Abstract P330: Optimizing Predictions of Infarct Core Using Machine Learning

Stroke ◽  
2021 ◽  
Vol 52 (Suppl_1) ◽  
Author(s):  
Rania Abdelkhaleq ◽  
Victor Lopez-Rivera ◽  
Sergio Salazar-Marioni ◽  
Songmi Lee ◽  
Youngran Kim ◽  
...  
Keyword(s):  
Machine Learning ◽  
Computed Tomography ◽  
Infarct Volume ◽  
Learning Model ◽  
Superior Performance ◽  
Infarct Core ◽  
Vessel Occlusion ◽  
Machine Learning Model ◽  
Selection For ◽  
Learning Analysis

Introduction: Evaluation of infarct core by advanced neuroimaging has facilitated patient selection for endovascular stroke therapy (EST), however the accuracy of machine-learning analysis compared to these modalities remains unexplored. We test the performance of computed tomography-Alberta Stroke Program Early Computed Tomography Score (CT- ASPECTS) vs. Computed Tomography Perfusion (CTP)-RAPID, vs. an extension of our novel machine-learning model, Deep Symmetry-sensitive Network (DeepSymNet [ref]), using the final infarct volume (FIV) in patients with rapid and successful endovascular reperfusion as the gold standard. Methods and Materials: We identified consecutive patients with large vessel occlusion acute ischemic stroke that underwent EST with TICI 2b/3 reperfusion. FIV was determined by volumetric measurements on 24-48h DWI MRI. The DeepSymNet algorithm combines symmetric and absolute brain representations and had been trained to predict CTP-RAPID core size from CTA source images acquired at presentation. Performance at predicting FIV was determined by Pearson’s correlation for CT- ASPECTS, CTP-RAPID, and DeepSymNet. Data are presented as median [IQR]. Results: Among the 76 patients that met inclusion criteria, 55.2% were male, the median age was 68 years [54-77], and 32.8% were White. 71% of the patients demonstrated an MCA occlusion, and 55% of all occlusions were left-sided. Median ASPECTS on presentation was 8 [7-8.5] and the median FIV was 10 mL [2-37]. ASPECTS, CTP-RAPID and DeepSymNet all correlated with FIV, with comparable performances from ASPECTS (R 2 =-0.398) and CTP-RAPID (R 2 =0.403) and superior performance by DeepSymNet (R 2 =-0.606)(Table). Conclusions: The DeepSymNet machine learning model analyzing CTA source images demonstrated superior performance to ASPECTS and CTP-RAPID in FIV prediction. These findings suggest machine learning models may provide improved predictions of infarct core and selection for EST.

scholarly journals Automated prediction of final infarct volume in patients with large-vessel occlusion acute ischemic stroke

2021 ◽  
Vol 51 (1) ◽  
pp. E13
Author(s):  
Rania Abdelkhaleq ◽  
Youngran Kim ◽  
Swapnil Khose ◽  
Peter Kan ◽  
Sergio Salazar-Marioni ◽  
...  
Keyword(s):  
Machine Learning ◽  
Ischemic Stroke ◽  
Acute Ischemic Stroke ◽  
Infarct Volume ◽  
Large Vessel Occlusion ◽  
Infarct Core ◽  
Vessel Occlusion ◽  
Automated Method ◽  
Machine Learning Model ◽  
Time Of Presentation

