A Machine Learning Approach to First Pass Reperfusion in Mechanical Thrombectomy: Prediction and Feature Analysis

A Deep Learning Approach to Predict Recanalization First-Pass Effect following Mechanical Thrombectomy in Patients with Acute Ischemic Stroke

BACKGROUND AND PURPOSE

Following endovascular thrombectomy in patients with large-vessel occlusion stroke, successful recanalization from 1 attempt, known as the first-pass effect, has correlated favorably with long-term outcomes. Pretreatment imaging may contain information that can be used to predict the first-pass effect. Recently, applications of machine learning models have shown promising results in predicting recanalization outcomes, albeit requiring manual segmentation. In this study, we sought to construct completely automated methods using deep learning to predict the first-pass effect from pretreatment CT and MR imaging.

MATERIALS AND METHODS

Our models were developed and evaluated using a cohort of 326 patients who underwent endovascular thrombectomy at UCLA Ronald Reagan Medical Center from 2014 to 2021. We designed a hybrid transformer model with nonlocal and cross-attention modules to predict the first-pass effect on MR imaging and CT series.

RESULTS

The proposed method achieved a mean 0.8506 (SD, 0.0712) for cross-validation receiver operating characteristic area under the curve (ROC-AUC) on MR imaging and 0.8719 (SD, 0.0831) for cross-validation ROC-AUC on CT. When evaluated on the prospective test sets, our proposed model achieved a mean ROC-AUC of 0.7967 (SD, 0.0335) with a mean sensitivity of 0.7286 (SD, 0.1849) and specificity of 0.8462 (SD, 0.1216) for MR imaging and a mean ROC-AUC of 0.8051 (SD, 0.0377) with a mean sensitivity of 0.8615 (SD, 0.1131) and specificity 0.7500 (SD, 0.1054) for CT, respectively, representing the first classification of the first-pass effect from MR imaging alone and the first automated first-pass effect classification method in CT.

CONCLUSIONS

Results illustrate that both nonperfusion MR imaging and CT from admission contain signals that can predict a successful first-pass effect following endovascular thrombectomy using our deep learning methods without requiring time-intensive manual segmentation.

Drug Burden Index Is a Modifiable Predictor of 30-Day Hospitalization in Community-Dwelling Older Adults With Complex Care Needs: Machine Learning Analysis of InterRAI Data

BACKGROUND

Older adults (≥65 years) account for a disproportionately high proportion of hospitalization and in-hospital mortality, some of which may be avoidable. Although machine learning (ML) models have already been built and validated for predicting hospitalization and mortality, there remains a significant need to optimize ML models further. Accurately predicting hospitalization may tremendously affect the clinical care of older adults as preventative measures can be implemented to improve clinical outcomes for the patient.

METHODS

In this retrospective cohort study, a data set of 14 198 community-dwelling older adults (≥65 years) with complex care needs from the International Resident Assessment Instrument-Home Care database was used to develop and optimize 3 ML models to predict 30-day hospitalization. The models developed and optimized were Random Forest (RF), XGBoost (XGB), and Logistic Regression (LR). Variable importance plots were generated for all 3 models to identify key predictors of 30-day hospitalization.

RESULTS

The area under the receiver-operating characteristics curve for the RF, XGB, and LR models were 0.97, 0.90, and 0.72, respectively. Variable importance plots identified the Drug Burden Index and alcohol consumption as important, immediately potentially modifiable variables in predicting 30-day hospitalization.

CONCLUSIONS

Identifying immediately potentially modifiable risk factors such as the Drug Burden Index and alcohol consumption is of high clinical relevance. If clinicians can influence these variables, they could proactively lower the risk of 30-day hospitalization. ML holds promise to improve the clinical care of older adults. It is crucial that these models undergo extensive validation through large-scale clinical studies before being utilized in the clinical setting.

iSPAN: Explainable prediction of outcomes post thrombectomy with Machine Learning

PURPOSE

This study aimed to develop and evaluate a machine learning model and a novel clinical score for predicting outcomes in stroke patients undergoing endovascular thrombectomy.

MATERIALS AND METHODS

This retrospective study included all patients aged over 18 years with an anterior circulation stroke treated at a thrombectomy centre from 2010 to 2020 with external validation. The primary outcome was day 90 mRS ≥3. Existing clinical scores (SPAN and PRE) and Machine Learning (ML) models were compared. A novel clinical score (iSPAN) was derived by adding an optimised weighting of the most important ML features to the SPAN.

