Application of Machine Learning to Automated Analysis of Cerebral Edema in Large Cohorts of Ischemic Stroke Patients

Introduction: Cerebral edema with resultant mass effect is a potentially fatal consequence of ischemic stroke, but early and sensitive biomarkers of brain tissue compression are lacking. To quantify brain mass effect, we developed a novel, automated segmentation method to delineate CSF spaces in CT images from ischemic stroke patients. Methods: CTs from sixteen acute ischemic stroke patients (median NIHSS 16.5, median age 61.5 yrs, 14-92 hrs after stroke onset) were included after informed consent was obtained. After infarction, conventional CSF segmentation using Hounsfield unit (HU) thresholding is suboptimal due to infarct hypodensity. Utilizing manually delineated infarct and CSF spaces as training samples, we augmented conventional HU threshold segmentation with level sets, sparse regression and random forest segmentation methods. Using leave-one-out cross-validation, the combined approach was compared to HU thresholding using Dice ratios (a measure of the overlap between the segmented and the ground-truth CSF spaces). Results: Shown is an example of a CT brain slice segmented by HU thresholding and the combined strategy: false negative (red), false positive (green), and true positive (yellow). The Dice ratios for HU thresholding and the combined approaches were 58.2±16.3% and 68.9±14.6%, respectively, demonstrating the significantly improved performance for the combined strategy (p=0.0014). Conclusions: We have developed an advanced image segmentation strategy to delineate CSF spaces which outperforms conventional HU thresholding. An automated CSF segmentation strategy will permit quantification of cerebral edema in a large population of stroke patients, as required for genetic studies, for example.

Download Full-text

Abstract TP215: Machine Learning Analysis of Gene Expression Data and Predictive Stroke Outcomes in Acute Ischemic Stroke

Stroke ◽

10.1161/str.51.suppl_1.tp215 ◽

2020 ◽

Vol 51 (Suppl_1) ◽

Author(s):

Sarah R Martha ◽

Qiang Cheng ◽

Liyu Gong ◽

Lisa Collier ◽

Stephanie Davis ◽

...

Keyword(s):

Gene Expression ◽

Machine Learning ◽

Ischemic Stroke ◽

Acute Ischemic Stroke ◽

Mechanical Thrombectomy ◽

Stroke Patients ◽

Stroke Outcomes ◽

Inflammatory Genes ◽

Patient Demographics ◽

Edema Volume

Background and Purpose: The ability to predict ischemic stroke outcomes in the first day of admission could be vital for patient counseling, rehabilitation, and care planning. The Blood and Clot Thrombectomy Registry and Collaboration (BACTRAC; clinicaltrials.gov NCT03153683) collects blood samples distal and proximal to the intracranial thrombus during mechanical thrombectomy. These samples are a novel resource in evaluating acute gene expression changes at the time of ischemic stroke. The purpose of this study was to identify inflammatory genes and patient demographics that are predictive of stroke outcomes (infarct and/or edema volume) in acute ischemic stroke patients. Methods: The BACTRAC study is a non-probability, convenience sampling of subjects (≥ 18 year olds) treated with mechanical thrombectomy for emergent large vessel occlusion. We evaluated relative concentrations of mRNA for gene expression in 84 inflammatory molecules in static blood distal and proximal to the intracranial thrombus from adults who underwent thrombectomy. We employed a machine learning method, Random Forest, utilizing the first set of enrolled subjects, to predict which inflammatory genes and patient demographics were important features for infarct and edema volumes. Results: We analyzed the first 28 subjects (age = 66 ± 15.48, 11 males) in the BACTRAC registry. Results from machine learning analyses demonstrate that the genes CCR4, IFNA2, IL9, CXCL3, Age, DM, IL7, CCL4, BMI, IL5, CCR3, TNF, and IL27 predict infarct volume. The genes IFNA2, IL5, CCL11, IL17C, CCR4, IL9, IL7, CCR3, IL27, DM, and CSF2 predict edema volume. There is an intersection of genes CCR4, IFNA2, IL9, IL7, IL5, CCR3 to both infarct and edema volumes. Overall, these genes depicts a microenvironment for chemoattraction and proliferation of autoimmune cells, particularly Th2 cells and neutrophils. Conclusions: Machine learning algorithms can be employed to develop predictive biomarker signatures for stroke outcomes in ischemic stroke patients, particularly in regard to identifying acute gene expression changes that occur during stroke.

