Deep Learning–Derived High-Level Neuroimaging Features Predict Clinical Outcomes for Large Vessel Occlusion

Multi-modality multi-task model for mRS prediction using diffusion-weighted resonance imaging

This study focuses on predicting the prognosis of acute ischemic stroke patients with focal neurologic symptoms using a combination of diffusion-weighted magnetic resonance imaging (DWI) and clinical information. The primary outcome is a poor functional outcome defined by a modified Rankin Scale (mRS) score of 3-6 after 3 months of stroke. Employing nnUnet for DWI lesion segmentation, the study utilizes both multi-task and multi-modality methodologies, integrating DWI and clinical data for prognosis prediction. Integrating the two modalities was shown to improve performance by 0.04 compared to using DWI only. The model achieves notable performance metrics, with a dice score of 0.7375 for lesion segmentation and an area under the curve of 0.8080 for mRS prediction. These results surpass existing scoring systems, showing a 0.16 improvement over the Totaled Health Risks in Vascular Events score. The study further employs grad-class activation maps to identify critical brain regions influencing mRS scores. Analysis of the feature map reveals the efficacy of the multi-tasking nnUnet in predicting poor outcomes, providing insights into the interplay between DWI and clinical data. In conclusion, the integrated approach demonstrates significant advancements in prognosis prediction for cerebral infarction patients, offering a superior alternative to current scoring systems.

Segmentation and Volume Estimation of the Habenula Using Deep Learning in Patients With Depression

Background

The habenula is involved in the pathophysiology of depression. However, its small structure limits the accuracy of segmentation methods, and the findings regarding its volume have been inconsistent. This study aimed to create a highly accurate habenula segmentation model using deep learning, test its generalizability to clinical magnetic resonance imaging, and examine differences between healthy participants and patients with depression.

Methods

This multicenter study included 382 participants (patients with depression: N = 234, women 47.0%; healthy participants: N = 148, women 37.8%). A 3-dimensional residual U-Net was used to create a habenula segmentation model on 3T magnetic resonance images. The reproducibility and generalizability of the predictive model were tested on various validation cohorts. Thereafter, differences between the habenula volume of healthy participants and that of patients with depression were examined.

Results

A Dice coefficient of 86.6% was achieved in the derivation cohort. The test-retest dataset showed a mean absolute percentage error of 6.66, indicating sufficiently high reproducibility. A Dice coefficient of >80% was achieved for datasets with different imaging conditions, such as magnetic field strengths, spatial resolutions, and imaging sequences, by adjusting the threshold. A significant negative correlation with age was observed in the general population, and this correlation was more pronounced in patients with depression (p < 10-7, r = -0.59). Habenula volume decreased with depression severity in women even when the effects of age and scanner were excluded (p = .019, η2 = 0.099).

Conclusions

Habenula volume could be a pathophysiologically relevant factor and diagnostic and therapeutic marker for depression, particularly in women.

Distinct brain morphometry patterns revealed by deep learning improve prediction of post-stroke aphasia severity

BACKGROUND

Emerging evidence suggests that post-stroke aphasia severity depends on the integrity of the brain beyond the lesion. While measures of lesion anatomy and brain integrity combine synergistically to explain aphasic symptoms, substantial interindividual variability remains unaccounted. One explanatory factor may be the spatial distribution of morphometry beyond the lesion (e.g., atrophy), including not just specific brain areas, but distinct three-dimensional patterns.

METHODS

Here, we test whether deep learning with Convolutional Neural Networks (CNNs) on whole brain morphometry (i.e., segmented tissue volumes) and lesion anatomy better predicts chronic stroke individuals with severe aphasia (N = 231) than classical machine learning (Support Vector Machines; SVMs), evaluating whether encoding spatial dependencies identifies uniquely predictive patterns.

RESULTS

CNNs achieve higher balanced accuracy and F1 scores, even when SVMs are nonlinear or integrate linear or nonlinear dimensionality reduction. Parity only occurs when SVMs access features learned by CNNs. Saliency maps demonstrate that CNNs leverage distributed morphometry patterns, whereas SVMs focus on the area around the lesion. Ensemble clustering of CNN saliencies reveals distinct morphometry patterns unrelated to lesion size, consistent across individuals, and which implicate unique networks associated with different cognitive processes as measured by the wider neuroimaging literature. Individualized predictions depend on both ipsilateral and contralateral features outside the lesion.

CONCLUSIONS

Three-dimensional network distributions of morphometry are directly associated with aphasia severity, underscoring the potential for CNNs to improve outcome prognostication from neuroimaging data, and highlighting the prospective benefits of interrogating spatial dependence at different scales in multivariate feature space.

Magnetic resonance imaging-based deep learning imaging biomarker for predicting functional outcomes after acute ischemic stroke

PURPOSE

Clinical risk scores are essential for predicting outcomes in stroke patients. The advancements in deep learning (DL) techniques provide opportunities to develop prediction applications using magnetic resonance (MR) images. We aimed to develop an MR-based DL imaging biomarker for predicting outcomes in acute ischemic stroke (AIS) and evaluate its additional benefit to current risk scores.

METHOD

This study included 3338 AIS patients. We trained a DL model using deep neural network architectures on MR images and radiomics to predict poor functional outcomes at three months post-stroke. The DL model generated a DL score, which served as the DL imaging biomarker. We compared the predictive performance of this biomarker to five risk scores on a holdout test set. Additionally, we assessed whether incorporating the imaging biomarker into the risk scores improved the predictive performance.

RESULTS

The DL imaging biomarker achieved an area under the receiver operating characteristic curve (AUC) of 0.788. The AUCs of the five studied risk scores were 0.789, 0.793, 0.804, 0.810, and 0.826, respectively. The imaging biomarker's predictive performance was comparable to four of the risk scores but inferior to one (p = 0.038). Adding the imaging biomarker to the risk scores improved the AUCs (p-values) to 0.831 (0.003), 0.825 (0.001), 0.834 (0.003), 0.836 (0.003), and 0.839 (0.177), respectively. The net reclassification improvement and integrated discrimination improvement indices also showed significant improvements (all p < 0.001).