RESULTS

812 patients were initially included (397 female, average age 73), 63 for external validation. The best performing clinical score and ML model were SPAN and XGB (sensitivity, specificity and accuracy 0.290, 0.967, 0.628 and 0.693, 0.783, 0.738 respectively). A significant difference was found overall and our XGB model was more accurate than SPAN (p < 0.0018). The most important features were Age, mTICI and total number of passes. The addition of 11 points for mTICI of ≤2B and 3 points for ≥3 passes to the SPAN achieved the best accuracy and was used to create the iSPAN. iSPAN was not significantly less accurate than our XGB model (p > 0.5). In the external validation set, iSPAN and SPAN achieved sensitivity, specificity, and accuracy of (0.735, 0.862, 0.79) and (0.471, 0.897, 0.67) respectively.

CONCLUSION

iSPAN incorporates machine-derived features to achieve better predictions compared to existing clinical scores. It is not inferior to our XGB model and is externally generalisable.

Development and Internal Validation of Machine Learning Models to Predict Mortality and Disability After Mechanical Thrombectomy for Acute Anterior Circulation Large Vessel Occlusion

OBJECTIVE

Mechanical thrombectomy (MT) improves outcomes in patients with LVO but many still experience mortality or severe disability. We sought to develop machine learning (ML) models that predict 90-day outcomes after MT for LVO.

METHODS

Consecutive patients who underwent MT for LVO between 2015-2021 at a Comprehensive Stroke Center were reviewed. Outcomes included 90-day favorable functional status (mRS 0-2), severe disability (mRS 4-6), and mortality. ML models were trained for each outcome using prethrombectomy data (pre) and with thrombectomy data (post).

RESULTS

Three hundred and fifty seven patients met the inclusion criteria. After model screening and hyperparameter tuning the top performing ML model for each outcome and timepoint was random forest (RF). Using only prethrombectomy features, the AUCs for the RFpre models were 0.73 (95% CI 0.62-0.85) for favorable functional status, 0.77 (95% CI 0.65-0.86) for severe disability, and 0.78 (95% CI 0.64-0.88) for mortality. All of these were better than a standard statistical model except for favorable functional status. Each RF model outperformed Pre, SPAN-100, THRIVE, and HIAT scores (P < 0.0001 for all). The most predictive features were premorbid mRS, age, and NIHSS. Incorporating MT data, the AUCs for the RFpost models were 0.80 (95% CI 0.67-0.90) for favorable functional status, 0.82 (95% CI 0.69-0.91) for severe disability, and 0.71 (95% CI 0.55-0.84) for mortality.

CONCLUSIONS

RF models accurately predicted 90-day outcomes after MT and performed better than standard statistical and clinical prediction models.

Application of machine learning approaches in predicting clinical outcomes in older adults - a systematic review and meta-analysis

BACKGROUND

Machine learning-based prediction models have the potential to have a considerable positive impact on geriatric care.

DESIGN

Systematic review and meta-analyses.

PARTICIPANTS

Older adults (≥ 65 years) in any setting.

INTERVENTION

Machine learning models for predicting clinical outcomes in older adults were evaluated. A random-effects meta-analysis was conducted in two grouped cohorts, where the predictive models were compared based on their performance in predicting mortality i) under and including 6 months ii) over 6 months.

OUTCOME MEASURES

Studies were grouped into two groups by the clinical outcome, and the models were compared based on the area under the receiver operating characteristic curve metric.

RESULTS

Thirty-seven studies that satisfied the systematic review criteria were appraised, and eight studies predicting a mortality outcome were included in the meta-analyses. We could only pool studies by mortality as there were inconsistent definitions and sparse data to pool studies for other clinical outcomes. The area under the receiver operating characteristic curve from the meta-analysis yielded a summary estimate of 0.80 (95% CI: 0.76 - 0.84) for mortality within 6 months and 0.81 (95% CI: 0.76 - 0.86) for mortality over 6 months, signifying good discriminatory power.

CONCLUSION

The meta-analysis indicates that machine learning models display good discriminatory power in predicting mortality. However, more large-scale validation studies are necessary. As electronic healthcare databases grow larger and more comprehensive, the available computational power increases and machine learning models become more sophisticated; there should be an effort to integrate these models into a larger research setting to predict various clinical outcomes.