OBJECTIVE In patients with large-vessel occlusion (LVO) acute ischemic stroke (AIS), determinations of infarct size play a key role in the identification of candidates for endovascular stroke therapy (EVT). An accurate, automated method to quantify infarct at the time of presentation using widely available imaging modalities would improve screening for EVT. Here, the authors aimed to compare the performance of three measures of infarct core at presentation, including an automated method using machine learning. METHODS Patients with LVO AIS who underwent successful EVT at four comprehensive stroke centers were identified. Patients were included if they underwent concurrent noncontrast head CT (NCHCT), CT angiography (CTA), and CT perfusion (CTP) with Rapid imaging at the time of presentation, and MRI 24 to 48 hours after reperfusion. NCHCT scans were analyzed using the Alberta Stroke Program Early CT Score (ASPECTS) graded by neuroradiology or neurology expert readers. CTA source images were analyzed using a previously described machine learning model named DeepSymNet (DSN). Final infarct volume (FIV) was determined from diffusion-weighted MRI sequences using manual segmentation. The primary outcome was the performance of the three infarct core measurements (NCHCT-ASPECTS, CTA with DSN, and CTP-Rapid) to predict FIV, which was measured using area under the receiver operating characteristic (ROC) curve (AUC) analysis. RESULTS Among 76 patients with LVO AIS who underwent EVT and met inclusion criteria, the median age was 67 years (IQR 54–76 years), 45% were female, and 37% were White. The median National Institutes of Health Stroke Scale score was 16 (IQR 12–22), and the median NCHCT-ASPECTS on presentation was 8 (IQR 7–8). The median time between when the patient was last known to be well and arrival was 156 minutes (IQR 73–303 minutes), and between NCHCT/CTA/CTP to groin puncture was 73 minutes (IQR 54–81 minutes). The AUC was obtained at three different cutoff points: 10 ml, 30 ml, and 50 ml FIV. At the 50-ml FIV cutoff, the AUC of ASPECTS was 0.74; of CTP core volume, 0.72; and of DSN, 0.82. Differences in AUCs for the three predictors were not significant for the three FIV cutoffs. CONCLUSIONS In a cohort of patients with LVO AIS in whom reperfusion was achieved, determinations of infarct core at presentation by NCHCT-ASPECTS and a machine learning model analyzing CTA source images were equivalent to CTP in predicting FIV. These findings have suggested that the information to accurately predict infarct core in patients with LVO AIS was present in conventional imaging modalities (NCHCT and CTA) and accessible by machine learning methods.

scholarly journals Development and external validation of a deep learning-based computed tomography classification system for COVID-19

2021 ◽  
Author(s):  
Yuki KATAOKA
Keyword(s):  
Machine Learning ◽  
Computed Tomography ◽  
Interstitial Pneumonia ◽  
External Validation ◽  
Usual Interstitial Pneumonia ◽  
Evaluation Process ◽  
High Sensitivity ◽  
Learning Model ◽  
Machine Learning Model ◽  
Ablation Study

Rationale: Currently available machine learning models for diagnosing COVID-19 based on computed tomography (CT) images are limited due to concerns regarding methodological flaws or underlying biases in the evaluation process. Objectives: We aimed to develop and externally validate a novel machine learning model that can classify CT image findings as positive or negative for SARS-CoV-2 reverse transcription polymerase chain reaction (RT-PCR).Methods: We used 3128 images from a wide variety of two-gate data sources for the development and ablation study of the machine learning model. A total of 633 COVID-19 cases and 2295 non-COVID-19 cases were included in the study. We randomly divided cases into a development set and ablation set at a ratio of 8:2. For the ablation study, we used another dataset including 150 cases of interstitial pneumonia among non-COVID-19 images. For external validation, we used 893 images from 740 consecutive patients at 11 acute care hospitals suspected of having COVID-19 at the time of diagnosis. The dataset included 343 COVID-19 patients. The reference standard was RT-PCR.Result: In ablation study, using interstitial pneumonia images, the specificity of the model were 0.986 for usual interstitial pneumonia pattern, 0.820 for non-specific interstitial pneumonia pattern, 0.400 for organizing pneumonia pattern. In the external validation study, the sensitivity and specificity of the model were 0.869 and 0.432, respectively, at the low-level cutoff, and 0.724 and 0.721, respectively, at the high-level cutoff.Conclusions: Our machine learning model exhibited a high sensitivity in external validation datasets and may assist physicians to rule out COVID-19 diagnosis in a timely manner. Further studies are warranted to improve model specificity.