CONCLUSIONS

Using DL techniques to create an MR-based imaging biomarker is feasible and enhances the predictive ability of current risk scores.

Deep learning on pre-procedural computed tomography and clinical data predicts outcome following stroke thrombectomy

BACKGROUND

Deep learning using clinical and imaging data may improve pre-treatment prognostication in ischemic stroke patients undergoing endovascular thrombectomy (EVT).

METHODS

Deep learning models were trained and tested on baseline clinical and imaging (CT head and CT angiography) data to predict 3-month functional outcomes in stroke patients who underwent EVT. Classical machine learning models (logistic regression and random forest classifiers) were constructed to compare their performance with the deep learning models. An external validation dataset was used to validate the models. The MR PREDICTS prognostic tool was tested on the external validation set, and its performance was compared with the deep learning and classical machine learning models.

RESULTS

A total of 975 patients (550 men; mean±SD age 67.5±15.1 years) were studied with 778 patients in the model development cohort and 197 in the external validation cohort. The deep learning model trained on baseline CT and clinical data, and the logistic regression model (clinical data alone) demonstrated the strongest discriminative abilities for 3-month functional outcome and were comparable (AUC 0.811 vs 0.817, Q=0.82). Both models exhibited superior prognostic performance than the other deep learning (CT head alone, CT head, and CT angiography) and MR PREDICTS models (all Q<0.05).

CONCLUSIONS

The discriminative performance of deep learning for predicting functional independence was comparable to logistic regression. Future studies should focus on whether incorporating procedural and post-procedural data significantly improves model performance.

Scaling behaviours of deep learning and linear algorithms for the prediction of stroke severity

Deep learning has allowed for remarkable progress in many medical scenarios. Deep learning prediction models often require 105-107 examples. It is currently unknown whether deep learning can also enhance predictions of symptoms post-stroke in real-world samples of stroke patients that are often several magnitudes smaller. Such stroke outcome predictions however could be particularly instrumental in guiding acute clinical and rehabilitation care decisions. We here compared the capacities of classically used linear and novel deep learning algorithms in their prediction of stroke severity. Our analyses relied on a total of 1430 patients assembled from the MRI-Genetics Interface Exploration collaboration and a Massachusetts General Hospital-based study. The outcome of interest was National Institutes of Health Stroke Scale-based stroke severity in the acute phase after ischaemic stroke onset, which we predict by means of MRI-derived lesion location. We automatically derived lesion segmentations from diffusion-weighted clinical MRI scans, performed spatial normalization and included a principal component analysis step, retaining 95% of the variance of the original data. We then repeatedly separated a train, validation and test set to investigate the effects of sample size; we subsampled the train set to 100, 300 and 900 and trained the algorithms to predict the stroke severity score for each sample size with regularized linear regression and an eight-layered neural network. We selected hyperparameters on the validation set. We evaluated model performance based on the explained variance (R2) in the test set. While linear regression performed significantly better for a sample size of 100 patients, deep learning started to significantly outperform linear regression when trained on 900 patients. Average prediction performance improved by ∼20% when increasing the sample size 9× [maximum for 100 patients: 0.279 ± 0.005 (R2, 95% confidence interval), 900 patients: 0.337 ± 0.006]. In summary, for sample sizes of 900 patients, deep learning showed a higher prediction performance than typically employed linear methods. These findings suggest the existence of non-linear relationships between lesion location and stroke severity that can be utilized for an improved prediction performance for larger sample sizes.

Predicting stroke outcome: A case for multimodal deep learning methods with tabular and CT Perfusion data

MOTIVATION

Acute ischemic stroke is one of the leading causes of morbidity and disability worldwide, often followed by a long rehabilitation period. To improve and personalize stroke rehabilitation, it is essential to provide a reliable prognosis to caregivers and patients. Deep learning techniques might improve the predictions by incorporating different data modalities. We present a multimodal approach to predict the functional status of acute ischemic stroke patients after their discharge based on tabular data and CT perfusion imaging.

METHODS

We conducted experiments on tabular, imaging, and multimodal deep learning architectures to predict dichotomized mRS scores 3 months after the event. The dataset was collected from a Dutch hospital and includes 98 CVA patients with a visible occlusion on their CT perfusion scan. Tabular data is based on the Dutch Acute Stroke Audit data, and imaging data consists of summed-up CT perfusion maps.

RESULTS

On the tabular data, TabNet outperformed our baselines with an AUC of 0.71, while ResNet-10 on the imaging data performed comparably with an AUC of 0.70. Our implementation of the multimodal DAFT architecture outperforms baselines as well as comparable studies by achieving an 0.75 AUC, and 0.80 F1 score. This was achieved with a final model of less than a hundred thousand optimizable parameters, and a dataset less than half the size of reference papers.

CONCLUSION

Overall, we demonstrate the feasibility of predicting the functional outcome for ischemic stroke patients and the usability of multimodal deep learning architectures for this purpose.

A weakly supervised deep learning model integrating noncontrasted computed tomography images and clinical factors facilitates haemorrhagic transformation prediction after intravenous thrombolysis in acute ischaemic stroke patients

BACKGROUND

Haemorrhage transformation (HT) is a serious complication of intravenous thrombolysis (IVT) in acute ischaemic stroke (AIS). Accurate and timely prediction of the risk of HT before IVT may change the treatment decision and improve clinical prognosis. We aimed to develop a deep learning method for predicting HT after IVT for AIS using noncontrast computed tomography (NCCT) images.

