A multi-institutional machine learning algorithm for prognosticating facial nerve injury following microsurgical resection of vestibular schwannoma.

Vestibular schwannomas (VS) are the most common tumor of the skull base with available treatment options that carry a risk of iatrogenic injury to the facial nerve, which can significantly impact patients' quality of life. As facial nerve outcomes remain challenging to prognosticate, we endeavored to utilize machine learning to decipher predictive factors relevant to facial nerve outcomes following microsurgical resection of VS. A database of patient-, tumor- and surgery-specific features was constructed via retrospective chart review of 242 consecutive patients who underwent microsurgical resection of VS over a 7-year study period. This database was then used to train non-linear supervised machine learning classifiers to predict facial nerve preservation, defined as House-Brackmann (HB) I vs. facial nerve injury, defined as HB II-VI, as determined at 6-month outpatient follow-up. A random forest algorithm demonstrated 90.5% accuracy, 90% sensitivity and 90% specificity in facial nerve injury prognostication. A random variable (rv) was generated by randomly sampling a Gaussian distribution and used as a benchmark to compare the predictiveness of other features. This analysis revealed age, body mass index (BMI), case length and the tumor dimension representing tumor growth towards the brainstem as prognosticators of facial nerve injury. When validated via prospective assessment of facial nerve injury risk, this model demonstrated 84% accuracy. Here, we describe the development of a machine learning algorithm to predict the likelihood of facial nerve injury following microsurgical resection of VS. In addition to serving as a clinically applicable tool, this highlights the potential of machine learning to reveal non-linear relationships between variables which may have clinical value in prognostication of outcomes for high-risk surgical procedures.

Mapping the Relationship of White Matter Lesions to Depression in Multiple Sclerosis.

BACKGROUND: Multiple sclerosis (MS) is an immune-mediated neurological disorder, and up to 50% of patients experience depression. We investigated how white matter network disruption is related to depression in MS. METHODS: Using electronic health records, 380 participants with MS were identified. Depressed individuals (MS+Depression group; n = 232) included persons who had an ICD-10 depression diagnosis, had a prescription for antidepressant medication, or screened positive via Patient Health Questionnaire (PHQ)-2 or PHQ-9. Age- and sex-matched nondepressed individuals with MS (MS-Depression group; n = 148) included persons who had no prior depression diagnosis, had no psychiatric medication prescriptions, and were asymptomatic on PHQ-2 or PHQ-9. Research-quality 3T structural magnetic resonance imaging was obtained as part of routine care. We first evaluated whether lesions were preferentially located within the depression network compared with other brain regions. Next, we examined if MS+Depression patients had greater lesion burden and if this was driven by lesions in the depression network. Primary outcome measures were the burden of lesions (e.g., impacted fascicles) within a network and across the brain. RESULTS: MS lesions preferentially affected fascicles within versus outside the depression network (ß = 0.09, 95% CI = 0.08 to 0.10, p < .001). MS+Depression patients had more lesion burden (ß = 0.06, 95% CI = 0.01 to 0.10, p = .015); this was driven by lesions within the depression network (ß = 0.02, 95% CI = 0.003 to 0.040, p = .020). CONCLUSIONS: We demonstrated that lesion location and burden may contribute to depression comorbidity in MS. MS lesions disproportionately impacted fascicles in the depression network. MS+Depression patients had more disease than MS-Depression patients, which was driven by disease within the depression network. Future studies relating lesion location to personalized depression interventions are warranted.

Quantifying interictal intracranial EEG to predict focal epilepsy.

Introduction: Intracranial EEG (IEEG) is used for 2 main purposes, to determine: (1) if epileptic networks are amenable to focal treatment and (2) where to intervene. Currently these questions are answered qualitatively and sometimes differently across centers. There is a need for objective, standardized methods to guide surgical decision making and to enable large scale data analysis across centers and prospective clinical trials. Methods: We analyzed interictal data from 101 patients with drug resistant epilepsy who underwent presurgical evaluation with IEEG. We chose interictal data because of its potential to reduce the morbidity and cost associated with ictal recording. 65 patients had unifocal seizure onset on IEEG, and 36 were non-focal or multi-focal. We quantified the spatial dispersion of implanted electrodes and interictal IEEG abnormalities for each patient. We compared these measures against the "5 Sense Score (5SS)," a pre-implant estimate of the likelihood of focal seizure onset, and assessed their ability to predict the clinicians' choice of therapeutic intervention and the patient outcome. Results: The spatial dispersion of IEEG electrodes predicted network focality with precision similar to the 5SS (AUC = 0.67), indicating that electrode placement accurately reflected pre-implant information. A cross-validated model combining the 5SS and the spatial dispersion of interictal IEEG abnormalities significantly improved this prediction (AUC = 0.79; p<0.05). The combined model predicted ultimate treatment strategy (surgery vs. device) with an AUC of 0.81 and post-surgical outcome at 2 years with an AUC of 0.70. The 5SS, interictal IEEG, and electrode placement were not correlated and provided complementary information. Conclusions: Quantitative, interictal IEEG significantly improved upon pre-implant estimates of network focality and predicted treatment with precision approaching that of clinical experts. We present this study as an important step in building standardized, quantitative tools to guide epilepsy surgery.