Abstract WMP15: Automated Volumetric Assessment of Infarct Core in Non-Contrast Computed Tomography in Patients With Acute Ischemic Stroke Secondary to Large Vessel Occlusion

Stroke ◽  
2020 ◽  
Vol 51 (Suppl_1) ◽  
Author(s):  
Alvaro Garcia-Tornel ◽  
Matias Deck ◽  
Marc Ribo ◽  
David Rodriguez-Luna ◽  
Jorge Pagola ◽  
...  
Keyword(s):  
Computed Tomography ◽  
Ischemic Stroke ◽  
Acute Ischemic Stroke ◽  
Functional Outcome ◽  
Infarct Volume ◽  
Large Vessel Occlusion ◽  
Infarct Core ◽  
Vessel Occlusion ◽  
Good Functional Outcome ◽  
Volumetric Assessment

Introduction: Perfusion imaging has emerged as an imaging tool to select patients with acute ischemic stroke (AIS) secondary to large vessel occlusion (LVO) for endovascular treatment (EVT). We aim to compare an automated method to assess the infarct ischemic core (IC) in Non-Contrast Computed Tomography (NCCT) with Computed Tomography Perfusion (CTP) imaging and its ability to predict functional outcome and final infarct volume (FIV). Methods: 494 patients with anterior circulation stroke treated with EVT were included. Volumetric assessment of IC in NCCT (eA-IC) was calculated using eASPECTS™ (Brainomix, Oxford). CTP was processed using availaible software considering CTP-IC as volume of Cerebral Blood Flow (CBF) <30% comparing with the contralateral hemisphere. FIV was calculated in patients with complete recanalization using a semiautomated method with a NCCT performed 48-72 hours after EVT. Complete recanalization was considered as modified Thrombolysis In Cerebral Ischemia (mTICI) ≥2B after EVT. Good functional outcome was defined as modified Rankin score (mRs) ≤2 at 90 days. Statistical analysis was performed to assess the correlation between EA-IC and CTP-IC and its ability to predict prognosis and FIV. Results: Median eA-IC and CTP-IC were 16 (IQR 7-31) and 8 (IQR 0-28), respectively. 419 patients (85%) achieved complete recanalization, and their median FIV was 17.5cc (IQR 5-52). Good functional outcome was achieved in 230 patients (47%). EA-IC and CTP-IC had moderate correlation between them (r=0.52, p<0.01) and similar correlation with FIV (r=0.52 and 0.51, respectively, p<0.01). Using ROC curves, both methods had similar performance in its ability to predict good functional outcome (EA-IC AUC 0.68 p<0.01, CTP-IC AUC 0.66 p<0.01). Multivariate analysis adjusted by confounding factors showed that eA-IC and CTP-IC predicted good functional outcome (for every 10cc and >40cc, OR 1.5, IC1.3-1.8, p<0.01 and OR 1.3, IC1.1-1.5, p<0.01, respectively). Conclusion: Automated volumetric assessment of infarct core in NCCT has similar performance predicting prognosis and final infarct volume than CTP. Prospective studies should evaluate a NCCT-core / vessel occlusion penumbra missmatch as an alternative method to select patients for EVT.

scholarly journals Text and Data Formatting for Machine Learning

2019 ◽  
Vol 9 (1) ◽  
pp. 2756-2760
Keyword(s):  
Machine Learning ◽  
Big Data ◽  
Data Analysis ◽  
Data Preprocessing ◽  
Learning Model ◽  
Poor Quality ◽  
Big Data Analysis ◽  
Model Quality ◽  
Machine Learning Model ◽  
Learning Analysis

Machine learning is a prominent tool for getting data from large amounts of information. Whereas a good amount of machine learning analysis has targeted on increasing the accuracy and potency of coaching and reasoning algorithms, there is less attention within the equally vital issues of observing the standard of information fed into the machine learning model. The standard of huge information is far away from good. Recent studies have shown that poor quality will bring serious errors to the result of big data analysis and this could have an effect on in making additional precise results from the information. Advantages of data preprocessing within the context of ML are advanced detection of errors, model-quality improves by the usage of better data, savings in engineering hours to debug issues

scholarly journals The trauma severity model: An ensemble machine learning approach to risk prediction

2017 ◽  
Author(s):  
Michael T. Gorczyca ◽  
Nicole C. Toscano ◽  
Julius D. Cheng
Keyword(s):  
Machine Learning ◽  
Risk Prediction ◽  
Generalized Linear Models ◽  
Linear Models ◽  
Learning Model ◽  
Superior Performance ◽  
List Type ◽  
Trauma Severity ◽  
Ensemble Machine Learning ◽  
Machine Learning Model

