Detecting CTP truncation artifacts in acute stroke imaging from the arterial input and the vascular output functions.

BACKGROUND: Current guidelines for CT perfusion (CTP) in acute stroke suggest acquiring scans with a minimal duration of 60-70 s. But even then, CTP analysis can be affected by truncation artifacts. Conversely, shorter acquisitions are still widely used in clinical practice and may, sometimes, be sufficient to reliably estimate lesion volumes. We aim to devise an automatic method that detects scans affected by truncation artifacts. METHODS: Shorter scan durations are simulated from the ISLES'18 dataset by consecutively removing the last CTP time-point until reaching a 10 s duration. For each truncated series, perfusion lesion volumes are quantified and used to label the series as unreliable if the lesion volumes considerably deviate from the original untruncated ones. Afterwards, nine features from the arterial input function (AIF) and the vascular output function (VOF) are derived and used to fit machine-learning models with the goal of detecting unreliably truncated scans. Methods are compared against a baseline classifier solely based on the scan duration, which is the current clinical standard. The ROC-AUC, precision-recall AUC and the F1-score are measured in a 5-fold cross-validation setting. RESULTS: The best performing classifier obtained an ROC-AUC of 0.982, precision-recall AUC of 0.985 and F1-score of 0.938. The most important feature was the AIFcoverage, measured as the time difference between the scan duration and the AIF peak. When using the AIFcoverage to build a single feature classifier, an ROC-AUC of 0.981, precision-recall AUC of 0.984 and F1-score of 0.932 were obtained. In comparison, the baseline classifier obtained an ROC-AUC of 0.954, precision-recall AUC of 0.958 and F1-Score of 0.875. CONCLUSIONS: Machine learning models fed with AIF and VOF features accurately detected unreliable stroke lesion measurements due to insufficient acquisition duration. The AIFcoverage was the most predictive feature of truncation and identified unreliable short scans almost as good as machine learning. We conclude that AIF/VOF based classifiers are more accurate than the scans' duration for detecting truncation. These methods could be transferred to perfusion analysis software in order to increase the interpretability of CTP outputs.

ISLES 2022: A multi-center magnetic resonance imaging stroke lesion segmentation dataset.

Magnetic resonance imaging (MRI) is an important imaging modality in stroke. Computer based automated medical image processing is increasingly finding its way into clinical routine. The Ischemic Stroke Lesion Segmentation (ISLES) challenge is a continuous effort to develop and identify benchmark methods for acute and sub-acute ischemic stroke lesion segmentation. Here we introduce an expert-annotated, multicenter MRI dataset for segmentation of acute to subacute stroke lesions ( https://doi.org/10.5281/zenodo.7153326 ). This dataset comprises 400 multi-vendor MRI cases with high variability in stroke lesion size, quantity and location. It is split into a training dataset of n = 250 and a test dataset of n = 150. All training data is publicly available. The test dataset will be used for model validation only and will not be released to the public. This dataset serves as the foundation of the ISLES 2022 challenge ( https://www.isles-challenge.org/ ) with the goal of finding algorithmic methods to enable the development and benchmarking of automatic, robust and accurate segmentation methods for ischemic stroke.

Prediction of Stroke Infarct Growth Rates by Baseline Perfusion Imaging.

BACKGROUND AND PURPOSE: Computed tomography perfusion imaging allows estimation of tissue status in patients with acute ischemic stroke. We aimed to improve prediction of the final infarct and individual infarct growth rates using a deep learning approach. METHODS: We trained a deep neural network to predict the final infarct volume in patients with acute stroke presenting with large vessel occlusions based on the native computed tomography perfusion images, time to reperfusion and reperfusion status in a derivation cohort (MR CLEAN trial [Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands]). The model was internally validated in a 5-fold cross-validation and externally in an independent dataset (CRISP study [CT Perfusion to Predict Response to Recanalization in Ischemic Stroke Project]). We calculated the mean absolute difference between the predictions of the deep learning model and the final infarct volume versus the mean absolute difference between computed tomography perfusion imaging processing by RAPID software (iSchemaView, Menlo Park, CA) and the final infarct volume. Next, we determined infarct growth rates for every patient. RESULTS: We included 127 patients from the MR CLEAN (derivation) and 101 patients of the CRISP study (validation). The deep learning model improved final infarct volume prediction compared with the RAPID software in both the derivation, mean absolute difference 34.5 versus 52.4 mL, and validation cohort, 41.2 versus 52.4 mL (P<0.01). We obtained individual infarct growth rates enabling the estimation of final infarct volume based on time and grade of reperfusion. CONCLUSIONS: We validated a deep learning-based method which improved final infarct volume estimations compared with classic computed tomography perfusion imaging processing. In addition, the deep learning model predicted individual infarct growth rates which could enable the introduction of tissue clocks during the management of acute stroke.

