Radiomics: A Primer on Processing Workflow and Analysis

Quantitative analysis of medical images can provide objective tools for diagnosis, prognostication, and disease monitoring. Radiomics refers to automated extraction of a large number of quantitative features from medical images for characterization of underlying pathologies. In this review, we will discuss the principles of radiomics, image preprocessing, feature extraction workflow, and statistical analysis. We will also address the limitations and future directions of radiomics.

Bedside detection of intracranial midline shift using portable magnetic resonance imaging

Neuroimaging is crucial for assessing mass effect in brain-injured patients. Transport to an imaging suite, however, is challenging for critically ill patients. We evaluated the use of a low magnetic field, portable MRI (pMRI) for assessing midline shift (MLS). In this observational study, 0.064 T pMRI exams were performed on stroke patients admitted to the neuroscience intensive care unit at Yale New Haven Hospital. Dichotomous (present or absent) and continuous MLS measurements were obtained on pMRI exams and locally available and accessible standard-of-care imaging exams (CT or MRI). We evaluated the agreement between pMRI and standard-of-care measurements. Additionally, we assessed the relationship between pMRI-based MLS and functional outcome (modified Rankin Scale). A total of 102 patients were included in the final study (48 ischemic stroke; 54 intracranial hemorrhage). There was significant concordance between pMRI and standard-of-care measurements (dichotomous, κ = 0.87; continuous, ICC = 0.94). Low-field pMRI identified MLS with a sensitivity of 0.93 and specificity of 0.96. Moreover, pMRI MLS assessments predicted poor clinical outcome at discharge (dichotomous: adjusted OR 7.98, 95% CI 2.07–40.04, p = 0.005; continuous: adjusted OR 1.59, 95% CI 1.11–2.49, p = 0.021). Low-field pMRI may serve as a valuable bedside tool for detecting mass effect.

The Initial Step Towards Establishing a Quantitative, Magnetic Resonance Imaging-Based Framework for Response Assessment of Spinal Metastases After Stereotactic Body Radiation Therapy

BACKGROUND

There are no established threshold values regarding the degree of growth on imaging when assessing response of spinal metastases treated with stereotactic body radiation therapy (SBRT).

OBJECTIVE

To determine a magnetic resonance imaging-based minimum detectable difference (MDD) in gross tumor volume (GTV) and its association with 1-yr radiation site-specific (RSS) progression-free survival (PFS).

METHODS

GTVs at baseline and first 2 post-SBRT scans (Post1 and Post2, respectively) for 142 spinal segments were contoured, and percentage volume change between scans calculated. One-year RSS PFS was acquired from medical records. The MDD was determined. The MDD was compared against optimal thresholds of GTV changes associated with 1-yr RSS PFS using Youden's J index, and receiver operating characteristic curves between timepoints compared to determine which timeframe had the best association.

RESULTS

A total of 17 of the 142 segments demonstrated progression. The MDD was 10.9%. Baseline-Post2 demonstrated the best performance (area under the curve [AUC] 0.90). Only Baseline-Post2 had an optimal threshold > MDD at 14.7%. Due to large distribution of GTVs, volumes were split into tertiles. Small tumors (GTV < 2 cc) had optimal thresholds of 42.0%, 71.3%, and 37.2% at Baseline-Post1 (AUC 0.81), Baseline-Post2 (AUC 0.89), and Post1-Post2 (AUC 0.77), respectively. Medium tumors (2 ≤ GTV ≤ 8.3 cc) all demonstrated optimal thresholds < MDD, with AUCs ranging from 0.65 to 0.84. Large tumors (GTV > 8.3 cc) had 2 timepoints where optimal thresholds > MDD: Baseline-Post2 (13.3%; AUC 0.97) and Post1-Post2 (11.8%; AUC 0.66). Baseline-Post2 had the best association with RSS PFS for all tertiles.

CONCLUSION

Given a MDD of 10.9%, for small GTVs, larger (>37%) changes were required before local failure could be determined, compared to 11% to 13% for medium/large tumors.

Similar admission NIHSS may represent larger tissue-at-risk in patients with right-sided versus left-sided large vessel occlusion

BACKGROUND

We investigated the effects of the side of large vessel occlusion (LVO) on post-thrombectomy infarct volume and clinical outcome with regard to admission National Institutes of Health Stroke Scale (NIHSS) score.

METHODS

We retrospectively identified patients with anterior LVO who received endovascular thrombectomy and follow-up MRI. Applying voxel-wise general linear models and multivariate analysis, we assessed the effects of occlusion side, admission NIHSS, and post-thrombectomy reperfusion (modified Thrombolysis in Cerebral Infarction, mTICI) on final infarct distribution and volume as well as discharge modified Rankin Scale (mRS) score.