Artificial Intelligence for Clinical Decision Support in Acute Ischemic Stroke: A Systematic Review

BACKGROUND

Established randomized trial-based parameters for acute ischemic stroke group patients into generic treatment groups, leading to attempts using various artificial intelligence (AI) methods to directly correlate patient characteristics to outcomes and thereby provide decision support to stroke clinicians. We review AI-based clinical decision support systems in the development stage, specifically regarding methodological robustness and constraints for clinical implementation.

METHODS

Our systematic review included full-text English language publications proposing a clinical decision support system using AI techniques for direct decision support in acute ischemic stroke cases in adult patients. We (1) describe data and outcomes used in those systems, (2) estimate the systems' benefits compared with traditional stroke diagnosis and treatment, and (3) reported concordance with reporting standards for AI in healthcare.

RESULTS

One hundred twenty-one studies met our inclusion criteria. Sixty-five were included for full extraction. In our sample, utilized data sources, methods, and reporting practices were highly heterogeneous.

CONCLUSIONS

Our results suggest significant validity threats, dissonance in reporting practices, and challenges to clinical translation. We outline practical recommendations for the successful implementation of AI research in acute ischemic stroke treatment and diagnosis.

Machine learning prediction of malignant middle cerebral artery infarction after mechanical thrombectomy for anterior circulation large vessel occlusion

OBJECTIVE

Prediction of malignant middle cerebral artery infarction (MMI) could identify patients for early intervention. We trained and internally validated a ML model that predicts MMI following mechanical thrombectomy (MT) for ACLVO.

METHODS

All patients who underwent MT for ACLVO between 2015 - 2021 at a single institution were reviewed. Data was divided into 80% training and 20% test sets. 10 models were evaluated on the training set. The top 3 models underwent hyperparameter tuning using grid search with nested 5-fold CV to optimize the area under the receiver operating curve (AUROC). Tuned models were evaluated on the test set and compared to logistic regression.

RESULTS

A total of 381 patients met the inclusion criteria. There were 50 (13.1%) patients who developed MMI. Out of the 10 ML models screened on the training set, the top 3 performing were neural network (median AUROC 0.78, IQR 0.72 - 0.83), support vector machine ([SVM] median AUROC 0.77, IQR 0.72 - 0.83), and random forest (median AUROC 0.75, IQR 0.68 - 0.81). On the test set, random forest (median AUROC 0.78, IQR 0.73 - 0.83) and neural network (median AUROC 0.78, IQR 0.73 - 0.83) were the top performing models, followed by SVM (median AUROC 0.77, IQR 0.70 - 0.83). These scores were significantly better than those for logistic regression (AUROC 0.72, IQR 0.66 - 0.78), individual risk factors, and the Malignant Brain Edema score (p < 0.001 for all).

CONCLUSION

ML models predicted MMI with good discriminative ability. They outperformed standard statistical techniques and individual risk factors.

Influential factors for the emission inspection results of urban in-use vehicles: From an ensemble learning perspective