Association of brain tissue cerebrospinal fluid fraction with age in healthy cognitively normal adults.

Background and purpose: Our objective was to apply multi-compartment T2 relaxometry in cognitively normal individuals aged 20-80 years to study the effect of aging on the parenchymal CSF fraction (CSFF), a potential measure of the subvoxel CSF space. Materials and methods: A total of 60 volunteers (age range, 22-80 years) were enrolled. Voxel-wise maps of short-T2 myelin water fraction (MWF), intermediate-T2 intra/extra-cellular water fraction (IEWF), and long-T2 CSFF were obtained using fast acquisition with spiral trajectory and adiabatic T2prep (FAST-T2) sequence and three-pool non-linear least squares fitting. Multiple linear regression analyses were performed to study the association between age and regional MWF, IEWF, and CSFF measurements, adjusting for sex and region of interest (ROI) volume. ROIs include the cerebral white matter (WM), cerebral cortex, and subcortical deep gray matter (GM). In each model, a quadratic term for age was tested using an ANOVA test. A Spearman's correlation between the normalized lateral ventricle volume, a measure of organ-level CSF space, and the regional CSFF, a measure of tissue-level CSF space, was computed. Results: Regression analyses showed that there was a statistically significant quadratic relationship with age for CSFF in the cortex (p = 0.018), MWF in the cerebral WM (p = 0.033), deep GM (p = 0.017) and cortex (p = 0.029); and IEWF in the deep GM (p = 0.033). There was a statistically highly significant positive linear relationship between age and regional CSFF in the cerebral WM (p < 0.001) and deep GM (p < 0.001). In addition, there was a statistically significant negative linear association between IEWF and age in the cerebral WM (p = 0.017) and cortex (p < 0.001). In the univariate correlation analysis, the normalized lateral ventricle volume correlated with the regional CSFF measurement in the cerebral WM (ρ = 0.64, p < 0.001), cortex (ρ = 0.62, p < 0.001), and deep GM (ρ = 0.66, p < 0.001). Conclusion: Our cross-sectional data demonstrate that brain tissue water in different compartments shows complex age-dependent patterns. Parenchymal CSFF, a measure of subvoxel CSF-like water in the brain tissue, is quadratically associated with age in the cerebral cortex and linearly associated with age in the cerebral deep GM and WM.

Mapping the relationship of white matter lesions to depression in multiple sclerosis.

Importance: Multiple sclerosis (MS) is an immune-mediated neurological disorder that affects nearly one million people in the United States. Up to 50% of patients with MS experience depression. Objective: To investigate how white matter network disruption is related to depression in MS. Design: Retrospective case-control study of participants who received research-quality 3-tesla neuroimaging as part of MS clinical care from 2010-2018. Analyses were performed from May 1 to September 30, 2022. Setting: Single-center academic medical specialty MS clinic. Participants: Participants with MS were identified via the electronic health record (EHR). All participants were diagnosed by an MS specialist and completed research-quality MRI at 3T. After excluding participants with poor image quality, 783 were included. Inclusion in the depression group (MS+Depression) required either: 1) ICD-10 depression diagnosis (F32-F34.*); 2) prescription of antidepressant medication; or 3) screening positive via Patient Health Questionnaire-2 (PHQ-2) or -9 (PHQ-9). Age- and sex-matched nondepressed comparators (MS-Depression) included persons with no depression diagnosis, no psychiatric medications, and were asymptomatic on PHQ-2/9. Exposure: Depression diagnosis. Main Outcomes and Measures: We first evaluated if lesions were preferentially located within the depression network compared to other brain regions. Next, we examined if MS+Depression patients had greater lesion burden, and if this was driven by lesions specifically in the depression network. Outcome measures were the burden of lesions (e.g., impacted fascicles) within a network and across the brain. Secondary measures included between-diagnosis lesion burden, stratified by brain network. Linear mixed-effects models were employed. Results: Three hundred-eighty participants met inclusion criteria, (232 MS+Depression: age[SD]=49[12], %females=86; 148 MS-Depression: age[SD]=47[13], %females=79). MS lesions preferentially affected fascicles within versus outside the depression network (ß=0.09, 95% CI=0.08-0.10, P<0.001). MS+Depression had more white matter lesion burden (ß=0.06, 95% CI=0.01-0.10, P=0.015); this was driven by lesions within the depression network (ß=0.02, 95% CI 0.003-0.040, P=0.020). Conclusions and Relevance: We provide new evidence supporting a relationship between white matter lesions and depression in MS. MS lesions disproportionately impacted fascicles in the depression network. MS+Depression had more disease than MS-Depression, which was driven by disease within the depression network. Future studies relating lesion location to personalized depression interventions are warranted.