RESULTS

We included 469 patients, 254 with left-sided and 215 with right-sided LVO. Admission NIHSS was higher in those with left-sided LVO (median (IQR) 16 (10-22)) than in those with right-sided LVO (14 (8-16), p>0.001). In voxel-wise analysis, worse post-thrombectomy reperfusion, lower admission NIHSS score, and poor discharge outcome were associated with right-hemispheric infarct lesions. In multivariate analysis, right-sided LVO was an independent predictor of larger final infarct volume (p=0.003). There was a significant three-way interaction between admission stroke severity (based on NIHSS), LVO side, and mTICI with regard to final infarct volume (p=0.041). Specifically, in patients with moderate stroke (NIHSS 6-15), incomplete reperfusion (mTICI 0-2b) was associated with larger final infarct volume (p<0.001) and worse discharge outcome (p=0.02) in right-sided compared with left-sided LVO.

CONCLUSIONS

When adjusted for admission NIHSS, worse post-thrombectomy reperfusion is associated with larger infarct volume and worse discharge outcome in right-sided versus left-sided LVO. This may represent larger tissue-at-risk in patients with right-sided LVO when applying admission NIHSS as a clinical biomarker for penumbra.

OTHR-15. Assessment of TRIPOD adherence in articles developing machine learning models for differentiation of glioma from brain metastasis

Purpose

Machine learning (ML) applications in predictive models in neuro-oncology have become an increasingly investigated subject of research. For their incorporation into clinical practice, rigorous assessment is needed to reduce bias. Several reports have indicated utility of ML applications in differentiation of glioma from brain metastasis. However, a systematic assessment of quality of methodology and reporting in these studies has not been done yet. We examined the adherence of 29 published reports in this field to the TRIPOD statement, which is similar to CLAIM checklist.

Materials and Methods

Our systematic review was conducted in accordance with PRISMA guidelines. Ovid Embase, Ovid MEDLINE, Cochrane trials (CENTRAL) and Web of science core-collection were searched. Keywords included artificial intelligence, machine learning, deep learning, radiomics, magnetic resonance imaging, glioma, and glioblastoma. Assessment of TRIPOD adherence in 29 eligible studies was performed. Individual item performance was assessed by adherence index (ADI), the ratio of mean achieved score to maximum score per TRIPOD item.

Results

In a preliminary analysis of 8 studies, the average TRIPOD adherence score was 0.48 (14.25/30 items fulfilled) with individual scores ranging from 0.27 (8/30) to 0.60 (18/30). Best overall item performance, with an ADI of 1, was seen in item 3 (Background/Objectives), 16 (Model performance) and 19 (Interpretation). Poorest performance was detected in item 1 (Title) and 2 (Abstract), followed by item 9 (Missing Data) with ADI of 0, 0 and 0.13, respectively.

Conclusion

Preliminary results underline the lack of reproducibility in ML studies on distinction between glioma and brain metastasis. An average TRIPOD adherence score of 0.48 indicates insufficient quality of reporting and outlines the need for increased utilization of quality scoring systems in study documentation. Systematic evaluation of quality score adherence will allow us to identify common flaws in this field for enabling translation of models into clinical workflow.

Rapid Assessment of Blood Pressure Variability and Outcome After Successful Thrombectomy

BACKGROUND AND PURPOSE

High blood pressure (BP) variability after endovascular stroke therapy is associated with poor outcome. Conventional BP variability measures require long recordings, limiting their utility as a risk assessment tool to guide clinical decision-making. Here, we performed rapid assessment of BP variability by spectral analysis and evaluated its association with early clinical improvement and long-term functional outcomes.

METHODS

We conducted a prospective study of 146 patients with anterior circulation ischemic stroke who underwent successful endovascular stroke therapy. Spectral analysis of 5-minute recordings of beat-to-beat BP was used to quantify BP variability. Outcomes included initial clinical response and modified Rankin Scale at 90 days.

RESULTS

Increased BP variability at high frequencies was independently associated with poor functional outcome at 90 days (adjusted odds ratio [aOR], 1.85 [95% CI, 1.07-3.25], P=0.03; low-/high-frequency ratio aOR, 0.67 [95% CI, 0.46-0.92], P=0.02) and reduced likelihood of an early neurological recovery (aOR, 0.62 [95% CI, 0.44-0.91], P=0.01 and aOR, 1.37 [95% CI, 1.03-1.87], P=0.04, respectively).

CONCLUSIONS

High-frequency BP oscillations after successful reperfusion may be harmful and associate with a decreased likelihood of neurological recovery and favorable functional outcomes. Rapid assessment of BP variability throughout the postreperfusion period is feasible and may allow for a more personalized BP management.

Site-Specific Variation in Radiomic Features of Head and Neck Squamous Cell Carcinoma and Its Impact on Machine Learning Models

Simple Summary

Head and neck squamous cell carcinoma (HNSCC) is the most common mucosal malignancy of the head and neck and a leading cause of cancer death. HNSCC arises from different primary anatomical locations that are typically combined during radiomic analyses assuming that the radiomic features, i.e., quantitative image-based features, are similar based on histopathologic characteristics. However, whether these quantitative features are comparable across tumor sites remains unknown. The aim of our retrospective study was to assess if systematic differences exist between radiomic features based on different tumor sites in HNSCC and how they might affect machine learning model performance in endpoint prediction. Using a population of 605 HNSCC patients, we observed significant differences in radiomic features of tumors from different locations and showed that these differences can impact machine learning model performance. This suggests that tumor site should be considered when developing and evaluating radiomics-based models.