AIFNet: Automatic vascular function estimation for perfusion analysis using deep learning.

Perfusion imaging is crucial in acute ischemic stroke for quantifying the salvageable penumbra and irreversibly damaged core lesions. As such, it helps clinicians to decide on the optimal reperfusion treatment. In perfusion CT imaging, deconvolution methods are used to obtain clinically interpretable perfusion parameters that allow identifying brain tissue abnormalities. Deconvolution methods require the selection of two reference vascular functions as inputs to the model: the arterial input function (AIF) and the venous output function, with the AIF as the most critical model input. When manually performed, the vascular function selection is time demanding, suffers from poor reproducibility and is subject to the professionals' experience. This leads to potentially unreliable quantification of the penumbra and core lesions and, hence, might harm the treatment decision process. In this work we automatize the perfusion analysis with AIFNet, a fully automatic and end-to-end trainable deep learning approach for estimating the vascular functions. Unlike previous methods using clustering or segmentation techniques to select vascular voxels, AIFNet is directly optimized at the vascular function estimation, which allows to better recognise the time-curve profiles. Validation on the public ISLES18 stroke database shows that AIFNet almost reaches inter-rater performance for the vascular function estimation and, subsequently, for the parameter maps and core lesion quantification obtained through deconvolution. We conclude that AIFNet has potential for clinical transfer and could be incorporated in perfusion deconvolution software.

Predicting Infarct Core From Computed Tomography Perfusion in Acute Ischemia With Machine Learning: Lessons From the ISLES Challenge.

BACKGROUND AND PURPOSE: The ISLES challenge (Ischemic Stroke Lesion Segmentation) enables globally diverse teams to compete to develop advanced tools for stroke lesion analysis with machine learning. Detection of irreversibly damaged tissue on computed tomography perfusion (CTP) is often necessary to determine eligibility for late-time-window thrombectomy. Therefore, the aim of ISLES-2018 was to segment infarcted tissue on CTP based on diffusion-weighted imaging as a reference standard. METHODS: The data, from 4 centers, consisted of 103 cases of acute anterior circulation large artery occlusion stroke who underwent diffusion-weighted imaging rapidly after CTP. Diffusion-weighted imaging lesion segmentation was performed manually and acted as a reference standard. The data were separated into 63 cases for training and 40 for testing, upon which quality metrics (dice score coefficient, Hausdorff distance, absolute lesion volume difference, etc) were computed to rank methods based on their overall performance. RESULTS: Twenty-four different teams participated in the challenge. Median time to CTP was 185 minutes (interquartile range, 180-238), the time between CTP and magnetic resonance imaging was 36 minutes (interquartile range, 25-79), and the median infarct lesion size was 15.2 mL (interquartile range, 5.7-45). The best performance for Dice score coefficient and absolute volume difference were 0.51 and 10.1 mL, respectively, from different teams. Based on the ranking criteria, the top team's algorithm demonstrated for average Dice score coefficient and average absolute volume difference 0.51 and 10.2 mL, respectively, outperforming the conventional threshold-based method (dice score coefficient, 0.3; volume difference, 15.3). Diverse algorithms were used, almost all based on deep learning, with top-ranked approaches making use of the raw perfusion data as well as methods to synthetically generate complementary information to boost prediction performance. CONCLUSIONS: Machine learning methods may predict infarcted tissue from CTP with improved accuracy compared with threshold-based methods used in clinical routine. This dataset will remain public and can be used to test improvement in algorithms over time.

Theoretical analysis and experimental validation of volume bias of soft Dice optimized segmentation maps in the context of inherent uncertainty.

The clinical interest is often to measure the volume of a structure, which is typically derived from a segmentation. In order to evaluate and compare segmentation methods, the similarity between a segmentation and a predefined ground truth is measured using popular discrete metrics, such as the Dice score. Recent segmentation methods use a differentiable surrogate metric, such as soft Dice, as part of the loss function during the learning phase. In this work, we first briefly describe how to derive volume estimates from a segmentation that is, potentially, inherently uncertain or ambiguous. This is followed by a theoretical analysis and an experimental validation linking the inherent uncertainty to common loss functions for training CNNs, namely cross-entropy and soft Dice. We find that, even though soft Dice optimization leads to an improved performance with respect to the Dice score and other measures, it may introduce a volume bias for tasks with high inherent uncertainty. These findings indicate some of the method's clinical limitations and suggest doing a closer ad-hoc volume analysis with an optional re-calibration step.