METHODS

We retrospectively collected data from 828 AIS patients undergoing recombinant tissue plasminogen activator (rt-PA) treatment within a 4.5-h time window (n = 665) or of undergoing urokinase treatment within a 6-h time window (n = 163) and divided them into the HT group (n = 69) and non-HT group (n = 759). HT was defined based on the criteria of the European Cooperative Acute Stroke Study-II trial. To address the problems of indiscernible features and imbalanced data, a weakly supervised deep learning (WSDL) model for HT prediction was constructed based on multiple instance learning and active learning using admission NCCT images and clinical information in addition to conventional deep learning models. Threefold cross-validation and transfer learning were performed to confirm the robustness of the network. Of note, the predictive value of the commonly used scales in clinics associated with NCCT images (i.e., the HAT and SEDAN score) was also analysed and compared to measure the feasibility of our proposed DL algorithms.

RESULTS

Compared to the conventional DL and ML models, the WSDL model had the highest AUC of 0.799 (95% CI 0.712-0.883). Significant differences were observed between the WSDL model and five ML models (P < 0.05). The prediction performance of the WSDL model outperforms the HAT and SEDAN scores at the optimal operating point (threshold = 1.5). Further subgroup analysis showed that the WSDL model performed better for symptomatic intracranial haemorrhage (AUC = 0.833, F1 score = 0.909).

CONCLUSIONS

Our WSDL model based on NCCT images had relatively good performance for predicting HT in AIS and may be suitable for assisting in clinical treatment decision-making.

Preserved Corticospinal Tract Revealed by Acute Perfusion Imaging Relates to Better Outcome After Thrombectomy in Stroke

BACKGROUND

The indication for mechanical thrombectomy (MT) in stroke patients with large vessel occlusion has been constantly expanded over the past years. Despite remarkable treatment effects at the group level in clinical trials, many patients remain severely disabled even after successful recanalization. A better understanding of this outcome variability will help to improve clinical decision-making on MT in the acute stage. Here, we test whether current outcome models can be refined by integrating information on the preservation of the corticospinal tract as a functionally crucial white matter tract derived from acute perfusion imaging.

METHODS

We retrospectively analyzed 162 patients with stroke and large vessel occlusion of the anterior circulation who were admitted to the University Medical Center Lübeck between 2014 and 2020 and underwent MT. The ischemic core was defined as fully automatized based on the acute computed tomography perfusion with cerebral blood volume data using outlier detection and clustering algorithms. Normative whole-brain structural connectivity data were used to infer whether the corticospinal tract was affected by the ischemic core or preserved. Ordinal logistic regression models were used to correlate this information with the modified Rankin Scale after 90 days.

RESULTS

The preservation of the corticospinal tract was associated with a reduced risk of a worse functional outcome in large vessel occlusion-stroke patients undergoing MT, with an odds ratio of 0.28 (95% CI, 0.15-0.53). This association was still significant after adjusting for multiple confounding covariables, such as age, lesion load, initial symptom severity, sex, stroke side, and recanalization status.

CONCLUSIONS

A preinterventional computed tomography perfusion-based surrogate of corticospinal tract preservation or disconnectivity is strongly associated with functional outcomes after MT. If validated in independent samples this concept could serve as a novel tool to improve current outcome models to better understand intersubject variability after MT in large vessel occlusion stroke.

Artificial Intelligence-Based Consumer Health Informatics Application: Scoping Review

BACKGROUND

There is no doubt that the recent surge in artificial intelligence (AI) research will change the trajectory of next-generation health care, making it more approachable and accessible to patients. Therefore, it is critical to research patient perceptions and outcomes because this trend will allow patients to be the primary consumers of health technology and decision makers for their own health.

OBJECTIVE

This study aimed to review and analyze papers on AI-based consumer health informatics (CHI) for successful future patient-centered care.

METHODS

We searched for all peer-reviewed papers in PubMed published in English before July 2022. Research on an AI-based CHI tool or system that reports patient outcomes or perceptions was identified for the scoping review.

RESULTS

We identified 20 papers that met our inclusion criteria. The eligible studies were summarized and discussed with respect to the role of the AI-based CHI system, patient outcomes, and patient perceptions. The AI-based CHI systems identified included systems in mobile health (13/20, 65%), robotics (5/20, 25%), and telemedicine (2/20, 10%). All the systems aimed to provide patients with personalized health care. Patient outcomes and perceptions across various clinical disciplines were discussed, demonstrating the potential of an AI-based CHI system to benefit patients.

CONCLUSIONS

This scoping review showed the trend in AI-based CHI systems and their impact on patient outcomes as well as patients' perceptions of these systems. Future studies should also explore how clinicians and health care professionals perceive these consumer-based systems and integrate them into the overall workflow.

Machine learning prediction of motor function in chronic stroke patients: a systematic review and meta-analysis

Background

Recent studies have reported that machine learning (ML), with a relatively strong capacity for processing non-linear data and adaptive ability, could improve the accuracy and efficiency of prediction. The article summarizes the published studies on ML models that predict motor function 3-6 months post-stroke.

Methods

A systematic literature search was conducted in PubMed, Embase, Cochorane and Web of Science as of April 3, 2023 for studies on ML prediction of motor function in stroke patients. The quality of the literature was assessed using the Prediction model Risk Of Bias Assessment Tool (PROBAST). A random-effects model was preferred for meta-analysis using R4.2.0 because of the different variables and parameters.

Results

A total of 44 studies were included in this meta-analysis, involving 72,368 patients and 136 models. Models were categorized into subgroups according to the predicted outcome Modified Rankin Scale cut-off value and whether they were constructed based on radiomics. C-statistics, sensitivity, and specificity were calculated. The random-effects model showed that the C-statistics of all models were 0.81 (95% CI: 0.79; 0.83) in the training set and 0.82 (95% CI: 0.80; 0.85) in the validation set. According to different Modified Rankin Scale cut-off values, C-statistics of ML models predicting Modified Rankin Scale>2(used most widely) in stroke patients were 0.81 (95% CI: 0.78; 0.84) in the training set, and 0.84 (95% CI: 0.81; 0.87) in the validation set. C-statistics of radiomics-based ML models in the training set and validation set were 0.81 (95% CI: 0.78; 0.84) and 0.87 (95% CI: 0.83; 0.90), respectively.