Download Full-text

Abstract WMP23: Automated ASPECTS Scoring of CT Scans for Acute Ischemic Stroke Patients Using Machine Learning

Stroke ◽

10.1161/str.49.suppl_1.wmp23 ◽

2018 ◽

Vol 49 (Suppl_1) ◽

Author(s):

Hulin Kuang ◽

Ericka Teleg ◽

Mohamed Najm ◽

Alexis T Wilson ◽

Sung I Sohn ◽

...

Keyword(s):

Machine Learning ◽

Ischemic Stroke ◽

Acute Ischemic Stroke ◽

Ct Scans ◽

Stroke Patients

Download Full-text

Machine learning volumetry of ischemic brain lesions on CT after thrombectomy—prospective diagnostic accuracy study in ischemic stroke patients

Neuroradiology ◽

10.1007/s00234-020-02419-7 ◽

2020 ◽

Vol 62 (10) ◽

pp. 1239-1245

Author(s):

Jiri Kral ◽

Martin Cabal ◽

Linda Kasickova ◽

Jaroslav Havelka ◽

Tomas Jonszta ◽

...

Keyword(s):

Machine Learning ◽

Ischemic Stroke ◽

Diagnostic Accuracy ◽

Brain Lesions ◽

Diagnostic Accuracy Study ◽

Stroke Patients ◽

Ischemic Brain ◽

Accuracy Study

Download Full-text

Using Machine Learning to Improve the Prediction of Functional Outcome in Ischemic Stroke Patients

IEEE/ACM Transactions on Computational Biology and Bioinformatics ◽

10.1109/tcbb.2018.2811471 ◽

2018 ◽

Vol 15 (6) ◽

pp. 1953-1959 ◽

Cited By ~ 13

Author(s):

Miguel Monteiro ◽

Ana Catarina Fonseca ◽

Ana Teresa Freitas ◽

Teresa Pinho e Melo ◽

Alexandre P. Francisco ◽

...

Keyword(s):

Machine Learning ◽

Ischemic Stroke ◽

Functional Outcome ◽

Stroke Patients

Download Full-text

Treatment Efficacy Analysis in Acute Ischemic Stroke Patients Using In Silico Modeling Based on Machine Learning: A Proof-of-Principle

Biomedicines ◽

10.3390/biomedicines9101357 ◽

2021 ◽

Vol 9 (10) ◽

pp. 1357

Author(s):

Anthony Winder ◽

Matthias Wilms ◽

Jens Fiehler ◽

Nils D. Forkert

Keyword(s):