Variations in the Circle of Willis in a large population sample using 3D TOF angiography: The Tromsø Study.

The main arteries that supply blood to the brain originate from the Circle of Willis (CoW). The CoW exhibits considerable anatomical variations which may have clinical importance, but the variability is insufficiently characterised in the general population. We assessed the anatomical variability of CoW variants in a community-dwelling sample (N = 1,864, 874 men, mean age = 65.4, range 40-87 years), and independent and conditional frequencies of the CoW's artery segments. CoW segments were classified as present or missing/hypoplastic (w/1mm diameter threshold) on 3T time-of-flight magnetic resonance angiography images. We also examined whether age and sex were associated with CoW variants. We identified 47 unique CoW variants, of which five variants constituted 68.5% of the sample. The complete variant was found in 11.9% of the subjects, and the most common variant (27.8%) was missing both posterior communicating arteries. Conditional frequencies showed patterns of interdependence across most missing segments in the CoW. CoW variants were associated with mean-split age (P = .0147), and there was a trend showing more missing segments with increasing age. We found no association with sex (P = .0526). Our population study demonstrated age as associated with CoW variants, suggesting reduced collateral capacity with older age.

Whole liver segmentation based on deep learning and manual adjustment for clinical use in SIRT.

PURPOSE: In selective internal radiation therapy (SIRT), an accurate total liver segmentation is required for activity prescription and absorbed dose calculation. Our goal was to investigate the feasibility of using automatic liver segmentation based on a convolutional neural network (CNN) for CT imaging in SIRT, and the ability of CNN to reduce inter-observer variability of the segmentation. METHODS: A multi-scale CNN was modified for liver segmentation for SIRT patients. The CNN model was trained with 139 datasets from three liver segmentation challenges and 12 SIRT patient datasets from our hospital. Validation was performed on 13 SIRT datasets and 12 challenge datasets. The model was tested on 40 SIRT datasets. One expert manually delineated the livers and adjusted the liver segmentations from CNN for 40 test SIRT datasets. Another expert performed the same tasks for 20 datasets randomly selected from the 40 SIRT datasets. The CNN segmentations were compared with the manual and adjusted segmentations from the experts. The difference between the manual segmentations was compared with the difference between the adjusted segmentations to investigate the inter-observer variability. Segmentation difference was evaluated through dice similarity coefficient (DSC), volume ratio (RV), mean surface distance (MSD), and Hausdorff distance (HD). RESULTS: The CNN segmentation achieved a median DSC of 0.94 with the manual segmentation and of 0.98 with the manually corrected CNN segmentation, respectively. The DSC between the adjusted segmentations is 0.98, which is 0.04 higher than the DSC between the manual segmentations. CONCLUSION: The CNN model achieved good liver segmentations on CT images of good image quality, with relatively normal liver shapes and low tumor burden. 87.5% of the 40 CNN segmentations only needed slight adjustments for clinical use. However, the trained model failed on SIRT data with low dose or contrast, lesions with large density difference from their surroundings, and abnormal liver position and shape. The abovementioned scenarios were not adequately represented in the training data. Despite this limitation, the current CNN is already a useful clinical tool which improves inter-observer agreement and therefore contributes to the standardization of the dosimetry. A further improvement is expected when the CNN will be trained with more data from SIRT patients.

Prediction of final infarct volume from native CT perfusion and treatment parameters using deep learning.

CT Perfusion (CTP) imaging has gained importance in the diagnosis of acute stroke. Conventional perfusion analysis performs a deconvolution of the measurements and thresholds the perfusion parameters to determine the tissue status. We pursue a data-driven and deconvolution-free approach, where a deep neural network learns to predict the final infarct volume directly from the native CTP images and metadata such as the time parameters and treatment. This would allow clinicians to simulate various treatments and gain insight into predicted tissue status over time. We demonstrate on a multicenter dataset that our approach is able to predict the final infarct and effectively uses the metadata. An ablation study shows that using the native CTP measurements instead of the deconvolved measurements improves the prediction.

ISLES 2016 and 2017-Benchmarking Ischemic Stroke Lesion Outcome Prediction Based on Multispectral MRI.