Abstract

Current radiomic studies of head and neck squamous cell carcinomas (HNSCC) are typically based on datasets combining tumors from different locations, assuming that the radiomic features are similar based on histopathologic characteristics. However, molecular pathogenesis and treatment in HNSCC substantially vary across different tumor sites. It is not known if a statistical difference exists between radiomic features from different tumor sites and how they affect machine learning model performance in endpoint prediction. To answer these questions, we extracted radiomic features from contrast-enhanced neck computed tomography scans (CTs) of 605 patients with HNSCC originating from the oral cavity, oropharynx, and hypopharynx/larynx. The difference in radiomic features of tumors from these sites was assessed using statistical analyses and Random Forest classifiers on the radiomic features with 10-fold cross-validation to predict tumor sites, nodal metastasis, and HPV status. We found statistically significant differences (p-value ≤ 0.05) between the radiomic features of HNSCC depending on tumor location. We also observed that differences in quantitative features among HNSCC from different locations impact the performance of machine learning models. This suggests that radiomic features may reveal biologic heterogeneity complementary to current gold standard histopathologic evaluation. We recommend considering tumor site in radiomic studies of HNSCC.

Admission computed tomography radiomic signatures outperform hematoma volume in predicting baseline clinical severity and functional outcome in the ATACH-2 trial intracerebral hemorrhage population

BACKGROUND AND PURPOSE

Radiomics provides a framework for automated extraction of high-dimensional feature sets from medical images. We aimed to determine radiomics signature correlates of admission clinical severity and medium-term outcome from intracerebral hemorrhage (ICH) lesions on baseline head computed tomography (CT).

METHODS

We used the ATACH-2 (Antihypertensive Treatment of Acute Cerebral Hemorrhage II) trial dataset. Patients included in this analysis (n = 895) were randomly allocated to discovery (n = 448) and independent validation (n = 447) cohorts. We extracted 1130 radiomics features from hematoma lesions on baseline noncontrast head CT scans and generated radiomics signatures associated with admission Glasgow Coma Scale (GCS), admission National Institutes of Health Stroke Scale (NIHSS), and 3-month modified Rankin Scale (mRS) scores. Spearman's correlation between radiomics signatures and corresponding target variables was compared with hematoma volume.

RESULTS

In the discovery cohort, radiomics signatures, compared to ICH volume, had a significantly stronger association with admission GCS (0.47 vs. 0.44, p = 0.008), admission NIHSS (0.69 vs. 0.57, p < 0.001), and 3-month mRS scores (0.44 vs. 0.32, p < 0.001). Similarly, in independent validation, radiomics signatures, compared to ICH volume, had a significantly stronger association with admission GCS (0.43 vs. 0.41, p = 0.02), NIHSS (0.64 vs. 0.56, p < 0.001), and 3-month mRS scores (0.43 vs. 0.33, p < 0.001). In multiple regression analysis adjusted for known predictors of ICH outcome, the radiomics signature was an independent predictor of 3-month mRS in both cohorts.

CONCLUSIONS

Limited by the enrollment criteria of the ATACH-2 trial, we showed that radiomics features quantifying hematoma texture, density, and shape on baseline CT can provide imaging correlates for clinical presentation and 3-month outcome. These findings couldtrigger a paradigm shift where imaging biomarkers may improve current modelsfor prognostication, risk-stratification, and treatment triage of ICH patients.

Early head CT in post-cardiac arrest patients: A helpful tool or contributor to self-fulfilling prophecy?

OBJECTIVE

Neuroprognostication guidelines suggest that early head computed tomography (HCT) might be useful in the evaluation of cardiac arrest (CA) patients following return of spontaneous circulation. We aimed to determine the impact of early HCT, performed within the first 6 h following CA, on decision-making following resuscitation.

METHODS

We identified a cohort of initially unconscious post-CA patients at a tertiary care academic medical center from 2012 to 2017. Variables pertaining to demographics, CA details, post-CA care, including neuroimaging and neurophysiologic testing, were abstracted retrospectively from the electronic medical records. Changes in management resulting from HCT findings were recorded. Blinded board-certified neurointensivists adjudicated HCT findings related to hypoxic-ischemic brain injury (HIBI) burden. The gray-white matter ratio (GWR) was also calculated.

RESULTS

Of 302 patients, 182 (60.2%) underwent HCT within six hours of CA (early HCT group). Approximately 1 in 4 early HCTs were abnormal (most commonly HIBI changes; 78.7%, n = 37), which resulted in a change in management in nearly half of cases (46.8%, n = 22). The most common changes in management were de-escalation in care [including transition to do not resuscitate status), withholding targeted temperature management, and withdrawal of life sustaining therapy (WLST)]. In cases with radiographic HIBI, mean [standard deviation] GWR was lower (1.20 [0.10] vs 1.30 [0.09], P < 0.001) and progression to brain death was higher (44.4% vs 2.9%; P < 0.001). The inter-rater reliability (IRR) of early HCT to determine presence of HIBI between radiology and three neurointensivists had a wide range (κ 0.13-0.66).