AbstractStatistical theory indicates that a flexible model can attain a lower generalization error than an inflexible model, provided that the setting is appropriate. This is highly relevant in the context of mortality risk prediction for trauma patients, as researchers have focused exclusively on the use of generalized linear models for risk prediction, and generalized linear models may be too inflexible to capture the potentially complex relationships in trauma data. Due to this, we propose a machine learning model, the Trauma Severity Model (TSM), for risk prediction. In order to validate TSM’s performance, this study compares TSM to three established risk prediction models: the Bayesian Logistic Injury Severity Score, the Harborview Assessment for Risk of Mortality, and the Trauma Mortality Prediction Model. Our results indicate that TSM has superior performance, and thereby provides improved risk prediction.Highlights:We propose an ensemble machine learning model for trauma risk prediction.A hyper-parameter search scheme is proposed for model development.We compare our model to established models for trauma risk prediction.Our model improves over established models for each performance metric considered.

scholarly journals A Novel Machine Learning Model Developed to Assist in Patient Selection for Outpatient Total Shoulder Arthroplasty

2019 ◽  
Vol 28 (13) ◽  
pp. e580-e585 ◽  
Author(s):  
Dustin R. Biron ◽  
Ishan Sinha ◽  
Justin E. Kleiner ◽  
Dilum P. Aluthge ◽  
Avi D. Goodman ◽  
...  
Keyword(s):  
Machine Learning ◽  
Patient Selection ◽  
Shoulder Arthroplasty ◽  
Total Shoulder Arthroplasty ◽  
Learning Model ◽  
Machine Learning Model ◽  
Selection For

scholarly journals Identifying main and interaction effects of risk factors to predict intensive care admission in patients hospitalized with COVID-19: a retrospective cohort study in Hong Kong

2020 ◽  
Author(s):  
Jiandong Zhou ◽  
Gary Tse ◽  
Sharen Lee ◽  
Tong Liu ◽  
William KK Wu ◽  
...  
Keyword(s):  
Machine Learning ◽  
Risk Factors ◽  
Hong Kong ◽  
Intensive Care ◽  
Cell Count ◽  
Learning Model ◽  
Older Age ◽  
Superior Performance ◽  
Icu Admission ◽  
Machine Learning Model

ABSTRACTBackgroundThe coronavirus disease 2019 (COVID-19) has become a pandemic, placing significant burdens on the healthcare systems. In this study, we tested the hypothesis that a machine learning approach incorporating hidden nonlinear interactions can improve prediction for Intensive care unit (ICU) admission.MethodsConsecutive patients admitted to public hospitals between 1st January and 24th May 2020 in Hong Kong with COVID-19 diagnosed by RT-PCR were included. The primary endpoint was ICU admission.ResultsThis study included 1043 patients (median age 35 (IQR: 32-37; 54% male). Nineteen patients were admitted to ICU (median hospital length of stay (LOS): 30 days, median ICU LOS: 16 days). ICU patients were more likely to be prescribed angiotensin converting enzyme inhibitors/angiotensin receptor blockers, anti-retroviral drugs lopinavir/ritonavir and remdesivir, ribavirin, steroids, interferon-beta and hydroxychloroquine. Significant predictors of ICU admission were older age, male sex, prior coronary artery disease, respiratory diseases, diabetes, hypertension and chronic kidney disease, and activated partial thromboplastin time, red cell count, white cell count, albumin and serum sodium. A tree-based machine learning model identified most informative characteristics and hidden interactions that can predict ICU admission. These were: low red cells with 1) male, 2) older age, 3) low albumin, 4) low sodium or 5) prolonged APTT. A five-fold cross validation confirms superior performance of this model over baseline models including XGBoost, LightGBM, random forests, and multivariate logistic regression.ConclusionsA machine learning model including baseline risk factors and their hidden interactions can accurately predict ICU admission in COVID-19.

Sign in / Sign up

Export Citation Format

Share Document