Conclusion

ML can be used as an assessment tool for predicting the motor function in patients with 3-6 months of post-stroke. Additionally, the study found that ML models with radiomics as a predictive variable were also demonstrated to have good predictive capabilities. This systematic review provides valuable guidance for the future optimization of ML prediction systems that predict poor motor outcomes in stroke patients.

Systematic review registration

https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42022335260, identifier: CRD42022335260.

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.

Artificial Intelligence in Acute Ischemic Stroke Subtypes According to Toast Classification: A Comprehensive Narrative Review

The correct recognition of the etiology of ischemic stroke (IS) allows tempestive interventions in therapy with the aim of treating the cause and preventing a new cerebral ischemic event. Nevertheless, the identification of the cause is often challenging and is based on clinical features and data obtained by imaging techniques and other diagnostic exams. TOAST classification system describes the different etiologies of ischemic stroke and includes five subtypes: LAAS (large-artery atherosclerosis), CEI (cardio embolism), SVD (small vessel disease), ODE (stroke of other determined etiology), and UDE (stroke of undetermined etiology). AI models, providing computational methodologies for quantitative and objective evaluations, seem to increase the sensitivity of main IS causes, such as tomographic diagnosis of carotid stenosis, electrocardiographic recognition of atrial fibrillation, and identification of small vessel disease in magnetic resonance images. The aim of this review is to provide overall knowledge about the most effective AI models used in the differential diagnosis of ischemic stroke etiology according to the TOAST classification. According to our results, AI has proven to be a useful tool for identifying predictive factors capable of subtyping acute stroke patients in large heterogeneous populations and, in particular, clarifying the etiology of UDE IS especially detecting cardioembolic sources.

Development and Validation of a Machine Learning Predictive Model for Cardiac Surgery-Associated Acute Kidney Injury

OBJECTIVE

We aimed to develop and validate a predictive machine learning (ML) model for cardiac surgery associated with acute kidney injury (CSA-AKI) based on a multicenter randomized control trial (RCT) and a Medical Information Mart for Intensive Care-IV (MIMIC-IV) dataset.

METHODS

This was a subanalysis from a completed RCT approved by the Ethics Committee of Fuwai Hospital in Beijing, China (NCT03782350). Data from Fuwai Hospital were randomly assigned, with 80% for the training dataset and 20% for the testing dataset. The data from three other centers were used for the external validation dataset. Furthermore, the MIMIC-IV dataset was also utilized to validate the performance of the predictive model. The area under the receiver operating characteristic curve (ROC-AUC), the precision-recall curve (PR-AUC), and the calibration brier score were applied to evaluate the performance of the traditional logistic regression (LR) and eleven ML algorithms. Additionally, the Shapley Additive Explanations (SHAP) interpreter was used to explain the potential risk factors for CSA-AKI.

RESULT

A total of 6495 eligible patients undergoing cardiopulmonary bypass (CPB) were eventually included in this study, 2416 of whom were from Fuwai Hospital (Beijing), for model development, 562 from three other cardiac centers in China, and 3517 from the MIMICIV dataset, were used, respectively, for external validation. The CatBoostClassifier algorithms outperformed other models, with excellent discrimination and calibration performance for the development, as well as the MIMIC-IV, datasets. In addition, the CatBoostClassifier achieved ROC-AUCs of 0.85, 0.67, and 0.77 and brier scores of 0.14, 0.19, and 0.16 in the testing, external, and MIMIC-IV datasets, respectively. Moreover, the utmost important risk factor, the N-terminal brain sodium peptide (NT-proBNP), was confirmed by the LASSO method in the feature section process. Notably, the SHAP explainer identified that the preoperative blood urea nitrogen level, prothrombin time, serum creatinine level, total bilirubin level, and age were positively correlated with CSA-AKI; preoperative platelets level, systolic and diastolic blood pressure, albumin level, and body weight were negatively associated with CSA-AKI.

CONCLUSIONS

The CatBoostClassifier algorithms outperformed other ML models in the discrimination and calibration of CSA-AKI prediction cardiac surgery with CPB, based on a multicenter RCT and MIMIC-IV dataset. Moreover, the preoperative NT-proBNP level was confirmed to be strongly related to CSA-AKI.

Interpretable deep learning for the prognosis of long-term functional outcome post-stroke using acute diffusion weighted imaging

Advances in deep learning can be applied to acute stroke imaging to build powerful and explainable prediction models that could supersede traditionally used biomarkers. We aimed to evaluate the performance and interpretability of a deep learning model based on convolutional neural networks (CNN) in predicting long-term functional outcome with diffusion-weighted imaging (DWI) acquired at day 1 post-stroke. Ischemic stroke patients (n = 322) were included from the ASTER and INSULINFARCT trials as well as the Pitié-Salpêtrière registry. We trained a CNN to predict long-term functional outcome assessed at 3 months with the modified Rankin Scale (dichotomized as good [mRS ≤ 2] vs. poor [mRS ≥ 3]) and compared its performance to two logistic regression models using lesion volume and ASPECTS. The CNN contained an attention mechanism, which allowed to visualize the areas of the brain that drove prediction. The deep learning model yielded a significantly higher area under the curve (0.83 95%CI [0.78-0.87]) than lesion volume (0.78 [0.73-0.83]) and ASPECTS (0.77 [0.71-0.83]) (p < 0.05). Setting all classifiers to the specificity as the deep learning model (i.e., 0.87 [0.82-0.92]), the CNN yielded a significantly higher sensitivity (0.67 [0.59-0.73]) than lesion volume (0.48 [0.40-0.56]) and ASPECTS (0.50 [0.41-0.58]) (p = 0.002). The attention mechanism revealed that the network learned to naturally attend to the lesion to predict outcome.