CONCLUSION

Early HCT identified abnormalities in 25% of cases and frequently influenced therapeutic decisions. Neuroimaging interpretation discrepancies between radiology and neurointensivists are common and agreement on severity of HIBI on early HCT is poor (k 0.11).

Imaging of Spontaneous Intracerebral Hemorrhage

Primary or nontraumatic spontaneous intracerebral hemorrhage (ICH) comprises approximately 15% to 20% of all stroke. ICH has a mortality of approximately 40% within the first month, and 75% mortality and morbidity rate within the first year. Despite reduction in overall stroke incidence, hemorrhagic stroke incidence has remained steady since 1980. Neuroimaging is critical in detection of ICH, determining the underlying cause, identification of patients at risk of hematoma expansion, and directing the treatment strategy. This article discusses the neuroimaging methods of ICH, imaging markers for clinical outcome prediction, and future research directions with attention to the latest evidence-based guidelines.

Fluid dynamics simulations show that facial masks can suppress the spread of COVID-19 in indoor environments

The Coronavirus disease outbreak of 2019 has been causing significant loss of life and unprecedented economical loss throughout the world. Social distancing and face masks are widely recommended around the globe in order to protect others and prevent the spread of the virus through breathing, coughing, and sneezing. To expand the scientific underpinnings of such recommendations, we carry out high-fidelity computational fluid dynamics simulations of unprecedented resolution and realism to elucidate the underlying physics of saliva particulate transport during human cough with and without facial masks. Our simulations: (a) are carried out under both a stagnant ambient flow (indoor) and a mild unidirectional breeze (outdoor); (b) incorporate the effect of human anatomy on the flow; (c) account for both medical and non-medical grade masks; and (d) consider a wide spectrum of particulate sizes, ranging from 10 micro m to 300 micro m. We show that during indoor coughing some saliva particulates could travel up to 0.48 m, 0.73 m, and 2.62 m for the cases with medical-grade, non-medical grade, and without facial masks, respectively. Thus, in indoor environments either medical or non-medical grade facial masks can successfully limit the spreading of saliva particulates to others. Under outdoor conditions with a unidirectional mild breeze, however, leakage flow through the mask can cause saliva particulates to be entrained into the energetic shear layers around the body and transported very fast at large distances by the turbulent flow, thus, limiting the effectiveness of facial masks.

Choice of imaging modality for pre-treatment staging of head and neck cancer impacts TNM staging

PURPOSE

The purpose of this retrospective cohort study was to determine whether there is a difference in the sensitivity of chest computed tomography (CT) versus ¹⁸F-fluorodeoxyglucose positron emission tomography with low-dose nonenhanced CT (¹⁸F-FDG PET/CT or PET/CT) in the detection of distant metastases in head and neck cancer, within a tertiary care setting.

MATERIALS AND METHODS

Patients with head and neck cancer, and known distant metastases, who underwent both ¹⁸F-FDG PET/CT with integrated low-dose nonenhanced CT and diagnostic chest CT prior to initiation of therapy from 2008 to 2017 were included. Two head and neck radiologists, blinded to all patient information and to each other's readings, reviewed the PET/CT or CT chest images for each patient and identified whether distant metastases were present. No radiologist read both modalities for a single patient. Concordance between imaging modalities was quantitatively analyzed using McNemar's test.

RESULTS

27 patients were included. McNemar's mid p-value analysis showed no significant difference in the detection of distant metastases (p = .6875). However, PET/CT detected distant metastases in three patients that chest CT did not, while chest CT identified distant metastatic disease in two patients that were negative on PET/CT.

CONCLUSIONS

While this study did not identify a statistically significant difference in sensitivity, five patients had distant metastases identified on only one of the two modalities. Use of a single modality would have resulted in inaccurate staging in 7-11% of patients in our study. The use of both modalities offers the greatest accuracy when providing stage-adapted oncologic treatment.

Prediction of post-radiotherapy locoregional progression in HPV-associated oropharyngeal squamous cell carcinoma using machine-learning analysis of baseline PET/CT radiomics

•

Radiomics quantitatively captures visually inappreciable imaging features.

•

PET/CT radiomics provides wholistic metabolic and structural tumor characterization.

•

Machine-learning algorithms can generate radiomics-based biomarkers for OPSCC.

•

PET/CT radiomics can predict post-radiotherapy locoregional progression in HPV-associated OPSCC.

•

Such biomarkers may improve patient selection for treatment de-intensification trials.