Quantitative Water Permeability Mapping of Blood-Brain-Barrier Dysfunction in Aging.

Blood-brain-barrier (BBB) dysfunction is a hallmark of aging and aging-related disorders, including cerebral small vessel disease and Alzheimer's disease. An emerging biomarker of BBB dysfunction is BBB water exchange rate (kW) as measured by diffusion-weighted arterial spin labeling (DW-ASL) MRI. We developed an improved DW-ASL sequence for Quantitative Permeability Mapping and evaluated whole brain and region-specific kW in a cohort of 30 adults without dementia across the age spectrum. In this cross-sectional study, we found higher kW values in the cerebral cortex (mean = 81.51 min-1, SD = 15.54) compared to cerebral white matter (mean = 75.19 min-1, SD = 13.85) (p < 0.0001). We found a similar relationship for cerebral blood flow (CBF), concordant with previously published studies. Multiple linear regression analysis with kW as an outcome showed that age was statistically significant in the cerebral cortex (p = 0.013), cerebral white matter (p = 0.033), hippocampi (p = 0.043), orbitofrontal cortices (p = 0.042), and precunei cortices (p = 0.009), after adjusting for sex and number of vascular risk factors. With CBF as an outcome, age was statistically significant only in the cerebral cortex (p = 0.026) and precunei cortices (p = 0.020). We further found moderate negative correlations between white matter hyperintensity (WMH) kW and WMH volume (r = -0.51, p = 0.02), and normal-appearing white matter (NAWM) and WMH volume (r = -0.44, p = 0.05). This work illuminates the relationship between BBB water exchange and aging and may serve as the basis for BBB-targeted therapies for aging-related brain disorders.

QSMRim-Net: Imbalance-aware learning for identification of chronic active multiple sclerosis lesions on quantitative susceptibility maps.

BACKGROUND AND PURPOSE: Chronic active multiple sclerosis (MS) lesions are characterized by a paramagnetic rim at the edge of the lesion and are associated with increased disability in patients. Quantitative susceptibility mapping (QSM) is an MRI technique that is sensitive to chronic active lesions, termed rim + lesions on the QSM. We present QSMRim-Net, a data imbalance-aware deep neural network that fuses lesion-level radiomic and convolutional image features for automated identification of rim + lesions on QSM. METHODS: QSM and T2-weighted-Fluid-Attenuated Inversion Recovery (T2-FLAIR) MRI of the brain were collected at 3 T for 172 MS patients. Rim + lesions were manually annotated by two human experts, followed by consensus from a third expert, for a total of 177 rim + and 3986 rim negative (rim-) lesions. Our automated rim + detection algorithm, QSMRim-Net, consists of a two-branch feature extraction network and a synthetic minority oversampling network to classify rim + lesions. The first network branch is for image feature extraction from the QSM and T2-FLAIR, and the second network branch is a fully connected network for QSM lesion-level radiomic feature extraction. The oversampling network is designed to increase classification performance with imbalanced data. RESULTS: On a lesion-level, in a five-fold cross validation framework, the proposed QSMRim-Net detected rim + lesions with a partial area under the receiver operating characteristic curve (pROC AUC) of 0.760, where clinically relevant false positive rates of less than 0.1 were considered. The method attained an area under the precision recall curve (PR AUC) of 0.704. QSMRim-Net out-performed other state-of-the-art methods applied to the QSM on both pROC AUC and PR AUC. On a subject-level, comparing the predicted rim + lesion count and the human expert annotated count, QSMRim-Net achieved the lowest mean square error of 0.98 and the highest correlation of 0.89 (95% CI: 0.86, 0.92). CONCLUSION: This study develops a novel automated deep neural network for rim + MS lesion identification using T2-FLAIR and QSM images.

Quantitative susceptibility mapping versus phase imaging to identify multiple sclerosis iron rim lesions with demyelination.