Emission inspection of motor vehicles (emission inspection) is a crucial player in solving the problem of motor vehicle exhaust pollution, and research on the features affecting emission inspection results and their importance is a basis for optimizing the environmental management of motor vehicles. However, there is no study on the multi-feature impact analysis of the emission inspection results. This hinders the emission inspection from playing a better guiding role in the policy formulation of motor vehicle management. In this paper, the ensemble learning algorithm and interpretable machine learning theory are used. Nineteen feature indicators and over 400,000 vehicle mass analysis system (VMAS) detection data in Chengdu were selected from the emission inspection database to construct prediction models for emission inspection results. Moreover, the factors affecting emission inspection results and their ranks by importance were also obtained. The results revealed that the environment has a strong influence on the outcomes from emission inspections (accounting for about one-third of the total effect). Besides, the following eight feature indicators displayed great effects on emission inspection results in sequence: emission inspection agency (18.38%), world manufacturer code (15.01%), vehicle usage days (9.60%), transmission type (9.41%), accumulated mileage (9.21%), emission standard (5.82%), temperature (5.54%), and driving mode (5.50%). In this study, prediction models for emission inspection results are established, and the results are interpreted based on the interpretable machine learning theory. It is considered that more attention should be paid to the effect of inspection differences among emission inspection agencies on fairness, as well as the effects of differences in world manufacturer and transmission type on vehicle deterioration in future research. The supervision of emission inspection agencies, training of inspectors, elimination of obsolete vehicles, and government-guided purchase should be strengthened. This study provides empirical support for optimizing the formulation of motor vehicle environmental management policies.Implications: Emission inspection of motor vehicles (emission inspection) is a crucial player in solving the problem of motor vehicle exhaust pollution. In this work, prediction models for emission of motor vehicles inspection results are established. The results revealed that following eight feature indicators displayed great effects on emission inspection results in sequence: emission inspection agency (18.38%), world manufacturer code (15.01%), vehicle usage days (9.60%), transmission type (9.41%), accumulated mileage (9.21%), emission standard (5.82%), temperature (5.54%), and driving mode (5.50%). It is considered that more attention should be paid to the effect of inspection differences among emission inspection agencies on fairness, as well as the effects of differences in world manufacturer and transmission type on vehicle deterioration in future research. The supervision of emission inspection agencies, training of inspectors, elimination of obsolete vehicles, and government-guided purchase should be strengthened. This study provides empirical support for optimizing the formulation of motor vehicle environmental management policies.

Revascularization Outcome Prediction for A Direct Aspiration-First Pass Technique (ADAPT) from Pre-Treatment Imaging and Machine Learning

A direct aspiration-first pass technique (ADAPT) has recently gained popularity for the treatment of large vessel ischemic stroke. Here, we sought to create a machine learning-based model that uses pre-treatment imaging metrics to predict successful outcomes for ADAPT in middle cerebral artery (MCA) stroke cases. In 119 MCA strokes treated by ADAPT, we calculated four imaging parameters-clot length, perviousness, distance from the internal carotid artery (ICA) and angle of interaction (AOI) between clot/catheter. We determined treatment success by first pass effect (FPE), and performed univariate analyses. We further built and validated multivariate machine learning models in a random train-test split (75%:25%) of our data. To test model stability, we repeated the machine learning procedure over 100 randomizations, and reported the average performances. Our results show that perviousness (p = 0.002) and AOI (p = 0.031) were significantly higher and clot length (p = 0.007) was significantly lower in ADAPT cases with FPE. A logistic regression model achieved the highest accuracy (74.2%) in the testing cohort, with an AUC = 0.769. The models had similar performance over the 100 train-test randomizations (average testing AUC = 0.768 ± 0.026). This study provides feasibility of multivariate imaging-based predictors for stroke treatment outcome. Such models may help operators select the most adequate thrombectomy approach.

Decompressive Hemicraniectomy in the Modern Era of Mechanical Thrombectomy

OBJECTIVES

We aim to determine the incidence of decompressive hemicraniectomy (DHC) in the modern era of mechanical thrombectomy techniques and improved revascularization outcomes.

METHODS

We performed a retrospective analysis of 512 patients admitted with acute ischemic strokes with anterior circulation large-vessel occlusion that were treated by mechanical thrombectomy from 2010-2019. The primary endpoint was the need for surgical decompression. Secondary endpoints were infarct size, hemorrhagic conversion, and functional outcome at hospital discharge.

RESULTS

Of the 512 patients, 18 (3.5%) underwent DHC at a median 2.0 days from stroke onset. The DHC group was significantly younger than the non-DHC group (P < 0.001), had worse reperfusion rates (P = 0.024) and larger infarct size (P < 0.001). Hemorrhagic conversion was more frequent in the DHC group but did not reach statistical significance (P = 0.08). From 2010-2015, 196 patients underwent a mechanical thrombectomy, 13 of whom (6.6%) required a DHC, while 316 patients underwent mechanical thrombectomy from 2016-2019 and only 5 patients required a DHC (1.6%; P = 0.002). Younger age (P < 0.001), urinary tract infection (P < 0.001) and increasing infarct size were significantly associated with needing a DHC. When controlling for other risk factors, higher thrombolysis in cerebral infarction score significantly reduced the need for decompressive hemicraniectomy (P = 0.004).

CONCLUSIONS

This is one of the largest single-center experiences demonstrating that improved recanalization decreased the need for DHC without increasing the risk of hemorrhagic conversion.