Locoregional failure remains a therapeutic challenge in oropharyngeal squamous cell carcinoma (OPSCC). We aimed to devise novel objective imaging biomarkers for prediction of locoregional progression in HPV-associated OPSCC. Following manual lesion delineation, 1037 PET and 1037 CT radiomic features were extracted from each primary tumor and metastatic cervical lymph node on baseline PET/CT scans. Applying random forest machine-learning algorithms, we generated radiomic models for censoring-aware locoregional progression prognostication (evaluated by Harrell's C-index) and risk stratification (evaluated in Kaplan-Meier analysis). A total of 190 patients were included; an optimized model yielded a median (interquartile range) C-index of 0.76 (0.66-0.81; p = 0.01) in prognostication of locoregional progression, using combined PET/CT radiomic features from primary tumors. Radiomics-based risk stratification reliably identified patients at risk for locoregional progression within 2-, 3-, 4-, and 5-year follow-up intervals, with log-rank p-values of p = 0.003, p = 0.001, p = 0.02, p = 0.006 in Kaplan-Meier analysis, respectively. Our results suggest PET/CT radiomic biomarkers can predict post-radiotherapy locoregional progression in HPV-associated OPSCC. Pending validation in large, independent cohorts, such objective biomarkers may improve patient selection for treatment de-intensification trials in this prognostically favorable OPSCC entity, and eventually facilitate personalized therapy.

Effects of Collateral Status on Infarct Distribution Following Endovascular Therapy in Large Vessel Occlusion Stroke

BACKGROUND AND PURPOSE

We aim to examine effects of collateral status and post-thrombectomy reperfusion on final infarct distribution and early functional outcome in patients with anterior circulation large vessel occlusion ischemic stroke.

METHODS

Patients with large vessel occlusion who underwent endovascular intervention were included in this study. All patients had baseline computed tomography angiography and follow-up magnetic resonance imaging. Collateral status was graded according to the criteria proposed by Miteff et al and reperfusion was assessed using the modified Thrombolysis in Cerebral Infarction (mTICI) system. We applied a multivariate voxel-wise general linear model to correlate the distribution of final infarction with collateral status and degree of reperfusion. Early favorable outcome was defined as a discharge modified Rankin Scale score ≤2.

RESULTS

Of the 283 patients included, 129 (46%) had good, 97 (34%) had moderate, and 57 (20%) had poor collateral status. Successful reperfusion (mTICI 2b/3) was achieved in 206 (73%) patients. Poor collateral status was associated with infarction of middle cerebral artery border zones, whereas worse reperfusion (mTICI scores 0-2a) was associated with infarction of middle cerebral artery territory deep white matter tracts and the posterior limb of the internal capsule. In multivariate regression models, both mTICI (P<0.001) and collateral status (P<0.001) were among independent predictors of final infarct volumes. However, mTICI (P<0.001), but not collateral status (P=0.058), predicted favorable outcome at discharge.

CONCLUSIONS

In this cohort of patients with large vessel occlusion stroke, both the collateral status and endovascular reperfusion were strongly associated with middle cerebral artery territory final infarct volumes. Our findings suggesting that baseline collateral status predominantly affected middle cerebral artery border zones infarction, whereas higher mTICI preserved deep white matter and internal capsule from infarction; may explain why reperfusion success-but not collateral status-was among the independent predictors of favorable outcome at discharge. Infarction of the lentiform nuclei was observed regardless of collateral status or reperfusion success.

Potential Added Value of PET/CT Radiomics for Survival Prognostication beyond AJCC 8th Edition Staging in Oropharyngeal Squamous Cell Carcinoma

Accurate risk-stratification can facilitate precision therapy in oropharyngeal squamous cell carcinoma (OPSCC). We explored the potential added value of baseline positron emission tomography (PET)/computed tomography (CT) radiomic features for prognostication and risk stratification of OPSCC beyond the American Joint Committee on Cancer (AJCC) 8th edition staging scheme. Using institutional and publicly available datasets, we included OPSCC patients with known human papillomavirus (HPV) status, without baseline distant metastasis and treated with curative intent. We extracted 1037 PET and 1037 CT radiomic features quantifying lesion shape, imaging intensity, and texture patterns from primary tumors and metastatic cervical lymph nodes. Utilizing random forest algorithms, we devised novel machine-learning models for OPSCC progression-free survival (PFS) and overall survival (OS) using “radiomics” features, “AJCC” variables, and the “combined” set as input. We designed both single- (PET or CT) and combined-modality (PET/CT) models. Harrell’s C-index quantified survival model performance; risk stratification was evaluated in Kaplan–Meier analysis. A total of 311 patients were included. In HPV-associated OPSCC, the best “radiomics” model achieved an average C-index ± standard deviation of 0.62 ± 0.05 (p = 0.02) for PFS prediction, compared to 0.54 ± 0.06 (p = 0.32) utilizing “AJCC” variables. Radiomics-based risk-stratification of HPV-associated OPSCC was significant for PFS and OS. Similar trends were observed in HPV-negative OPSCC. In conclusion, radiomics imaging features extracted from pre-treatment PET/CT may provide complimentary information to the current AJCC staging scheme for survival prognostication and risk-stratification of HPV-associated OPSCC.