Deep learning-based personalised outcome prediction after acute ischaemic stroke

BACKGROUND

Whether deep learning models using clinical data and brain imaging can predict the long-term risk of major adverse cerebro/cardiovascular events (MACE) after acute ischaemic stroke (AIS) at the individual level has not yet been studied.

METHODS

A total of 8590 patients with AIS admitted within 5 days of symptom onset were enrolled. The primary outcome was the occurrence of MACEs (a composite of stroke, acute myocardial infarction or death) over 12 months. The performance of deep learning models (DeepSurv and Deep-Survival-Machines (DeepSM)) and traditional survival models (Cox proportional hazards (CoxPH) and random survival forest (RSF)) were compared using the time-dependent concordance index ([Formula: see text] index).

RESULTS

Given the top 1 to all 60 clinical factors according to feature importance, CoxPH and RSF yielded [Formula: see text] index of 0.7236-0.8222 and 0.7279-0.8335, respectively. Adding image features improved the performance of deep learning models and traditional models assisted by deep learning models. DeepSurv and DeepSM yielded the best [Formula: see text] index of 0.8496 and 0.8531 when images were added to all 39 relevant clinical factors, respectively. In feature importance, brain image was consistently ranked highly. Deep learning models automatically extracted the image features directly from personalised brain images and predicted the risk and date of future MACEs at the individual level.

CONCLUSIONS

Deep learning models using clinical data and brain images could improve the prediction of MACEs and provide personalised outcome prediction for patients with AIS. Deep learning models will allow us to develop more accurate and tailored prognostic prediction systems that outperform traditional models.

Prognostic Value of Combined Radiomic Features from Follow-Up DWI and T2-FLAIR in Acute Ischemic Stroke

The biological pathways involved in lesion formation after an acute ischemic stroke (AIS) are poorly understood. Despite successful reperfusion treatment, up to two thirds of patients with large vessel occlusion remain functionally dependent. Imaging characteristics extracted from DWI and T2-FLAIR follow-up MR sequences could aid in providing a better understanding of the lesion constituents. We built a fully automated pipeline based on a tree ensemble machine learning model to predict poor long-term functional outcome in patients from the MR CLEAN-NO IV trial. Several feature sets were compared, considering only imaging, only clinical, or both types of features. Nested cross-validation with grid search and a feature selection procedure based on SHapley Additive exPlanations (SHAP) was used to train and validate the models. Considering features from both imaging modalities in combination with clinical characteristics led to the best prognostic model (AUC = 0.85, 95%CI [0.81, 0.89]). Moreover, SHAP values showed that imaging features from both sequences have a relevant impact on the final classification, with texture heterogeneity being the most predictive imaging biomarker. This study suggests the prognostic value of both DWI and T2-FLAIR follow-up sequences for AIS patients. If combined with clinical characteristics, they could lead to better understanding of lesion pathophysiology and improved long-term functional outcome prediction.

Deep Learning in Ischemic Stroke Imaging Analysis: A Comprehensive Review

Ischemic stroke is a cerebrovascular disease with a high morbidity and mortality rate, which poses a serious challenge to human health and life. Meanwhile, the management of ischemic stroke remains highly dependent on manual visual analysis of noncontrast computed tomography (CT) or magnetic resonance imaging (MRI). However, artifacts and noise of the equipment as well as the radiologist experience play a significant role on diagnostic accuracy. To overcome these defects, the number of computer-aided diagnostic (CAD) methods for ischemic stroke is increasing substantially during the past decade. Particularly, deep learning models with massive data learning capabilities are recognized as powerful auxiliary tools for the acute intervention and guiding prognosis of ischemic stroke. To select appropriate interventions, facilitate clinical practice, and improve the clinical outcomes of patients, this review firstly surveys the current state-of-the-art deep learning technology. Then, we summarized the major applications in acute ischemic stroke imaging, particularly in exploring the potential function of stroke diagnosis and multimodal prognostication. Finally, we sketched out the current problems and prospects.

Machine learning based outcome prediction of large vessel occlusion of the anterior circulation prior to thrombectomy in patients with wake-up stroke

PURPOSE

Outcome prediction of large vessel occlusion of the anterior circulation in patients with wake-up stroke is important to identify patients that will benefit from thrombectomy. Currently, mismatch concepts that require MRI or CT-Perfusion (CTP) are recommended to identify these patients. We evaluated machine learning algorithms in their ability to discriminate between patients with favorable (defined as a modified Rankin Scale (mRS) score of 0-2) and unfavorable (mRS 3-6) outcome and between patients with poor (mRS5-6) and non-poor (mRS 0-4) outcome.

METHODS

Data of 8395 patients that were treated between 2018 and 2020 from the nationwide registry of the German Society for Neuroradiology was retrospectively analyzed. Five models were trained with clinical variables and Alberta Stroke Program Early CT Score (ASPECTS). The model with the highest accuracy was validated with a test dataset with known stroke onset and with a test dataset that consisted only of wake-up strokes.

RESULTS

2419 patients showed poor and 3310 patients showed favorable outcome. The best performing Random Forest model achieved a sensitivity of 0.65, a specificity of 0.81 and an AUC of 0.79 on the test dataset of patients with wake-up stroke in the classification analysis between favorable and unfavorable outcome and a sensitivity of 0.42, a specificity of 0.83 and an AUC of 0.72 in the classification analysis between poor and non-poor outcome.

CONCLUSION

Machine learning algorithms have the potential to aid in the decision making for thrombectomy in patients with wake-up stroke especially in hospitals, where emergency CTP or MRI imaging is not available.

Machine Learning Risk Prediction for Incident Heart Failure in Patients With Atrial Fibrillation

BACKGROUND

Atrial fibrillation (AF) increases the risk of heart failure (HF); however, little focus is placed on the risk stratification for, and prevention of, incident HF in patients with AF.