BACKGROUND AND PURPOSE: To compare quantitative susceptibility mapping (QSM) and high-pass-filtered (HPF) phase imaging for (1) identifying chronic active rim lesions with more myelin damage and (2) distinguishing patients with increased clinical disability in multiple sclerosis. METHODS: Eighty patients were scanned with QSM for paramagnetic rim detection and Fast Acquisition with Spiral Trajectory and T2prep for myelin water fraction (MWF). Chronic lesions were classified based on the presence/absence of rim on HPF and QSM images. A lesion-level linear mixed-effects model with MWF as the outcome was used to compare myelin damage among the lesion groups. A multiple patient-level linear regression model was fit to establish the association between Expanded Disease Status Scale (EDSS) and the log of the number of rim lesions. RESULTS: Of 2062 lesions, 188 (9.1%) were HPF rim+/QSM rim+, 203 (9.8%) were HPF rim+/QSM rim-, and the remainder had no rim. In the linear mixed-effects model, HPF rim+/QSM rim+ lesions had significantly lower MWF than both HPF rim+/QSM rim- (p < .001) and HPF rim-/QSM rim- (p < .001) lesions, while the MWF difference between HPF rim+/QSM rim- and HPF rim-/QSM rim- lesions was not statistically significant (p = .130). Holding all other factors constant, the log number of QSM rim+ lesion was associated with EDSS increase (p = .044). The association between the log number of HPF rim+ lesions and EDSS was not statistically significant (p = .206). CONCLUSIONS: QSM identifies paramagnetic rim lesions that on average have more myelin damage and stronger association with clinical disability than those detected by phase imaging.

Global cerebrospinal fluid as a zero-reference regularization for brain quantitative susceptibility mapping.

BACKGROUND AND PURPOSE: The objective ofthis study was to demonstrate a global cerebrospinal fluid (CSF) method for a consistent and automated zero referencing of brain quantitative susceptibility mapping (QSM). METHODS: Whole brain CSF mask was automatically segmented by thresholding the gradient echo transverse relaxation ( R2∗) map, and regularization was employed to enforce uniform susceptibility distribution within the CSF volume in the field-to-susceptibility inversion. This global CSF regularization method was compared with a prior ventricular CSF regularization. Both reconstruction methods were compared in a repeatability study of 12 healthy subjects using t-test on susceptibility measurements, and in patient studies of 17 multiple sclerosis (MS) and 10 Parkinson's disease (PD) patients using Wilcoxon rank-sum test on radiological scores. RESULTS: In scan-rescan experiments, global CSF regularization provided more consistent CSF volume as well as higher repeatability of QSM measurements than ventricular CSF regularization with a smaller bias: -2.7 parts per billion (ppb) versus -0.13 ppb (t-test p<0.05) and a narrower 95% limits of agreement: [-7.25, 6.99] ppb versus [-16.60, 11.19 ppb] (f-test p<0.05). In PD and MS patients, global CSF regularization reduced smoothly varying shadow artifacts and significantly improved the QSM quality score (p<0.001). CONCLUSIONS: The proposed whole brain CSF method for QSM zero referencing improves repeatability and image quality of brain QSM compared to the ventricular CSF method.

Diagnostic accuracy of shuttle CT angiography (CTA) and helical CTA in the diagnosis of vasospasm.

PURPOSE: To evaluate the diagnostic accuracy of computed tomography angiography (CTA) acquired with shuttle technique (CTAs) and helical CTA (CTAh) for vasospasm, using digital subtraction angiography (DSA) obtained within 24 h as reference standard. METHODS: Thirty-six patients with suspected vasospasm in the setting of aneurysmal subarachnoid hemorrhage (ASAH, 30/36) or acute inflammatory/infectious conditions (6/36) who underwent CTAs (17/36) or CTAh (19/36) followed by DSA within 24 h were identified retrospectively. Presence of vasospasm in the proximal cerebral arterial segments was assessed qualitatively and semi-quantitatively on CTA and subsequent DSA. Sensitivity, specificity, and receiver operating characteristic (ROC) curves were calculated. Inter-rater variability was assessed using Cohen's kappa. RESULTS: On CTAs, 35% of patients had low and 65% had high vasospasm burden. On CTAh, 37% had low and 63% had high vasospasm burden. ROC analysis demonstrated an AUC of 0.87 for CTAs (95%CI 0.67-1.00, p = 0.015) and 0.88 for CTAh (0.72-1.00, p = 0.028). Cohen's kappa was 0.843 (95%CI 0.548-1.000). Thresholding with Youden's J index, CTAs had a sensitivity of 85.71% (95%CI 48.69 to 99.27) and specificity of 66.67% (35.42 to 87.94). CTAh had sensitivity of 100% (56.55 to 100.00) and specificity of 78.57% (52.41 to 92.43). CONCLUSION: CTAs and CTAh yielded similar sensitivity, specificity, and AUC values on ROC analysis for the detection of vasospasm using DSA as reference standard. Our findings suggest that CTAs is a promising alternative to CTAh especially in patients requiring serial imaging, given its potential advantages of decreased radiation exposure, contrast dose, and cost-effectiveness.