A Fully Automated Method for Segmenting Arteries and Quantifying Vessel Radii on Magnetic Resonance Angiography Images of Varying Projection Thickness

PURPOSE

Precise quantification of cerebral arteries can help with differentiation and prognostication of cerebrovascular disease. Existing image processing and segmentation algorithms for magnetic resonance angiography (MRA) are limited to the analysis of either 2D maximum intensity projection images or the entire 3D volume. The goal of this study was to develop a fully automated, hybrid 2D-3D method for robust segmentation of arteries and accurate quantification of vessel radii using MRA at varying projection thicknesses.

METHODS

A novel algorithm that employs an adaptive Frangi filter for segmentation of vessels followed by estimation of vessel radii is presented. The method was evaluated on MRA datasets and corresponding manual segmentations from three healthy subjects for various projection thicknesses. In addition, the vessel metrics were computed in four additional subjects. Three synthetically generated angiographic datasets resembling brain vasculature were also evaluated under different noise levels. Dice similarity coefficient, Jaccard Index, F-score, and concordance correlation coefficient were used to measure the segmentation accuracy of manual versus automatic segmentation.

RESULTS

Our new adaptive filter rendered accurate representations of vessels, maintained accurate vessel radii, and corresponded better to manual segmentation at different projection thicknesses than prior methods. Validation with synthetic datasets under low contrast and noisy conditions revealed accurate quantification of vessels without distortions.

CONCLUSION

We have demonstrated a method for automatic segmentation of vascular trees and the subsequent generation of a vessel radii map. This novel technique can be applied to analyze arterial structures in healthy and diseased populations and improve the characterization of vascular integrity.

Applications of radiomics in precision diagnosis, prognostication and treatment planning of head and neck squamous cell carcinomas

Recent advancements in computational power, machine learning, and artificial intelligence technology have enabled automated evaluation of medical images to generate quantitative diagnostic and prognostic biomarkers. Such objective biomarkers are readily available and have the potential to improve personalized treatment, precision medicine, and patient selection for clinical trials. In this article, we explore the merits of the most recent addition to the “-omics” concept for the broader field of head and neck cancer – “Radiomics”. This review discusses radiomics studies focused on (molecular) characterization, classification, prognostication and treatment guidance for head and neck squamous cell carcinomas (HNSCC). We review the underlying hypothesis, general concept and typical workflow of radiomic analysis, and elaborate on current and future challenges to be addressed before routine clinical application.

Machine Learning Decision Tree Models for Differentiation of Posterior Fossa Tumors Using Diffusion Histogram Analysis and Structural MRI Findings

We applied machine learning algorithms for differentiation of posterior fossa tumors using apparent diffusion coefficient (ADC) histogram analysis and structural MRI findings. A total of 256 patients with intra-axial posterior fossa tumors were identified, of whom 248 were included in machine learning analysis, with at least 6 representative subjects per each tumor pathology. The ADC histograms of solid components of tumors, structural MRI findings, and patients' age were applied to construct decision models using Classification and Regression Tree analysis. We also compared different machine learning classification algorithms (i.e., naïve Bayes, random forest, neural networks, support vector machine with linear and polynomial kernel) for dichotomized differentiation of the 5 most common tumors in our cohort: metastasis (n = 65), hemangioblastoma (n = 44), pilocytic astrocytoma (n = 43), ependymoma (n = 27), and medulloblastoma (n = 26). The decision tree model could differentiate seven tumor histopathologies with terminal nodes yielding up to 90% accurate classification rates. In receiver operating characteristics (ROC) analysis, the decision tree model achieved greater area under the curve (AUC) for differentiation of pilocytic astrocytoma (p = 0.020); and atypical teratoid/rhabdoid tumor ATRT (p = 0.001) from other types of neoplasms compared to the official clinical report. However, neuroradiologists' interpretations had greater accuracy in differentiating metastases (p = 0.001). Among different machine learning algorithms, random forest models yielded the highest accuracy in dichotomized classification of the 5 most common tumor types; and in multiclass differentiation of all tumor types random forest yielded an averaged AUC of 0.961 in training datasets, and 0.873 in validation samples. Our study demonstrates the potential application of machine learning algorithms and decision trees for accurate differentiation of brain tumors based on pretreatment MRI. Using easy to apply and understandable imaging metrics, the proposed decision tree model can help radiologists with differentiation of posterior fossa tumors, especially in tumors with similar qualitative imaging characteristics. In particular, our decision tree model provided more accurate differentiation of pilocytic astrocytomas from ATRT than by neuroradiologists in clinical reads.

Abstract TP11: CTA Collateral Status and Final Infarct Distribution Following Thrombectomy in Stroke Patients With Large Vessel Occlusion

Aim: The angiographic collateral status is a major predictor of final infarct volume in patients with large vessel occlusion (LVO). In this study, we assessed the effects of collateral status on final infarct lesion distribution after thrombectomy.