Performance of models highly depend not only on the used algorithm but also the data set it was applied to. This makes the comparison of newly developed tools to previously published approaches difficult. Either researchers need to implement others' algorithms first, to establish an adequate benchmark on their data, or a direct comparison of new and old techniques is infeasible. The Ischemic Stroke Lesion Segmentation (ISLES) challenge, which has ran now consecutively for 3 years, aims to address this problem of comparability. ISLES 2016 and 2017 focused on lesion outcome prediction after ischemic stroke: By providing a uniformly pre-processed data set, researchers from all over the world could apply their algorithm directly. A total of nine teams participated in ISLES 2015, and 15 teams participated in ISLES 2016. Their performance was evaluated in a fair and transparent way to identify the state-of-the-art among all submissions. Top ranked teams almost always employed deep learning tools, which were predominately convolutional neural networks (CNNs). Despite the great efforts, lesion outcome prediction persists challenging. The annotated data set remains publicly available and new approaches can be compared directly via the online evaluation system, serving as a continuing benchmark (www.isles-challenge.org).

Smaller caliber of the internal carotid artery in patients with ipsilateral aplasia of the A1 segment of the anterior cerebral artery: a study with CTA.

To measure the diameter and the transsectional area of the internal carotid arteries (ICA) on CT Angiography (CTA) in patients with aplasia of the A1-segment of the ACA (A1) and in patients with symmetrical A1, the mean diameter and area of the ICA on both sides were measured at a level of 2 cm below the skull base with a commercially available CT software in 41 consecutive patients with aplasia of A1 observed during a 12-month period on CTA and in 41 control patients with symmetrical A1. The mean diameter of the ipsilateral ICA was 3.83 ± 0.60 mm versus 4.86 ± 0.60 mm as mean diameter of the contralateral ICA and versus 4.40 ± 0.60 mm as mean diameter of both ICAs in the control group of patients. The mean area of the ipsilateral ICA was 11.58 ± 3.80 mm2 versus 18. 82 ± 7.39 mm2 as mean area of the contralateral ICA and versus 15.29 ± 4.42 mm2 as mean area of both ICA in the control group of patients. These differences are statistically highly significant. In patients with symmetrical A1, there was no statistical difference between the diameter or area of both internal carotid arteries. In conclusion, in patients with aplasia of A1, the ipsilateral diameter and area of the cervical ICA is smaller than the diameter and area of the contralateral ICA and smaller than the diameter and area of both internal carotid arteries in patients with symmetrical A1.

Automated DWI analysis can identify patients within the thrombolysis time window of 4.5 hours.

OBJECTIVE: To develop an automated model based on diffusion-weighted imaging (DWI) to detect patients within 4.5 hours after stroke onset and compare this method to the visual DWI-FLAIR (fluid-attenuated inversion recovery) mismatch. METHODS: We performed a subanalysis of the "DWI-FLAIR mismatch for the identification of patients with acute ischemic stroke within 4.5 hours of symptom onset" (PRE-FLAIR) and the "AX200 for ischemic stroke" (AXIS 2) trials. We developed a prediction model with data from the PRE-FLAIR study by backward logistic regression with the 4.5-hour time window as dependent variable and the following explanatory variables: age and median relative DWI (rDWI) signal intensity, interquartile range (IQR) rDWI signal intensity, and volume of the core. We obtained the accuracy of the model to predict the 4.5-hour time window and validated our findings in an independent cohort from the AXIS 2 trial. We compared the receiver operating characteristic curve to the visual DWI-FLAIR mismatch. RESULTS: In the derivation cohort of 118 patients, we retained the IQR rDWI as explanatory variable. A threshold of 0.39 was most optimal in selecting patients within 4.5 hours after stroke onset resulting in a sensitivity of 76% and specificity of 63%. The accuracy was validated in an independent cohort of 200 patients. The predictive value of the area under the curve of 0.72 (95% confidence interval 0.64-0.80) was similar to the visual DWI-FLAIR mismatch (area under the curve = 0.65; 95% confidence interval 0.58-0.72; p for difference = 0.18). CONCLUSIONS: An automated analysis of DWI performs at least as good as the visual DWI-FLAIR mismatch in selecting patients within the 4.5-hour time window.

ISLES 2015 - A public evaluation benchmark for ischemic stroke lesion segmentation from multispectral MRI.