OBJECTIVES

This study aimed to construct and validate a machine learning (ML) prediction model for HF hospitalization in patients with AF.

METHODS

The Fushimi AF Registry is a community-based prospective survey of patients with AF in Fushimi-ku, Kyoto, Japan. We divided the data set of the registry into derivation (n = 2,383) and validation (n = 2,011) cohorts. An ML model was built to predict the incidence of HF hospitalization using the derivation cohort, and predictive ability was examined using the validation cohort.

RESULTS

HF hospitalization occurred in 606 patients (14%) during a median follow-up period of 4.4 years in the entire registry. Data of transthoracic echocardiography and biomarkers were frequently nominated as important predictive variables across all 6 ML models. The ML model based on a random forest algorithm using 7 variables (age, history of HF, creatinine clearance, cardiothoracic ratio on x-ray, left ventricular [LV] ejection fraction, LV end-systolic diameter, and LV asynergy) had high prediction performance (area under the receiver operating characteristics curve [AUC]: 0.75) and was significantly superior to the Framingham HF risk model (AUC: 0.67; P < 0.001). Based on Kaplan-Meier curves, the ML model could stratify the risk of HF hospitalization during the follow-up period (log-rank; P < 0.001).

CONCLUSIONS

The ML model revealed important predictors and helped us to stratify the risk of HF, providing opportunities for the prevention of HF in patients with AF.

Pre-thrombectomy prognostic prediction of large-vessel ischemic stroke using machine learning: A systematic review and meta-analysis

Introduction

Machine learning (ML) methods are being increasingly applied to prognostic prediction for stroke patients with large vessel occlusion (LVO) treated with endovascular thrombectomy. This systematic review aims to summarize ML-based pre-thrombectomy prognostic models for LVO stroke and identify key research gaps.

Methods

Literature searches were performed in Embase, PubMed, Web of Science, and Scopus. Meta-analyses of the area under the receiver operating characteristic curves (AUCs) of ML models were conducted to synthesize model performance.

Results

Sixteen studies describing 19 models were eligible. The predicted outcomes include functional outcome at 90 days, successful reperfusion, and hemorrhagic transformation. Functional outcome was analyzed by 10 conventional ML models (pooled AUC=0.81, 95% confidence interval [CI]: 0.77-0.85, AUC range: 0.68-0.93) and four deep learning (DL) models (pooled AUC=0.75, 95% CI: 0.70-0.81, AUC range: 0.71-0.81). Successful reperfusion was analyzed by three conventional ML models (pooled AUC=0.72, 95% CI: 0.56-0.88, AUC range: 0.55-0.88) and one DL model (AUC=0.65, 95% CI: 0.62-0.68).

Conclusions

Conventional ML and DL models have shown variable performance in predicting post-treatment outcomes of LVO without generally demonstrating superiority compared to existing prognostic scores. Most models were developed using small datasets, lacked solid external validation, and at high risk of potential bias. There is considerable scope to improve study design and model performance. The application of ML and DL methods to improve the prediction of prognosis in LVO stroke, while promising, remains nascent.

Systematic review registration

https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021266524, identifier CRD42021266524.

Deep Learning to Predict Traumatic Brain Injury Outcomes in the Low-Resource Setting

OBJECTIVE

Traumatic brain injury (TBI) disproportionately affects low- and middle-income countries (LMICs). In these settings, accurate patient prognostication is both difficult and essential for high-quality patient care. With the ultimate goal of enhancing TBI triage in LMICs, we aim to develop the first deep learning model to predict outcomes after TBI and compare its performance with that of less complex algorithms.

METHODS

TBI patients' data were prospectively collected in Kampala, Uganda, from 2016 to 2020. To predict good versus poor outcome at hospital discharge, we created deep neural network, shallow neural network, and elastic-net regularized logistic regression models. Predictors included 13 easily acquirable clinical variables. We assessed model performance with 5-fold cross-validation to calculate areas under both the receiver operating characteristic curve and precision-recall curve (AUPRC), in addition to standardized partial AUPRC to focus on comparisons at clinically relevant operating points.

RESULTS

We included 2164 patients for model training, of which 12% had poor outcomes. The deep neural network performed best as measured by the area under the receiver operating characteristic curve (0.941) and standardized partial AUPRC in region maximizing recall (0.291), whereas the shallow neural network was best by the area under the precision-recall curve (0.770). In several other comparisons, the elastic-net regularized logistic regression was noninferior to the neural networks.

CONCLUSIONS

We present the first use of deep learning for TBI prognostication, with an emphasis on LMICs, where there is great need for decision support to allocate limited resources. Optimal algorithm selection depends on the specific clinical setting; deep learning is not a panacea, though it may have a role in these efforts.

FeMA: Feature matching auto-encoder for predicting ischaemic stroke evolution and treatment outcome

Although, predicting ischaemic stroke evolution and treatment outcome provide important information one step towards individual treatment planning, classifying functional outcome and modelling the brain tissue evolution remains a challenge due to data complexity and visually subtle changes in the brain. We propose a novel deep learning approach, Feature Matching Auto-encoder (FeMA) that consists of two stages, predicting ischaemic stroke evolution at one week without voxel-wise annotation and predicting ischaemic stroke treatment outcome at 90 days from a baseline scan. In the first stage, we introduce feature similarity and consistency objective, and in the second stage, we show that adding stroke evolution information increase the performance of functional outcome prediction. Comparative experiments demonstrate that our proposed method is more effective to extract representative follow-up features and achieves the best results for functional outcome of stroke treatment.

