOUR

T1 - Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke

AU - Tolhuisen, Manon L.

AU - Hoving, Jan W.

AU - Koopman, Miou S.

AU - Kappelhof, Manon

AU - van Voorst, Henk

AU - Bruggeman, Agnetha E.

AU - Demchuck, Adam M.

AU - Dippel, Diederik W.J.

AU - Emmer, Bart J.

AU - Bracard, Serge

AU - Guillemin, Francis

AU - van Oostenbrugge, Robert J.

AU - Mitchell, Peter J.

AU - van Zwam, Wim H.

AU - Hill, Michael D.

AU - Roos, Yvo B.W.E.M.

AU - Jovin, Tudor G.

AU - Berkhemer, Olvert A.

AU - Campbell, Bruce C.V.

AU - Saver, Jeffrey

AU - White, Phil

AU - Muir, Keith W.

AU - Goyal, Mayank

AU - Marquering, Henk A.

AU - Majoie, Charles B.

AU - Caan, Matthan W.A.

AU - on behalf of the, CLEAN NO IV and HERMES investigators

PY - 2022/8

Y1 - 2022/8

N2 - Infarct volume (FIV) on follow-up diffusion-weighted imaging (FU-DWI) is only moderately associated with functional outcome in acute ischemic stroke patients. However, FU-DWI may contain other imaging biomarkers that could aid in improving outcome prediction models for acute ischemic stroke. We included FU-DWI data from the HERMES, ISLES, and MR CLEAN-NO IV databases. Lesions were segmented using a deep learning model trained on the HERMES and ISLES datasets. We assessed the performance of three classifiers in predicting functional independence for the MR CLEAN-NO IV trial cohort based on: (1) FIV alone, (2) the most important features obtained from a trained convolutional autoencoder (CAE), and (3) radiomics. Furthermore, we investigated feature importance in the radiomic-feature-based model. For outcome prediction, we included 206 patients: 144 scans were included in the training set, 21 in the validation set, and 41 in the test set. The classifiers that included the CAE and the radiomic features showed AUC values of 0.88 and 0.81, respectively, while the model based on FIV had an AUC of 0.79. This difference was not found to be statistically significant. Feature importance results showed that lesion intensity heterogeneity received more weight than lesion volume in outcome prediction. This study suggests that predictions of functional outcome should not be based on FIV alone and that FU-DWI images capture additional prognostic information.

AB - Infarct volume (FIV) on follow-up diffusion-weighted imaging (FU-DWI) is only moderately associated with functional outcome in acute ischemic stroke patients. However, FU-DWI may contain other imaging biomarkers that could aid in improving outcome prediction models for acute ischemic stroke. We included FU-DWI data from the HERMES, ISLES, and MR CLEAN-NO IV databases. Lesions were segmented using a deep learning model trained on the HERMES and ISLES datasets. We assessed the performance of three classifiers in predicting functional independence for the MR CLEAN-NO IV trial cohort based on: (1) FIV alone, (2) the most important features obtained from a trained convolutional autoencoder (CAE), and (3) radiomics. Furthermore, we investigated feature importance in the radiomic-feature-based model. For outcome prediction, we included 206 patients: 144 scans were included in the training set, 21 in the validation set, and 41 in the test set. The classifiers that included the CAE and the radiomic features showed AUC values of 0.88 and 0.81, respectively, while the model based on FIV had an AUC of 0.79. This difference was not found to be statistically significant. Feature importance results showed that lesion intensity heterogeneity received more weight than lesion volume in outcome prediction. This study suggests that predictions of functional outcome should not be based on FIV alone and that FU-DWI images capture additional prognostic information.

KW - acute ischemic stroke

KW - follow-up DWI

KW - functional independence

KW - infarct core image features

KW - infarct core segmentation

KW - support vector machine

UR - https://www.scopus.com/pages/publications/85137408323

U2 - 10.3390/diagnostics12081786

DO - 10.3390/diagnostics12081786

M3 - Artículo

AN - SCOPUS:85137408323

SN - 2075-4418

VL - 12

JO - Diagnostics

JF - Diagnostics

IS - 8

M1 - 1786

ER -

Outcome Prediction Based on Automatically Extracted Infarct Core Image Features in Patients with Acute Ischemic Stroke

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this