Brain oxygen extraction fraction mapping in patients with multiple sclerosis.

We aimed to demonstrate the feasibility of whole brain oxygen extraction fraction (OEF) mapping for measuring lesion specific and regional OEF abnormalities in multiple sclerosis (MS) patients. In 22 MS patients and 11 healthy controls (HC), OEF and neural tissue susceptibility (χn) maps were computed from MRI multi-echo gradient echo data. In MS patients, 80 chronic active lesions with hyperintense rim on quantitative susceptibility mapping were identified, and the mean OEF and χn within the rim and core were compared using linear mixed-effect model analysis. The rim showed higher OEF and χn than the core: relative to their adjacent normal appearing white matter, OEF contrast = -6.6 ± 7.0% vs. -9.8 ± 7.8% (p < 0.001) and χn contrast = 33.9 ± 20.3 ppb vs. 25.7 ± 20.5 ppb (p = 0.017). Between MS and HC, OEF and χn were compared using a linear regression model in subject-based regions of interest. In the whole brain, compared to HC, MS had lower OEF, 30.4 ± 3.3% vs. 21.4 ± 4.4% (p < 0.001), and higher χn, -23.7 ± 7.0 ppb vs. -11.3 ± 7.7 ppb (p = 0.018). Our feasibility study suggests that OEF may serve as a useful quantitative marker of tissue oxygen utilization in MS.

Differential frequency in imaging-based outcome measurement: Bias in real-world oncology comparative-effectiveness studies.

BACKGROUND: Comparative-effectiveness studies using real-world data (RWD) can be susceptible to surveillance bias. In solid tumor oncology studies, analyses of endpoints such as progression-free survival (PFS) are based on progression events detected by imaging assessments. This study aimed to evaluate the potential bias introduced by differential imaging assessment frequency when using electronic health record (EHR)-derived data to investigate the comparative effectiveness of cancer therapies. METHODS: Using a nationwide de-identified EHR-derived database, we first analyzed imaging assessment frequency patterns in patients diagnosed with advanced non-small cell lung cancer (aNSCLC). We used those RWD inputs to develop a discrete event simulation model of two treatments where disease progression was the outcome and PFS was the endpoint. Using this model, we induced bias with differential imaging assessment timing and quantified its effect on observed versus true treatment effectiveness. We assessed percent bias in the estimated hazard ratio (HR). RESULTS: The frequency of assessments differed by cancer treatment types. In simulated comparative-effectiveness studies, PFS HRs estimated using real-world imaging assessment frequencies differed from the true HR by less than 10% in all scenarios (range: 0.4% to -9.6%). The greatest risk of biased effect estimates was found comparing treatments with widely different imaging frequencies, most exaggerated in disease settings where time to progression is very short. CONCLUSIONS: This study provided evidence that the frequency of imaging assessments to detect disease progression can differ by treatment type in real-world patients with cancer and may induce some bias in comparative-effectiveness studies in some situations.

ALL-Net: Anatomical information lesion-wise loss function integrated into neural network for multiple sclerosis lesion segmentation.

Accurate detection and segmentation of multiple sclerosis (MS) brain lesions on magnetic resonance images are important for disease diagnosis and treatment. This is a challenging task as lesions vary greatly in size, shape, location, and image contrast. The objective of our study was to develop an algorithm based on deep convolutional neural network integrated with anatomic information and lesion-wise loss function (ALL-Net) for fast and accurate automated segmentation of MS lesions. Distance transformation mapping was used to construct a convolutional module that encoded lesion-specific anatomical information. To overcome the lesion size imbalance during network training and improve the detection of small lesions, a lesion-wise loss function was developed in which individual lesions were modeled as spheres of equal size. On the ISBI-2015 longitudinal MS lesion segmentation challenge dataset (19 subjects in total), ALL-Net achieved an overall score of 93.32 and was amongst the top performing methods. On the larger Cornell MS dataset (176 subjects in total), ALL-Net significantly improved both voxel-wise metrics (Dice improvement of 3.9% to 35.3% with p-values ranging from p < 0.01 to p < 0.0001, and AUC of voxel-wise precision-recall curve improvement of 2.1% to 29.8%) and lesion-wise metrics (lesion-wise F1 score improvement of 12.6% to 29.8% with all p-values p < 0.0001, and AUC of lesion-wise ROC curve improvement of 1.4% to 20.0%) compared to leading publicly available MS lesion segmentation tools.