Ischemic stroke is the most common cerebrovascular disease, and its diagnosis, treatment, and study relies on non-invasive imaging. Algorithms for stroke lesion segmentation from magnetic resonance imaging (MRI) volumes are intensely researched, but the reported results are largely incomparable due to different datasets and evaluation schemes. We approached this urgent problem of comparability with the Ischemic Stroke Lesion Segmentation (ISLES) challenge organized in conjunction with the MICCAI 2015 conference. In this paper we propose a common evaluation framework, describe the publicly available datasets, and present the results of the two sub-challenges: Sub-Acute Stroke Lesion Segmentation (SISS) and Stroke Perfusion Estimation (SPES). A total of 16 research groups participated with a wide range of state-of-the-art automatic segmentation algorithms. A thorough analysis of the obtained data enables a critical evaluation of the current state-of-the-art, recommendations for further developments, and the identification of remaining challenges. The segmentation of acute perfusion lesions addressed in SPES was found to be feasible. However, algorithms applied to sub-acute lesion segmentation in SISS still lack accuracy. Overall, no algorithmic characteristic of any method was found to perform superior to the others. Instead, the characteristics of stroke lesion appearances, their evolution, and the observed challenges should be studied in detail. The annotated ISLES image datasets continue to be publicly available through an online evaluation system to serve as an ongoing benchmarking resource (www.isles-challenge.org).

Simultaneous segmentation and anatomical labeling of the cerebral vasculature.

We present a novel algorithm for the simultaneous segmentation and anatomical labeling of the cerebral vasculature. Unlike existing approaches that first attempt to obtain a good segmentation and then perform labeling, we optimize for both by simultaneously taking into account the image evidence and the prior knowledge about the geometry and connectivity of the vasculature. This is achieved by first constructing an overcomplete graph capturing the vasculature, and then selecting and labeling the subset of edges that most likely represents the true vasculature. We formulate the latter problem as an Integer Program (IP), which can be solved efficiently to provable optimality. We evaluate our approach on a publicly available dataset of 50 cerebral MRA images, and demonstrate that it compares favorably against state-of-the-art methods.

A reliable and time-saving semiautomatic spike-template-based analysis of interictal EEG-fMRI.

OBJECTIVE: A prerequisite for the implementation of interictal electroencephalography-correlated functional magnetic resonance imaging (EEG-fMRI) in the presurgical work-up for epilepsy surgery is straightforward processing. We propose a new semi-automatic method as alternative for the challenging and time-consuming visual spike identification. METHODS: Our method starts from a patient-specific spike-template, built by averaging spikes recorded on the EEG outside the scanner. Spatiotemporal cross-correlations between the template and the EEG measured during fMRI were calculated. To minimize false-positive detections, this time course of cross-correlations was binarized by means of a spike-template-specific threshold determined in healthy controls. To inform our model for statistical parametric mapping, this binarized regressor was convolved with the canonical hemodynamic response function. We validated our "template-based" method in 21 adult patients with refractory focal epilepsy with a well-defined epileptogenic zone and interictal spikes during EEG-fMRI. Sensitivity and specificity for detecting the epileptogenic zone were calculated and represented in receiver operating characteristic (ROC) curves. Our approach was compared with a previously proposed semiautomatic "topography-based" method that used the topographic amplitude distribution of spikes as a starting point for correlation-based fitting. RESULTS: Good diagnostic performance could be reached with our template-based method. The optimal area under the ROC curve was 0.77. Diagnostic performance of the topography-based method was overall low. SIGNIFICANCE: Our new template-based method is more standardized and time-saving than visual spike identification on intra-scanner EEG recordings, and preserves good diagnostic performance for detecting the epileptogenic zone.

Simultaneous segmentation and anatomical labeling of the cerebral vasculature.

We present a novel algorithm for the simultaneous segmentation and anatomical labeling of the cerebral vasculature. The method first constructs an overcomplete graph capturing the vasculature. It then selects and labels the subset of edges that most likely represents the true vasculature. Unlike existing approaches that first attempt to obtain a good segmentation and then perform labeling, we jointly optimize for both by simultaneously taking into account the image evidence and the prior knowledge about the geometry and connectivity of the vasculature. This results in an Integer Program (IP), which we solve optimally using a branch-and-cut algorithm. We evaluate our approach on a public dataset of 50 cerebral MRA images, and demonstrate that it compares favorably against state-of-the-art methods.

Anatomical labeling of the circle of willis using maximum a posteriori graph matching.

A new method for anatomically labeling the vasculature is presented and applied to the Circle of Willis. Our method converts the segmented vasculature into a graph that is matched with an annotated graph atlas in a maximum a posteriori (MAP) way. The MAP matching is formulated as a quadratic binary programming problem which can be solved efficiently. Unlike previous methods, our approach can handle non tree-like vasculature and large topological differences. The method is evaluated in a leave-one-out test on MRA of 30 subjects where it achieves a sensitivity of 93% and a specificity of 85% with an average error of 1.5 mm on matching bifurcations in the vascular graph.