Dataset on acute stroke risk stratification from CT angiographic radiomics

Voxel-Wise Map of Intracerebral Hemorrhage Locations Associated With Worse Outcomes

BACKGROUND

Prior studies on the clinical impact of intracerebral hemorrhage (ICH) location have used visual localization of hematomas to neuroanatomical structures. However, hematomas often cross neuroanatomical structure boundaries with inter-reviewer variability in visual localization. To address these limitations, we applied voxel-wise analysis to identify brain regions where ICH presence is independently predictive of worse outcomes.

METHODS

We included consecutive patients with acute spontaneous ICH from a comprehensive stroke center in a derivation cohort and validated the results in patients from the control arm of a multicenter clinical trial. Using general linear models, we created and publicly shared a voxel-wise map of brain regions where ICH presence was associated with higher 3-month modified Rankin Scale scores, independent of hematoma volume and clinical risk factors. We also determined the optimal overlap threshold between baseline hematoma and voxel-wise map to categorize ICH location into high versus low risk.

RESULTS

Excluding those with missing variables, head computed tomography processing pipeline failure and poor scan quality, 559 of 780 patients were included in derivation (mean age, 69.3±14.5 years; 311 [55.6%] males) and 345 of 500 (mean age, 62.5±12.9 years; 206 [59.7%] males) in validation cohorts. In a voxel-wise analysis, ICH presence in deep white matter, thalami, caudate, midbrain, and pons was associated with worse outcomes. At the patient level, >22% overlap of baseline hematoma with voxel-wise map optimally binarized ICH location to high- versus low-risk categories. In both the derivation and validation cohorts, a high-risk ICH location was independently associated with worse outcomes (higher 3-month modified Rankin Scale score), after adjusting for patients' age, symptom severity at admission, baseline hematoma volume, and the presence of intraventricular hemorrhage, with adjusted odds ratios of 2 ([95% CI, 1.3-3.0] P=0.001) and 1.7 ([95% CI, 1.1-2.9] P=0.027), respectively.

CONCLUSIONS

We created and publicly shared a voxel-wise map of brain regions where hematoma presence predicts worse outcomes, independent of volume and clinical risk factors.

Predicting Functional Outcomes of Endovascular Thrombectomy in Acute Ischemic Stroke Using a Clinical-Radiomics Nomogram

BACKGROUND

Endovascular thrombectomy (EVT) is recommended for acute ischemic stroke due to large-vessel occlusion. However, approximately 50% of patients still experience poor outcomes after the procedure. This study aimed to assess whether a nomogram model that integrates computed tomography angiography radiomics features and clinical variables can predict EVT outcomes in patients with acute ischemic stroke.

METHODS

A total of 159 patients undergoing EVT were randomly divided into training and validation groups at a 7:3 ratio. A modified Rankin Scale score ≤ 2 at 90 days indicated a favorable outcome. We used univariate and multivariate logistic regression to identify analytic and radiomics predictors and create predictive models. Model performance was evaluated using the area under the curve, Hosmer-Lemeshow test, and decision curve analysis for discrimination, calibration, and clinical utility.

RESULTS

A 19-feature radiomics signature reached an area under the curve of 0.79. Combining it with age, baseline National Institutes of Health Stroke Scale score, diabetes, and statin use increased the area under the curve of the clinical-radiomics nomogram to 0.85. Both decision curve and calibration curve analyses showed strong performance.

CONCLUSIONS

Combining a radiomics nomogram with clinical predictors could effectively forecast EVT outcomes in patients with acute anterior circulation large vessel occlusion stroke.