Functional connectome reorganization relates to post-stroke motor recovery and structural and functional disconnection.

Motor recovery following ischemic stroke is contingent on the ability of surviving brain networks to compensate for damaged tissue. In rodent models, sensory and motor cortical representations have been shown to remap onto intact tissue around the lesion site, but remapping to more distal sites (e.g. in the contralesional hemisphere) has also been observed. Resting state functional connectivity (FC) analysis has been employed to study compensatory network adaptations in humans, but mechanisms and time course of motor recovery are not well understood. Here, we examine longitudinal FC in 23 first-episode ischemic pontine stroke patients and utilize a graph matching approach to identify patterns of functional connectivity reorganization during recovery. We quantified functional reorganization between several intervals ranging from 1 week to 6 months following stroke, and demonstrated that the areas that undergo functional reorganization most frequently are in cerebellar/subcortical networks. Brain regions with more structural and functional connectome disruption due to the stroke also had more remapping over time. Finally, we show that functional reorganization is correlated with the extent of motor recovery in the early to late subacute phases, and furthermore, individuals with greater baseline motor impairment demonstrate more extensive early subacute functional reorganization (from one to two weeks post-stroke) and this reorganization correlates with better motor recovery at 6 months. Taken together, these results suggest that our graph matching approach can quantify recovery-relevant, whole-brain functional connectivity network reorganization after stroke.

Fully automated detection of paramagnetic rims in multiple sclerosis lesions on 3T susceptibility-based MR imaging.

BACKGROUND AND PURPOSE: The presence of a paramagnetic rim around a white matter lesion has recently been shown to be a hallmark of a particular pathological type of multiple sclerosis lesion. Increased prevalence of these paramagnetic rim lesions is associated with a more severe disease course in MS, but manual identification is time-consuming. We present APRL, a method to automatically detect paramagnetic rim lesions on 3T T2*-phase images. METHODS: T1-weighted, T2-FLAIR, and T2*-phase MRI of the brain were collected at 3T for 20 subjects with MS. The images were then processed with automated lesion segmentation, lesion center detection, lesion labelling, and lesion-level radiomic feature extraction. A total of 951 lesions were identified, 113 (12%) of which contained a paramagnetic rim. We divided our data into a training set (16 patients, 753 lesions) and a testing set (4 patients, 198 lesions), fit a random forest classification model on the training set, and assessed our ability to classify paramagnetic rim lesions on the test set. RESULTS: The number of paramagnetic rim lesions per subject identified via our automated lesion labelling method was highly correlated with the gold standard count per subject, r = 0.86 (95% CI [0.68, 0.94]). The classification algorithm using radiomic features classified lesions with an area under the curve of 0.82 (95% CI [0.74, 0.92]). CONCLUSION: This study develops a fully automated technique, APRL, for the detection of paramagnetic rim lesions using standard T1 and FLAIR sequences and a T2*phase sequence obtained on 3T MR images.

Predictors of acute deep venous thrombosis in patients hospitalized for COVID-19.