Methods: Acute ischemic stroke patients with occluded terminal ICA and/or MCA M1 segment who underwent thrombectomy and had a follow up MRI within a week were included. The angiographic collateral status was evaluated on pre-thrombectomy CTA and graded according to Miteff et al. (Brain 2009;132(8):2231-8). The final infarct lesion was segmented on DWI; and using voxel-wise general linear model, we determined the correlation of final infarct volume with post-thrombectomy TICI (thrombolysis in cerebral infarction) score, and collateral status - as a covariate.

Results: Among 106 patients with terminal ICA and/or MCA M1 occlusion in analysis, final infarct volume had a significant correlation with TICI reperfusion score (rho=0.384, p<0.001), CTA collaterals (rho=0.221, p=0.023), and TICIxCollaterals interaction term (rho=0.446, p<0.001). Voxel-wise analysis (Figure) showed that better reperfusion after thrombectomy (i.e. higher TICI) was associated with preservation of MCA territory cortex and deep white matter (green). The voxel-wise interaction analysis of TICI and CTA collateral status showed that poor collateral status is associated with infarction of the MCA-PCA border zone (red). Alternatively, good collaterals may preserve the peripheral edges of the MCA territory and MCA-ACA border zone (blue).

Conclusion: A successful thrombectomy in LVO stroke patients can preserve the cortical and deep white matter of MCA territory - including eloquent speech and motor regions - while CTA collateral status mainly determines the fate of the MCA-PCA border zone. On the other hand, lentiform nuclei tend to infarct despite successful reperfusion and good CTA collateral status.

Abstract WMP101: Prediction of Clinical Outcome in Supratentorial Intracerebral Hemorrhage: Application of Baseline Ct Scan Radiomics Feature Extraction and Machine Learning Classifiers

Aim: Radiomics refers to automatic extraction of numerous quantitative features from medical images to supplement visual assessment. Machine-learning algorithms provide a suitable statistical methodology for devising predictive classifiers based on large radiomics datasets. We aimed to predict intracerebral hemorrhage (ICH) outcome by applying machine-learning classifiers to both clinical data and hematoma radiomics features.

Methods: Patients enrolled in the Yale Longitudinal Study of ICH were included if they had (1) spontaneous supratentorial ICH, (2) baseline CT scan, (3) known admission Glasgow Coma Scale (GCS), and (4) 3-month modified Rankin Scale (mRS). A total of 1134 radiomics features related to the intensity, shape, texture, and waveform were extracted from manually segmented ICH lesions on baseline CT. Clinical variables were patients’ age, gender, GCS, presence of intraventricular hemorrhage, and thalamic ICH. We calculated the averaged receiver operating characteristics (ROC) area under curve (AUC) in outcome prediction among 100 repeats of 5-fold cross-validation (x500 iterations) for different combinations of feature selection and machine-learning algorithms.

Results: A total of 119 ICH patients were included, of whom 60 had poor outcome (mRS ≥4). Among different combinations, lasso regression feature selection and partial least square (PLS) classification model yielded the highest accuracy in outcome prediction (Figure), with an averaged (95% confidence interval) ROC AUC of 0.86 (0.83 - 0.89) using clinical variables “only”, versus 0.92 (0.89 - 0.95) using combination of clinical variables and 54 radiomics features selected by lasso regression. Among radiomics features selected by lasso regression, ICH lesion flatness had the highest variable importance and was the only shape feature selected.

Conclusion: Addition of ICH lesion radiomics to clinical variables using machine-learning models can improve outcome prediction.

Comparative Effectiveness of Endovascular Thrombectomy in Elderly Stroke Patients

Background and Purpose- Strokes in patients aged ≥80 years are common, and advanced age is associated with relatively poor poststroke functional outcome. The current guidelines do not recommend an upper age limit for endovascular thrombectomy (EVT). The purpose of this study is to evaluate the effectiveness of EVT in acute stroke because of large vessel occlusion for elderly patients >age 80 years. Methods- A Markov decision analytic model was constructed from a societal perspective to evaluate health outcomes in terms of quality-adjusted life years (QALYs) after EVT for acute ischemic stroke because of large vessel occlusion in patients above age 80 years. Age-specific input parameters were obtained from the most recent/comprehensive literature. Good outcome was defined as a modified Rankin Scale score ≤2. Probabilistic, 1-way, and 2-way sensitivity analyses were performed for both healthy patients and patients with disability at baseline. Results- Base case calculation showed in functionally independent patients at baseline, intravenous thrombolysis (IVT) with tPA (tissue-type plasminogen activator) only to be the better strategy with 3.76 QALYs compared to 2.93 QALYs for patients undergoing EVT. The difference in outcome is 0.83 QALY (equivalent to 303 days of life in perfect health). For patients with baseline disability, IVT only yields a utility of 1.92 QALYs and EVT yields a utility of 1.65 QALYs. The difference is 0.27 QALYs (equivalent to 99 days of life in perfect health). Multiple sensitivity analyses showed that the effectiveness of EVT is significantly determined by the morbidity and mortality after both IVT and EVT strategies, respectively. Conclusions- Our study demonstrates the impact of relevant factors on the effectiveness of EVT in patients above 80 years of age. Morbidity and mortality after both IVT and EVT strategies significantly influence the outcomes in both healthy and disabled patients at baseline. Better identification of patients not benefiting from IVT would optimize the selective use of EVT thereby improving its effectiveness.