Precision medicine in stroke: towards personalized outcome predictions using artificial intelligence

Stroke ranks among the leading causes for morbidity and mortality worldwide. New and continuously improving treatment options such as thrombolysis and thrombectomy have revolutionized acute stroke treatment in recent years. Following modern rhythms, the next revolution might well be the strategic use of the steadily increasing amounts of patient-related data for generating models enabling individualized outcome predictions. Milestones have already been achieved in several health care domains, as big data and artificial intelligence have entered everyday life. The aim of this review is to synoptically illustrate and discuss how artificial intelligence approaches may help to compute single-patient predictions in stroke outcome research in the acute, subacute and chronic stage. We will present approaches considering demographic, clinical and electrophysiological data, as well as data originating from various imaging modalities and combinations thereof. We will outline their advantages, disadvantages, their potential pitfalls and the promises they hold with a special focus on a clinical audience. Throughout the review we will highlight methodological aspects of novel machine-learning approaches as they are particularly crucial to realize precision medicine. We will finally provide an outlook on how artificial intelligence approaches might contribute to enhancing favourable outcomes after stroke.

Machine Learning-Based Model for Predicting Incidence and Severity of Acute Ischemic Stroke in Anterior Circulation Large Vessel Occlusion

Objectives: Patients with anterior circulation large vessel occlusion are at high risk of acute ischemic stroke, which could be disabling or fatal. In this study, we applied machine learning to develop and validate two prediction models for acute ischemic stroke (Model 1) and severity of neurological impairment (Model 2), both caused by anterior circulation large vessel occlusion (AC-LVO), based on medical history and neuroimaging data of patients on admission.

Methods: A total of 1,100 patients with AC- LVO from the Second Hospital of Hebei Medical University in North China were enrolled, of which 713 patients presented with acute ischemic stroke (AIS) related to AC- LVO and 387 presented with the non-acute ischemic cerebrovascular event. Among patients with the non-acute ischemic cerebrovascular events, 173 with prior stroke or TIA were excluded. Finally, 927 patients with AC-LVO were entered into the derivation cohort. In the external validation cohort, 150 patients with AC-LVO from the Hebei Province People's Hospital, including 99 patients with AIS related to AC- LVO and 51 asymptomatic AC-LVO patients, were retrospectively reviewed. We developed four machine learning models [logistic regression (LR), regularized LR (RLR), support vector machine (SVM), and random forest (RF)], whose performance was internally validated using 5-fold cross-validation. The performance of each machine learning model for the area under the receiver operating characteristic curve (ROC-AUC) was compared and the variables of each algorithm were ranked.

Results: In model 1, among the included patients with AC-LVO, 713 (76.9%) and 99 (66%) suffered an acute ischemic stroke in the derivation and external validation cohorts, respectively. The ROC-AUC of LR, RLR and SVM were significantly higher than that of the RF in the external validation cohorts [0.66 (95% CI 0.57–0.74) for LR, 0.66 (95% CI 0.57–0.74) for RLR, 0.55 (95% CI 0.45–0.64) for RF and 0.67 (95% CI 0.58–0.76) for SVM]. In model 2, 254 (53.9%) and 31 (37.8%) patients suffered disabling ischemic stroke in the derivation and external validation cohorts, respectively. There was no difference in AUC among the four machine learning algorithms in the external validation cohorts.

Conclusions: Machine learning methods with multiple clinical variables have the ability to predict acute ischemic stroke and the severity of neurological impairment in patients with AC-LVO.

Predicting functional outcome in patients with acute brainstem infarction using deep neuroimaging features

BACKGROUND AND PURPOSE

Acute brainstem infarctions can lead to serious functional impairments. We aimed to predict functional outcomes in patients with acute brainstem infarction using deep neuroimaging features extracted by convolutional neural networks (CNNs).

METHODS

This nationwide multicenter stroke registry study included 1482 patients with acute brainstem infarction. We applied CNNs to automatically extract deep neuroimaging features from diffusion-weighted imaging. Deep learning models based on clinical features, laboratory features, conventional imaging features (infarct volume, number of infarctions), and deep neuroimaging features were trained to predict functional outcomes at 3 months poststroke. Unfavorable outcome was defined as modified Rankin Scale score of 3 or higher at 3 months. The models were evaluated by comparing the area under the receiver operating characteristic curve (AUC).

RESULTS

A model based solely on 14 deep neuroimaging features from CNNs achieved an extremely high AUC of 0.975 (95% confidence interval [CI] = 0.934-0.997) and significantly outperformed the model combining clinical, laboratory, and conventional imaging features (0.772, 95% CI = 0.691-0.847, p < 0.001) in prediction of functional outcomes. The deep neuroimaging model also demonstrated significant improvement over traditional prognostic scores. In an interpretability analysis, the deep neuroimaging features displayed a significant correlation with age, National Institutes of Health Stroke Scale score, infarct volume, and inflammation factors.

CONCLUSIONS

Deep learning models can successfully extract objective neuroimaging features from the routine radiological data in an automatic manner and aid in predicting the functional outcomes in patients with brainstem infarction at 3 months with very high accuracy.

Nijeholt G, van Es A, van der Lugt A, Majoie C, Burke J, Roozenbeek B, Lingsma H, Dippel D. Outcome Prediction Models for Endovascular Treatment of Ischemic Stroke: Systematic Review and External Validation

Supplemental Digital Content is available in the text.

Prediction models for outcome of patients with acute ischemic stroke who will undergo endovascular treatment have been developed to improve patient management. The aim of the current study is to provide an overview of preintervention models for functional outcome after endovascular treatment and to validate these models with data from daily clinical practice.

Value of machine learning to predict functional outcome of endovascular treatment for acute ischaemic stroke of the posterior circulation

PURPOSE

Clinical outcomes vary considerably among individuals with vessel occlusion of the posterior circulation. In the present study we evaluated machine learning algorithms in their ability to discriminate between favourable and unfavourable outcomes in patients with endovascular treatment of acute ischaemic stroke of the posterior circulation.