Machine Learning ◽

Ischemic Stroke ◽

Treatment Group ◽

Acute Ischemic Stroke ◽

In Silico ◽

Mechanical Thrombectomy ◽

Intravenous Thrombolysis ◽

Stroke Patients ◽

New Devices

Interventional neuroradiology is characterized by engineering- and experience-driven device development with design improvements every few months. However, clinical validation of these new devices requires lengthy and expensive randomized controlled trials. This contribution proposes a machine learning-based in silico study design to evaluate new devices more quickly with a small sample size. Acute diffusion- and perfusion-weighted MRI, segmented one-week follow-up imaging, and clinical variables were available for 90 acute ischemic stroke patients. Three treatment option-specific random forest models were trained to predict the one-week follow-up lesion segmentation for (1) patients successfully recanalized using intra-arterial mechanical thrombectomy, (2) patients successfully recanalized using intravenous thrombolysis, and (3) non-recanalizing patients as an analogue for conservative treatment for each patient in the sample, independent of the true group membership. A repeated-measures analysis of the three predicted follow-up lesions for each patient revealed significantly larger lesions for the non-recanalizing group compared to the successful intravenous thrombolysis treatment group, which in turn showed significantly larger lesions compared to the successful mechanical thrombectomy treatment group (p < 0.001). A groupwise comparison of the true follow-up lesions for the three treatment options showed the same trend but did not reach statistical significance (p = 0.19). We conclude that the proposed machine learning-based in silico trial design leads to clinically feasible results and can support new efficacy studies by providing additional power and potential early intermediate results.

Download Full-text

A Nomogram Model to Predict Malignant Cerebral Edema in Ischemic Stroke Patients Treated with Endovascular Thrombectomy: An Observational Study

Neuropsychiatric Disease and Treatment ◽

10.2147/ndt.s279303 ◽

2020 ◽

Vol Volume 16 ◽

pp. 2913-2920

Author(s):

Mingyang Du ◽

Xianjun Huang ◽

Shun Li ◽

Lili Xu ◽

Bin Yan ◽

...

Keyword(s):

Ischemic Stroke ◽

Observational Study ◽

Cerebral Edema ◽

Stroke Patients ◽

Endovascular Thrombectomy

Download Full-text

Abstract P308: Machine Learning Based Models of the 30-Day Readmission for Stroke Patients Using Electronic Health Record Data

Stroke ◽

10.1161/str.52.suppl_1.p308 ◽

2021 ◽

Vol 52 (Suppl_1) ◽

Author(s):

Negar Darabi ◽

Niyousha Hosseinichimeh ◽

Anthony Noto ◽

Ramin Zand ◽

Vida Abedi

Keyword(s):

Machine Learning ◽

Ischemic Stroke ◽

Care Delivery ◽

Gradient Boosting ◽

Support Vector ◽

Stroke Patients ◽

Electronic Health Record Data ◽

Percutaneous Gastrostomy ◽

Extreme Gradient Boosting ◽

Electronic Health

Background: At a personalized level, identification of patients at higher risk of 30-day readmission and in need of special clinical attention could lower their chances of readmission. While at a system’s level, reducing hospital readmission improves the overall quality of care delivery and reduces the associated cost burden. Objective: To enhance understanding of the predictors of 30-day readmission after ischemic stroke and identify high-risk individuals. We aimed to compare the performance and the predictive power of machine learning-based methods and identify the best model. Method: The electronic health records (EHR) of acute ischemic stroke patients were extracted from two tertiary centers within the Geisinger Health System between January 1, 2015, and October 7, 2018. A total of 61 variables, including clinical variables, demographical characteristics, discharge status, and type of health insurance were used in this study. Patients were randomly split for model development (80%) and testing (20%). Random forest, gradient boosting machine, extreme gradient boosting (XGBoost), support vector machine, and logistic regression, were developed to predict the 30-day readmission after stroke. The models were evaluated based on the area under the curve (AUC), sensitivity, specificity, and positive predictive value (PPV). Results: A total of 3,184 patients with ischemic stroke (mean age: 71±13.90 years, men: 51.06%) were included in this study. From the 3,184, 301 (9.40%) were readmitted within 30-day. The best performance was obtained when XGBoost was used with ROSE-sampling. The AUC for the test set was 0.74 (95% CI: 0.64-0.78) with PPV of 0.43. The top four predictors of the 30-day readmission model were National Institutes of Health Stroke Scale score above 24, insert an indwelling urinary catheter, hypercoagulable state, and percutaneous gastrostomy. Conclusions: Machine learning model can be designed to predict 30-day readmission after stroke using structured data from EHR. Among the five algorithms analyzed, XGBoost had the best performance.

Download Full-text