Poor Outcomes Related to Anterior Extension of Large Hemispheric Infarction: Topographic Analysis of GAMES-RP Trial MRI Scans

BACKGROUND

We aimed to assess the correlation of lesion location and clinical outcome in patients with large hemispheric infarction (LHI).

METHODS

We analyzed admission MRI data from the GAMES-RP trial, which enrolled patients with anterior circulation infarct volumes of 82-300 cm³ within 10 hours of onset. Infarct lesions were segmented and co-registered onto MNI-152 brain space. Voxel-wise general linear models were applied to assess location-outcome correlations after correction for infarct volume as a co-variate.

RESULTS

We included 83 patients with known 3-month modified Rankin scale (mRS). In voxel-wise analysis, there was significant correlation between admission infarct lesions involving the anterior cerebral artery (ACA) territory and its middle cerebral artery (MCA) border zone with both higher 3-month mRS and post-stroke day 3 and 7 National Institutes of Health Stroke Scale (NIHSS) total score and arm/leg subscores. Higher NIHSS total scores from admission through poststroke day 2 correlated with left MCA infarcts. In multivariate analysis, ACA territory infarct volume (P = .001) and admission NIHSS (P = .005) were independent predictors of 3-month mRS. Moreover, in a subgroup of 36 patients with infarct lesions involving right MCA-ACA border zone, intravenous (IV) glibenclamide (BIIB093; glyburide) treatment was the only independent predictor of 3-month mRS in multivariate regression analysis (P = .016).

CONCLUSIONS

Anterior extension of LHI with involvement of ACA territory and ACA-MCA border zone is an independent predictor of poor functional outcome, likely due to impairment of arm/leg motor function. If confirmed in larger cohorts, infarct topology may potentially help triage LHI patients who may benefit from IV glibenclamide.

CLINICAL TRIAL REGISTRATION

URL: https://www.clinicaltrials.gov. Unique identifier: NCT01794182.

Stroke atlas of the brain: Voxel-wise density-based clustering of infarct lesions topographic distribution

Objective

The supply territories of main cerebral arteries are predominantly identified based on distribution of infarct lesions in patients with large arterial occlusion; whereas, there is no consensus atlas regarding the supply territories of smaller end-arteries. In this study, we applied a data-driven approach to construct a stroke atlas of the brain using hierarchical density clustering in large number of infarct lesions, assuming that voxels/regions supplied by a common end-artery tend to infarct together.

Methods

A total of 793 infarct lesions on MRI scans of 458 patients were segmented and coregistered to MNI-152 standard brain space. Applying a voxel-wise data-driven hierarchical density clustering algorithm, we identified those voxels that were most likely to be part of same infarct lesions in our dataset. A step-wise clustering scheme was applied, where the clustering threshold was gradually decreased to form the first 20 mother (>50 cm³) or main (1–50 cm³) clusters in addition to any possible number of tiny clusters (<1 cm³); and then, any resultant mother clusters were iteratively subdivided using the same scheme. Also, in a randomly selected 2/3 subset of our cohort, a bootstrapping cluster analysis with 100 permutations was performed to assess the statistical robustness of proposed clusters.

Results

Approximately 91% of the MNI-152 brain mask was covered by 793 infarct lesions across patients. The covered area of brain was parcellated into 4 mother, 16 main, and 123 tiny clusters at the first hierarchy level. Upon iterative clustering subdivision of mother clusters, the brain tissue was eventually parcellated into 1 mother cluster (62.6 cm³), 181 main clusters (total volume 1107.3 cm³), and 917 tiny clusters (total volume of 264.8 cm³). In bootstrap analysis, only 0.12% of voxels, were labelled as “unstable” – with a greater reachability distance in cluster scheme compared to their corresponding mean bootstrapped reachability distance. On visual assessment, the mother/main clusters were formed along supply territories of main cerebral arteries at initial hierarchical levels, and then tiny clusters emerged in deep white matter and gray matter nuclei prone to small vessel ischemic infarcts.

Conclusions

Applying voxel-wise data-driven hierarchical density clustering on a large number of infarct lesions, we have parcellated the brain tissue into clusters of voxels that tend to be part of same infarct lesion, and presumably representing end-arterial supply territories. This hierarchical stroke atlas of the brain is shared publicly, and can potentially be applied for future infarct location-outcome analysis.

•

Using data-driven density clustering, a hierarchical brain atlas is constructed to identify voxels likely to infarct together.

•

Different clusters can potentially be extracted from dendrogram through thresholding at different reachability thresholds.

•

The hierarchical stroke atlas hypothetically represents the detailed anatomical distribution of distal arties in the brain.

•

The stroke atlas is made publicly available for potential future location-outcome correlation studies in stroke patients.