METHODS

This retrospective study evaluated three algorithms (generalised linear model, K-nearest neighbour and random forest) to predict functional outcomes at dismissal of 30 patients with acute occlusion of the basilar artery who were treated with thrombectomy. Input variables encompassed baseline as well as peri and postprocedural data. Favourable outcome was defined as a modified Rankin scale score of 0-2 and unfavourable outcome was defined as a modified Rankin scale score of 3-6. The performance of the algorithms was assessed with the area under the receiver operating curve and with confusion matrixes.

RESULTS

Successful reperfusion was achieved in 83%, with 30% of the patients having a favourable outcome. The area under the curve was 0.93 for the random forest model, 0.86 for the K-nearest neighbour model and 0.78 for the generalised linear model. The accuracy was 0.69 for the generalised linear model and 0.84 for the random forest and the K nearest neighbour models.

CONCLUSION

Favourable and unfavourable outcomes at dismissal of patients with acute ischaemic stroke of the posterior circulation can be predicted immediately after the follow-up non-enhanced computed tomography using machine learning.

Current uses, emerging applications, and clinical integration of artificial intelligence in neuroradiology

Artificial intelligence (AI) is a branch of computer science with a variety of subfields and techniques, exploited to serve as a deductive tool that performs tasks originally requiring human cognition. AI tools and its subdomains are being incorporated into healthcare delivery for the improvement of medical data interpretation encompassing clinical management, diagnostics, and prognostic outcomes. In the field of neuroradiology, AI manifested through deep machine learning and connected neural networks (CNNs) has demonstrated incredible accuracy in identifying pathology and aiding in diagnosis and prognostication in several areas of neurology and neurosurgery. In this literature review, we survey the available clinical data highlighting the utilization of AI in the field of neuroradiology across multiple neurological and neurosurgical subspecialties. In addition, we discuss the emerging role of AI in neuroradiology, its strengths and limitations, as well as future needs in strengthening its role in clinical practice. Our review evaluated data across several subspecialties of neurology and neurosurgery including vascular neurology, spinal pathology, traumatic brain injury (TBI), neuro-oncology, multiple sclerosis, Alzheimer's disease, and epilepsy. AI has established a strong presence within the realm of neuroradiology as a successful and largely supportive technology aiding in the interpretation, diagnosis, and even prognostication of various pathologies. More research is warranted to establish its full scientific validity and determine its maximum potential to aid in optimizing and providing the most accurate imaging interpretation.

Improving Ischemic Stroke Care With MRI and Deep Learning Artificial Intelligence

ABSTRACT

Advanced magnetic resonance imaging has been used as selection criteria for both acute ischemic stroke treatment and secondary prevention. The use of artificial intelligence, and in particular, deep learning, to synthesize large amounts of data and to understand better how clinical and imaging data can be leveraged to improve stroke care promises a new era of stroke care. In this article, we review common deep learning model structures for stroke imaging, evaluation metrics for model performance, and studies that investigated deep learning application in acute ischemic stroke care and secondary prevention.

Prediction of Progression to Severe Stroke in Initially Diagnosed Anterior Circulation Ischemic Cerebral Infarction

Purpose: Accurate prediction of the progression to severe stroke in initially diagnosed nonsevere patients with acute-subacute anterior circulation nonlacuna ischemic infarction (ASACNLII) is important in making clinical decision. This study aimed to apply a machine learning method to predict if the initially diagnosed nonsevere patients with ASACNLII would progress to severe stroke by using diffusion-weighted images and clinical information on admission. Methods: This retrospective study enrolled 344 patients with ASACNLII from June 2017 to August 2020 on admission, and 108 cases progressed to severe stroke during hospitalization within 3-21 days. The entire data were randomized into a training set (n = 271) and an independent test set (n = 73). A U-Net neural network was employed for automatic segmentation and volume measurement of the ischemic lesions. Predictive models were developed and used for evaluating the progression to severe stroke using different feature sets (the volume data, the clinical data, and the combination) and machine learning methods (random forest, support vector machine, and logistic regression). Results: The U-Net showed high correlation with manual segmentation in terms of Dice coefficient of 0.806 and R 2 value of the volume measurements of 0.960 in the test set. The random forest classifier of the volume + clinical combination achieved the best area under the receiver operating characteristic curve of 0.8358 (95% CI 0.7321-0.9269), and the accuracy, sensitivity, and specificity were 0.7780 (0.7397-0.7945), 0.7695 (0.6102-0.9074), and 0.8686 (0.6923-1.0), respectively. The Shapley additive explanation diagram showed the volume variable as the most important predictor. Conclusion: The U-Net was fully automatic and showed a high correlation with manual segmentation. An integrated approach combining clinical variables and stroke lesion volumes that were derived from the advanced machine learning algorithms had high accuracy in predicting the progression to severe stroke in ASACNLII patients.

Machine Learning for Clinical Outcome Prediction

Clinical decision-making in healthcare is already being influenced by predictions or recommendations made by data-driven machines. Numerous machine learning applications have appeared in the latest clinical literature, especially for outcome prediction models, with outcomes ranging from mortality and cardiac arrest to acute kidney injury and arrhythmia. In this review article, we summarize the state-of-the-art in related works covering data processing, inference, and model evaluation, in the context of outcome prediction models developed using data extracted from electronic health records. We also discuss limitations of prominent modeling assumptions and highlight opportunities for future research.

Artificial Intelligence and Deep Learning in Neuroradiology: Exploring the New Frontier

There have been many recently published studies exploring machine learning (ML) and deep learning applications within neuroradiology. The improvement in performance of these techniques has resulted in an ever-increasing number of commercially available tools for the neuroradiologist. In this narrative review, recent publications exploring ML in neuroradiology are assessed with a focus on several key clinical domains. In particular, major advances are reviewed in the context of: (1) intracranial hemorrhage detection, (2) stroke imaging, (3) intracranial aneurysm screening, (4) multiple sclerosis imaging, (5) neuro-oncology, (6) head and tumor imaging, and (7) spine imaging.