The Evolution of Artificial Intelligence in Medical Imaging: From Computer Science to Machine and Deep Learning

Artificial intelligence (AI), the wide spectrum of technologies aiming to give machines or computers the ability to perform human-like cognitive functions, began in the 1940s with the first abstract models of intelligent machines. Soon after, in the 1950s and 1960s, machine learning algorithms such as neural networks and decision trees ignited significant enthusiasm. More recent advancements include the refinement of learning algorithms, the development of convolutional neural networks to efficiently analyze images, and methods to synthesize new images. This renewed enthusiasm was also due to the increase in computational power with graphical processing units and the availability of large digital databases to be mined by neural networks. AI soon began to be applied in medicine, first through expert systems designed to support the clinician's decision and later with neural networks for the detection, classification, or segmentation of malignant lesions in medical images. A recent prospective clinical trial demonstrated the non-inferiority of AI alone compared with a double reading by two radiologists on screening mammography. Natural language processing, recurrent neural networks, transformers, and generative models have both improved the capabilities of making an automated reading of medical images and moved AI to new domains, including the text analysis of electronic health records, image self-labeling, and self-reporting. The availability of open-source and free libraries, as well as powerful computing resources, has greatly facilitated the adoption of deep learning by researchers and clinicians. Key concerns surrounding AI in healthcare include the need for clinical trials to demonstrate efficacy, the perception of AI tools as 'black boxes' that require greater interpretability and explainability, and ethical issues related to ensuring fairness and trustworthiness in AI systems. Thanks to its versatility and impressive results, AI is one of the most promising resources for frontier research and applications in medicine, in particular for oncological applications.

Radiomics-Based Prediction of Collateral Status from CT Angiography of Patients Following a Large Vessel Occlusion Stroke

BACKGROUND

A major driver of individual variation in long-term outcomes following a large vessel occlusion (LVO) stroke is the degree of collateral arterial circulation. We aimed to develop and evaluate machine-learning models that quantify LVO collateral status using admission computed tomography angiography (CTA) radiomics.

METHODS

We extracted 1116 radiomic features from the anterior circulation territories from admission CTAs of 600 patients experiencing an acute LVO stroke. We trained and validated multiple machine-learning models for the prediction of collateral status based on consensus from two neuroradiologists as ground truth. Models were first trained to predict (1) good vs. intermediate or poor, or (2) good vs. intermediate or poor collateral status. Then, model predictions were combined to determine a three-tier collateral score (good, intermediate, or poor). We used the receiver operating characteristics area under the curve (AUC) to evaluate prediction accuracy.

RESULTS

We included 499 patients in training and 101 in an independent test cohort. The best-performing models achieved an averaged cross-validation AUC of 0.80 ± 0.05 for poor vs. intermediate/good collateral and 0.69 ± 0.05 for good vs. intermediate/poor, and AUC = 0.77 (0.67-0.87) and AUC = 0.78 (0.70-0.90) in the independent test cohort, respectively. The collateral scores predicted by the radiomics model were correlated with (rho = 0.45, p = 0.002) and were independent predictors of 3-month clinical outcome (p = 0.018) in the independent test cohort.

CONCLUSIONS

Automated tools for the assessment of collateral status from admission CTA-such as the radiomics models described here-can generate clinically relevant and reproducible collateral scores to facilitate a timely treatment triage in patients experiencing an acute LVO stroke.

Peculiar Neurological Examination Secondary to Persistent Primitive Hypoglossal Artery

A persistent primitive hypoglossal artery (PPHA) is an anatomical variant resulting in persistent carotid-vertebrobasilar anastomoses. This variant arises from the distal cervical segment of the internal carotid artery (ICA) between C1 and C3 and passes through an enlarged hypoglossal canal to join the basilar circulation. This case report describes a 60-year-old male with an acute ischemic event secondary to an occlusion in the right ICA and PPHA, resulting in a unique physical examination. Digital subtraction angiography (DSA) was utilized to visualize occlusion of the right common carotid artery with no residual right internal carotid artery or right vertebral artery filling. The patient's presenting symptoms yielded a unique neurological examination, making it difficult to localize a solitary lesion, which would account for the patient's acute presentation. In retrospect, with angiography revealing a right PPHA, his presentation fit more thoroughly with the clinical picture. With the increased utility of neuro-endovascular procedures, clinicians have a higher probability of encountering diverse angiographical findings. With this case report, we would like to familiarize practitioners with the presence of PPHA, present unique imaging findings involving typically isolated vascular territories, and stress the importance of clinical judgment when making decisions regarding stroke care.