ABSTRACT: Deep venous thrombosis (DVT) is associated with high mortality in coronavirus disease 2019 (COVID-19) but there remains uncertainty about the benefit of anti-coagulation prophylaxis and how to decide when ultrasound screening is indicated. We aimed to determine parameters predicting which COVID-19 patients are at risk of DVT and to assess the benefit of prophylactic anti-coagulation.Adult hospitalized patients with positive severe acute respiratory syndrome coronavirus-2 reverse transcription-polymerase chain reaction (RT-PCR) undergoing venous duplex ultrasound for DVT assessment (n = 451) were retrospectively reviewed. Clinical and laboratory data within 72 hours of ultrasound were collected. Using split sampling and a 10-fold cross-validation, a random forest model was developed to find the most important variables for predicting DVT. Different d-dimer cutoffs were examined for classification of DVT. We also compared the rate of DVT between the patients going and not going under thromboprophylaxis.DVT was found in 65 (14%) of 451 reverse transcription-polymerase chain reaction positive patients. The random forest model, trained and cross-validated on 2/3 of the original sample (n = 301), had area under the receiver operating characteristic curve = 0.91 (95% confidence interval [CI]: 0.85-0.97) for prediction of DVT in the test set (n = 150), with sensitivity = 93% (95%CI: 68%-99%) and specificity = 82% (95%CI: 75%-88%). The following variables had the highest importance: d-dimer, thromboprophylaxis, systolic blood pressure, admission to ultrasound interval, and platelets. Thromboprophylaxis reduced DVT risk 4-fold from 26% to 6% (P < .001), while anti-coagulation therapy led to hemorrhagic complications in 14 (22%) of 65 patients with DVT including 2 fatal intra-cranial hemorrhages. D-dimer was the most important predictor with area under curve = 0.79 (95%CI: 0.73-0.86) by itself, and a 5000 ng/mL threshold at 7 days postCOVID-19 symptom onset had 75% (95%CI: 53%-90%) sensitivity and 81% (95%CI: 72%-88%) specificity. In comparison with d-dimer alone, the random forest model showed 68% versus 32% specificity at 95% sensitivity, and 44% versus 23% sensitivity at 95% specificity.D-dimer >5000 ng/mL predicts DVT with high accuracy suggesting regular monitoring with d-dimer in the early stages of COVID-19 may be useful. A random forest model improved the prediction of DVT. Thromboprophylaxis reduced DVT in COVID-19 patients and should be considered in all patients. Full anti-coagulation therapy has a risk of life-threatening hemorrhage.

Heuristic scoring method utilizing FDG-PET statistical parametric mapping in the evaluation of suspected Alzheimer disease and frontotemporal lobar degeneration.

Distinguishing frontotemporal lobar degeneration (FTLD) and Alzheimer Disease (AD) on FDG-PET based on qualitative review alone can pose a diagnostic challenge. SPM has been shown to improve diagnostic performance in research settings, but translation to clinical practice has been lacking. Our purpose was to create a heuristic scoring method based on statistical parametric mapping z-scores. We aimed to compare the performance of the scoring method to the initial qualitative read and a machine learning (ML)-based method as benchmarks. FDG-PET/CT or PET/MRI of 65 patients with suspected dementia were processed using SPM software, yielding z-scores from either whole brain (W) or cerebellar (C) normalization relative to a healthy cohort. A non-ML, heuristic scoring system was applied using region counts below a preset z-score cutoff. W z-scores, C z-scores, or WC z-scores (z-scores from both W and C normalization) served as features to build random forest models. The neurological diagnosis was used as the gold standard. The sensitivity of the non-ML scoring system and the random forest models to detect AD was higher than the initial qualitative read of the standard FDG-PET [0.89-1.00 vs. 0.22 (95% CI, 0-0.33)]. A categorical random forest model to distinguish AD, FTLD, and normal cases had similar accuracy than the non-ML scoring model (0.63 vs. 0.61). Our non-ML-based scoring system of SPM z-scores approximated the diagnostic performance of a ML-based method and demonstrated higher sensitivity in the detection of AD compared to qualitative reads. This approach may improve the diagnostic performance.

Estimation of Multiple Sclerosis lesion age on magnetic resonance imaging.

We introduce the first-ever statistical framework for estimating the age of Multiple Sclerosis (MS) lesions from magnetic resonance imaging (MRI). Estimating lesion age is an important step when studying the longitudinal behavior of MS lesions and can be used in applications such as studying the temporal dynamics of chronic active MS lesions. Our lesion age estimation models use first order radiomic features over a lesion derived from conventional T1 (T1w) and T2 weighted (T2w) and fluid attenuated inversion recovery (FLAIR), T1w with gadolinium contrast (T1w+c), and Quantitative Susceptibility Mapping (QSM) MRI sequences as well as demographic information. For this analysis, we have a total of 32 patients with 53 new lesions observed at 244 time points. A one or two step random forest model for lesion age is fit on a training set using a lesion volume cutoff of 15 mm3 or 50 mm3. We explore the performance of nine different modeling scenarios that included various combinations of the MRI sequences and demographic information and a one or two step random forest models, as well as simpler models that only uses the mean radiomic feature from each MRI sequence. The best performing model on a validation set is a model that uses a two-step random forest model on the radiomic features from all of the MRI sequences with demographic information using a lesion volume cutoff of 50 mm3. This model has a mean absolute error of 7.23 months (95% CI: [6.98, 13.43]) and a median absolute error of 5.98 months (95% CI: [5.26, 13.25]) in the validation set. For this model, the predicted age and actual age have a statistically significant association (p-value <0.001) in the validation set.

PItcHPERFeCT: Primary Intracranial Hemorrhage Probability Estimation using Random Forests on CT.

