TY - JOUR
T1 - Fully Automated Thrombus Segmentation on CT Images of Patients with Acute Ischemic Stroke
AU - Mojtahedi, Mahsa
AU - Kappelhof, Manon
AU - Ponomareva, Elena
AU - Tolhuisen, Manon
AU - Jansen, Ivo
AU - Bruggeman, Agnetha A. E.
AU - Dutra, Bruna G.
AU - Yo, Lonneke
AU - Lecouffe, Natalie
AU - Hoving, Jan W.
AU - van Voorst, Henk
AU - Brouwer, Josje
AU - Terreros, Nerea Arrarte
AU - Konduri, Praneeta
AU - Meijer, Frederick J. A.
AU - Appelman, Auke
AU - Treurniet, Kilian M.
AU - Coutinho, Jonathan M.
AU - Roos, Yvo
AU - van Zwam, Wim
AU - Dippel, Diederik
AU - Gavves, Efstratios
AU - Emmer, Bart J.
AU - Majoie, Charles
AU - Marquering, Henk
N1 - Funding Information: Funding: This project is part of the Artificial Intelligence for Early Imaging Based Patient Selection in Acute Ischemic Stroke (AIRBORNE), which is funded by Top Sector Life Sciences & Health and Nicolab B.V. The commercial entities mentioned hereafter did not fund this project directly: MR CLEAN NO-IV was funded through the CONTRAST consortium, which acknowledges the support from the Netherlands Cardiovascular Research Initiative, an initiative of the Dutch Heart Foundation (CVON2015-01: CONTRAST), and from the Brain Foundation Netherlands (HA2015.01.06). Additionally, it is financed by the Ministry of Economic Affairs by means of the PPP allowance made available by the Top Sector Life Sciences & Health to stimulate public–private partnerships (LSHM17016). The MR CLEAN trial was supported by the Dutch Heart Foundation and by unrestricted grants from AngioCare BV, Medtronic/Covidien/EV3, MEDAC Gmbh/LAMEPRO, Penumbra Inc., Stryker, and Top Medical/Concentric. Funding Information: Conflicts of Interest: H.M. is co-founder and shareholder of Nicolab; C.M., Y.R., E.P. and I.J. are shareholders of Nicolab. I.J. and E.P. are employed by the company Nicolab. C.M. reports grants from CVON/Dutch Heart Foundation, TWIN Foundation, European Commission, Health Evaluation Netherlands, and Stryker (paid to the institution). Funding Information: This project is part of the Artificial Intelligence for Early Imaging Based Patient Selection in Acute Ischemic Stroke (AIRBORNE), which is funded by Top Sector Life Sciences & Health and Nicolab B.V. The commercial entities mentioned hereafter did not fund this project directly: MR CLEAN NO-IV was funded through the CONTRAST consortium, which acknowledges the support from the Netherlands Cardiovascular Research Initiative, an initiative of the Dutch Heart Foundation (CVON2015-01: CONTRAST), and from the Brain Foundation Netherlands (HA2015.01.06). Addition-ally, it is financed by the Ministry of Economic Affairs by means of the PPP allowance made available by the Top Sector Life Sciences & Health to stimulate public–private partnerships (LSHM17016). The MR CLEAN trial was supported by the Dutch Heart Foundation and by unrestricted grants from AngioCare BV, Medtronic/Covidien/EV3, MEDAC Gmbh/LAMEPRO, Penumbra Inc., Stryker, and Top Medical/Concentric. Publisher Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland.
PY - 2022/3/1
Y1 - 2022/3/1
N2 - Thrombus imaging characteristics are associated with treatment success and functional outcomes in stroke patients. However, assessing these characteristics based on manual annotations is labor intensive and subject to observer bias. Therefore, we aimed to create an automated pipeline for consistent and fast full thrombus segmentation. We used multicenter, multi-scanner datasets of anterior circulation stroke patients with baseline NCCT and CTA for training (n = 228) and testing (n = 100). We first found the occlusion location using StrokeViewer LVO and created a bounding box around it. Subsequently, we trained dual modality U-Net based convolutional neural networks (CNNs) to segment the thrombus inside this bounding box. We experimented with: (1) U-Net with two input channels for NCCT and CTA, and U-Nets with two encoders where (2) concatenate, (3) add, and (4) weighted-sum operators were used for feature fusion. Furthermore, we proposed a dynamic bounding box algorithm to adjust the bounding box. The dynamic bounding box algorithm reduces the missed cases but does not improve Dice. The two-encoder U-Net with a weighted-sum feature fusion shows the best performance (surface Dice 0.78, Dice 0.62, and 4% missed cases). Final segmentation results have high spatial accuracies and can therefore be used to determine thrombus characteristics and potentially benefit radiologists in clinical practice.
AB - Thrombus imaging characteristics are associated with treatment success and functional outcomes in stroke patients. However, assessing these characteristics based on manual annotations is labor intensive and subject to observer bias. Therefore, we aimed to create an automated pipeline for consistent and fast full thrombus segmentation. We used multicenter, multi-scanner datasets of anterior circulation stroke patients with baseline NCCT and CTA for training (n = 228) and testing (n = 100). We first found the occlusion location using StrokeViewer LVO and created a bounding box around it. Subsequently, we trained dual modality U-Net based convolutional neural networks (CNNs) to segment the thrombus inside this bounding box. We experimented with: (1) U-Net with two input channels for NCCT and CTA, and U-Nets with two encoders where (2) concatenate, (3) add, and (4) weighted-sum operators were used for feature fusion. Furthermore, we proposed a dynamic bounding box algorithm to adjust the bounding box. The dynamic bounding box algorithm reduces the missed cases but does not improve Dice. The two-encoder U-Net with a weighted-sum feature fusion shows the best performance (surface Dice 0.78, Dice 0.62, and 4% missed cases). Final segmentation results have high spatial accuracies and can therefore be used to determine thrombus characteristics and potentially benefit radiologists in clinical practice.
KW - CT angiography
KW - CT imaging
KW - Convolutional neural network (CNN)
KW - Ischemic stroke
KW - Segmentation
KW - Thrombus
KW - U-Net
UR - http://www.scopus.com/inward/record.url?scp=85126933935&partnerID=8YFLogxK
U2 - https://doi.org/10.3390/diagnostics12030698
DO - https://doi.org/10.3390/diagnostics12030698
M3 - Article
C2 - 35328251
SN - 2075-4418
VL - 12
JO - Diagnostics
JF - Diagnostics
IS - 3
M1 - 698
ER -