Neural Network–derived Perfusion Maps for the Assessment of Lesions in Patients with Acute Ischemic Stroke

AbstractPurposeIn this study we investigate whether a Convolutional Neural Network (CNN) can generate clinically relevant parametric maps from CT perfusion data in a clinical setting of patients with acute ischemic stroke.MethodsTraining of the CNN was done on a subset of 100 perfusion data, while 15 samples were used as validation. All the data used for the training/validation of the network and to generate ground truth (GT) maps, using a state-of-the-art deconvolution-algorithm, were previously pre-processed using a standard pipeline. Validation was carried out through manual segmentation of infarct core and penumbra on both CNN-derived maps and GT maps. Concordance among segmented lesions was assessed using the Dice and the Pearson correlation coefficients across lesion volumes.ResultsMean Dice scores from two different raters and the GT maps were > 0.70 (good-matching). Inter-rater concordance was also high and strong correlation was found between lesion volumes of CNN maps and GT maps (0.99, 0.98).ConclusionOur CNN-based approach generated clinically relevant perfusion maps that are comparable to state-of-the-art perfusion analysis methods based on deconvolution of the data. Moreover, the proposed technique requires less information to estimate the ischemic core and thus might allow the development of novel perfusion protocols with lower radiation dose.

Download Full-text

Improvement of automatic ischemic stroke lesion segmentation in CT perfusion maps using a learned deep neural network

Computers in Biology and Medicine ◽

10.1016/j.compbiomed.2021.104849 ◽

2021 ◽

Vol 137 ◽

pp. 104849

Author(s):

Mohsen Soltanpour ◽

Russ Greiner ◽

Pierre Boulanger ◽

Brian Buck

Keyword(s):

Neural Network ◽

Ischemic Stroke ◽

Deep Neural Network ◽

Ct Perfusion ◽

Lesion Segmentation ◽

Stroke Lesion ◽

Perfusion Maps

Download Full-text

Is there more valuable information in PWI datasets for a voxel-wise acute ischemic stroke tissue outcome prediction than what is represented by typical perfusion maps?

10.1117/12.2043490 ◽

2014 ◽

Cited By ~ 1

Author(s):

Nils Daniel Forkert ◽

Susanne Siemonsen ◽

Michael Dalski ◽

Tobias Verleger ◽

Andre Kemmling ◽

...

Keyword(s):

Ischemic Stroke ◽

Acute Ischemic Stroke ◽

Outcome Prediction ◽

Perfusion Maps

Download Full-text

A Convolutional Neural Network for Anterior Intra-Arterial Thrombus Detection and Segmentation on Non-Contrast Computed Tomography of Patients with Acute Ischemic Stroke

Applied Sciences ◽

10.3390/app10144861 ◽

2020 ◽

Vol 10 (14) ◽

pp. 4861

Author(s):

Manon L. Tolhuisen ◽

Elena Ponomareva ◽

Anne M. M. Boers ◽

Ivo G. H. Jansen ◽

Miou S. Koopman ◽

...

Keyword(s):

Neural Network ◽

Computed Tomography ◽

Ischemic Stroke ◽

Convolutional Neural Network ◽

Acute Ischemic Stroke ◽

Ground Truth ◽

Correct Identification ◽

Anterior Circulation ◽

Vessel Occlusion ◽

Comprehensive Stroke Center

The aim of this study was to develop a convolutional neural network (CNN) that automatically detects and segments intra-arterial thrombi on baseline non-contrast computed tomography (NCCT) scans. We retrospectively collected computed tomography (CT)-scans of patients with an anterior circulation large vessel occlusion (LVO) from the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands trial, both for training (n = 86) and validation (n = 43). For testing we included patients with (n = 58) and without (n = 45) an LVO from our comprehensive stroke center. Ground truth was established by consensus between two experts using both CT angiography and NCCT. We evaluated the CNN for correct identification of a thrombus, its location and thrombus segmentation and compared these with the results of a neurologist in training and expert neuroradiologist. Sensitivity of the CNN thrombus detection was 0.86, vs. 0.95 and 0.79 for the neuroradiologists. Specificity was 0.65 for the network vs. 0.58 and 0.82 for the neuroradiologists. The CNN correctly identified the location of the thrombus in 79% of the cases, compared to 81% and 77% for the neuroradiologists. The sensitivity and specificity for thrombus identification and the rate for correct thrombus location assessment by the CNN were similar to those of expert neuroradiologists.

Download Full-text

Abstract 3048: Real-World Clinical Use of CT Perfusion for Diagnosis and Prediction of Lesion Growth in Acute Ischemic Stroke

Stroke ◽

10.1161/str.43.suppl_1.a3048 ◽

2012 ◽

Vol 43 (suppl_1) ◽

Author(s):

Branko N Huisa ◽

William P Neil ◽

Nhu T Bruce ◽

Marcel Maya ◽

Benedict Pereira ◽

...

Keyword(s):

Ischemic Stroke ◽

Acute Ischemic Stroke ◽

Real World ◽

Cerebral Blood Volume ◽

Ct Perfusion ◽

Medical Center ◽

Brain Mri ◽

Academic Medical Center ◽

Acute Ischemia ◽

Perfusion Maps

Background: Diffusion-weighted imaging (DWI) detects acute ischemia with a high sensitivity. In research centers, qualitative CT perfusion (CTP) mapping correlates well with DWI and may accurately differentiate the infarct core from ischemic penumbra. The value of the CTP in real-world clinical practice, however, has not been fully established. We investigated the yield of CTP - derived cerebral blood volume (CBV) and mean transient time (MTT) for the detection of cerebral ischemia in a sample of acute ischemic stroke (AIS) patients. Methods: In a large metropolitan academic medical center that is a certified Primary Stroke Center (PSC) we retrospectively studied 162 patients who presented between January 2008 and July 2010 with symptoms suggestive of AIS. All patients had an initial Code Brain protocol including non-contrast head CT, CTP, and CTA. As clinically indicated, some patients underwent follow up brain MRI within 48 hours. Acute perfusion maps were derived in real time by a trained operator. From the obtained images CBV, MTT and DWI lesion volumes were manually traced using planimetry (ImageJ v1.42) by two stroke neurologists blinded to clinical information. Volumes were calculated using the Cavaleri theorem. Sensitivity, specificity and statistical analysis were calculated using Graph Pad 5.0. Results: Of 162 patients with acute stroke-like symptoms, 73 had DWI lesions. The sensitivity and specificity to detect abnormal DWI signals were 23% and 100%, for CBV; and 43.8% and 98.9% for MTT. For DWI lesions ≥5ml the yield was 59.3% for CVB and 77.8% for MTT. For lesions ≥10ml the yield was 68.4% for CBV and 89.5% for MTT. In patients with NIHSS ≥5, CBV predicted abnormal DWI in 22.6% and MTT in 35.5%. In patients with NIHSS ≥10, CBV and MTT, both had a yield of 50.0%. A CBV - MTT mismatch of >25% predicted MRI lesion extension in 81.25% of the cases. There were small but significant correlations for DWI versus CBV lesion volumes ( r 2 0.32, P= 0.0001), and for DWI versus MTT lesion volumes ( r 2 0.29, P <0.0001). Correlation between DWI and perfusion maps for MCA territory infarcts were CBV ( r 2 0.3, P <0.0001) and MTT ( r 2 0.45, P <0.0001). Conclusions: In real-world deployment during a Code Brain protocol in a busy PSC, acute imaging with CTP did not predict DWI lesions on brain MRI with sufficient accuracy. In patients with large lesions the predictive value was better.

Download Full-text