INTRODUCTION: Intracerebral hemorrhage (ICH), where a blood vessel ruptures into areas of the brain, accounts for approximately 10-15% of all strokes. X-ray computed tomography (CT) scanning is largely used to assess the location and volume of these hemorrhages. Manual segmentation of the CT scan using planimetry by an expert reader is the gold standard for volume estimation, but is time-consuming and has within- and across-reader variability. We propose a fully automated segmentation approach using a random forest algorithm with features extracted from X-ray computed tomography (CT) scans. METHODS: The Minimally Invasive Surgery plus rt-PA in ICH Evacuation (MISTIE) trial was a multi-site Phase II clinical trial that tested the safety of hemorrhage removal using recombinant-tissue plasminogen activator (rt-PA). For this analysis, we use 112 baseline CT scans from patients enrolled in the MISTE trial, one CT scan per patient. ICH was manually segmented on these CT scans by expert readers. We derived a set of imaging predictors from each scan. Using 10 randomly-selected scans, we used a first-pass voxel selection procedure based on quantiles of a set of predictors and then built 4 models estimating the voxel-level probability of ICH. The models used were: 1) logistic regression, 2) logistic regression with a penalty on the model parameters using LASSO, 3) a generalized additive model (GAM) and 4) a random forest classifier. The remaining 102 scans were used for model validation.For each validation scan, the model predicted the probability of ICH at each voxel. These voxel-level probabilities were then thresholded to produce binary segmentations of the hemorrhage. These masks were compared to the manual segmentations using the Dice Similarity Index (DSI) and the correlation of hemorrhage volume of between the two segmentations. We tested equality of median DSI using the Kruskal-Wallis test across the 4 models. We tested equality of the median DSI from sets of 2 models using a Wilcoxon signed-rank test. RESULTS: All results presented are for the 102 scans in the validation set. The median DSI for each model was: 0.89 (logistic), 0.885 (LASSO), 0.88 (GAM), and 0.899 (random forest). Using the random forest results in a slightly higher median DSI compared to the other models. After Bonferroni correction, the hypothesis of equality of median DSI was rejected only when comparing the random forest DSI to the DSI from the logistic (p < 0.001), LASSO (p < 0.001), or GAM (p < 0.001) models. In practical terms the difference between the random forest and the logistic regression is quite small. The correlation (95% CI) between the volume from manual segmentation and the predicted volume was 0.93 (0.9,0.95) for the random forest model. These results indicate that random forest approach can achieve accurate segmentation of ICH in a population of patients from a variety of imaging centers. We provide an R package (https://github.com/muschellij2/ichseg) and a Shiny R application online (http://johnmuschelli.com/ich_segment_all.html) for implementing and testing the proposed approach.

PREVAIL: Predicting Recovery through Estimation and Visualization of Active and Incident Lesions.

OBJECTIVE: The goal of this study was to develop a model that integrates imaging and clinical information observed at lesion incidence for predicting the recovery of white matter lesions in multiple sclerosis (MS) patients. METHODS: Demographic, clinical, and magnetic resonance imaging (MRI) data were obtained from 60 subjects with MS as part of a natural history study at the National Institute of Neurological Disorders and Stroke. A total of 401 lesions met the inclusion criteria and were used in the study. Imaging features were extracted from the intensity-normalized T1-weighted (T1w) and T2-weighted sequences as well as magnetization transfer ratio (MTR) sequence acquired at lesion incidence. T1w and MTR signatures were also extracted from images acquired one-year post-incidence. Imaging features were integrated with clinical and demographic data observed at lesion incidence to create statistical prediction models for long-term damage within the lesion. VALIDATION: The performance of the T1w and MTR predictions was assessed in two ways: first, the predictive accuracy was measured quantitatively using leave-one-lesion-out cross-validated (CV) mean-squared predictive error. Then, to assess the prediction performance from the perspective of expert clinicians, three board-certified MS clinicians were asked to individually score how similar the CV model-predicted one-year appearance was to the true one-year appearance for a random sample of 100 lesions. RESULTS: The cross-validated root-mean-square predictive error was 0.95 for normalized T1w and 0.064 for MTR, compared to the estimated measurement errors of 0.48 and 0.078 respectively. The three expert raters agreed that T1w and MTR predictions closely resembled the true one-year follow-up appearance of the lesions in both degree and pattern of recovery within lesions. CONCLUSION: This study demonstrates that by using only information from a single visit at incidence, we can predict how a new lesion will recover using relatively simple statistical techniques. The potential to visualize the likely course of recovery has implications for clinical decision-making, as well as trial enrichment.