Koskimäki, Janne

Features of new bleeding on conventional imaging in cerebral cavernous malformations (CCMs) often disappear after several weeks, yet the risk of rebleeding persists long thereafter. Increases in mean lesional quantitative susceptibility mapping (QSM) ≥ 6% on MRI during 1 year of prospective surveillance have been associated with new symptomatic hemorrhage (SH) during that period. The authors hypothesized that QSM at a single time point reflects features of hemorrhage in the prior year or potential bleeding in the subsequent year. Twenty-eight features were extracted from 265 QSM acquisitions in 120 patients enrolled in a prospective trial readiness project, and machine learning methods examined associations with SH and biomarker bleed (QSM increase ≥ 6%) in prior and subsequent years. QSM features including sum variance, variance, and correlation had lower average values in lesions with SH in the prior year (p < 0.05, false discovery rate corrected). A support-vector machine classifier recurrently selected sum average, mean lesional QSM, sphericity, and margin sharpness features to distinguish biomarker bleeds in the prior year (area under the curve = 0.61, 95% CI 0.52–0.70; p = 0.02). No QSM features were associated with a subsequent bleed. These results provide proof of concept that machine learning may derive features of QSM reflecting prior hemorrhagic activity, meriting further investigation. Clinical trial registration no.: NCT03652181 ( ClinicalTrials.gov )

Introduction and Hypothesis: New bleeding in cerebral cavernous malformations (CCM) heralds increased risk of future hemorrhage for several years, yet conventional imaging only detects new bleeding that occurred in the prior weeks. A biomarker of hemorrhage could help identifying high risk lesions. An increase of mean lesional quantitative susceptibility mapping (QSM) ≥6% on MRI has been adjudicated as reflecting new bleeding in CCM during longitudinal follow-up. However, mean lesional QSM from a single acquisition could not diagnose or prognosticate a bleed. We hypothesize that machine learning (ML) may identify diagnostic and prognostic features of bleeding within QSM maps at a single point in time. Material and Methods: Two hundred and sixty-five QSM maps of CCM lesions were acquired in 120 patients enrolled in National Institute of Health (NIH) multisite trial readiness project (NCT03652181). Each map was classified (Yes/No) in association with symptomatic hemorrhage (SH) and/or biomarker event with QSM increase ≥6% in the prior (diagnostic association) and subsequent (prognostic association) year. Twenty-eight features were extracted including 14 texture, 5 first-order statistical, as well as 3 size, shape, and morphological. Five-fold cross-validation was conducted on a support-vector machine (SVM) with linear stepwise kernel for both diagnostic and prognostic associations. Performance of individual features and composite classifiers was evaluated using student t-test ( p <0.05) with Bonferroni-correction and receiver operating characteristic (ROC) analysis, area under the curve (AUC). Results: Lesions with SH in the prior year had lower average values for sum variance (AUC=0.79, p<0.001), variance (AUC=0.79, p<0.001), and correlation (AUC=0.71, p=0.004) as compared to lesions without SH. The SVM classification method tasked with distinguishing lesions with mean QSM increase ≥6% in the prior year included recurrent selection of sum average, mean, sphericity and margin sharpness, yielding an AUC of 0.61 (p=0.02). In the prognostic cohort, no individual feature nor any SVM classification model was able to distinguish cases with a subsequent clinical bleed or biomarker event. Conclusion: This proof-of-concept suggests that ML may assist in deriving features on QSM maps acquired at a single timepoint, which reflect prior hemorrhagic activity in CCM. Further investigation is planned in larger cohorts and in conjunction with clinical trials of bleeding in CCM.

BackgroundQuantitative susceptibility mapping (QSM) and dynamic contrast‐enhanced quantitative permeability (DCEQP) on magnetic resonance (MR) have been shown to correlate with neurovascular disease progression as markers of vascular leakage and hemosiderin deposition. Applying these techniques as monitoring biomarkers in clinical trials will be necessary; however, their validation across multiple MR platforms and institutions has not been rigorously verified.PurposeTo validate quantitative measurement of MR biomarkers on multiple instruments at different institutions.Study TypePhantom validation between platforms and institutions.Phantom ModelT1/susceptibility phantom, two‐compartment dynamic flow phantom.Field Strength/Sequence3T/QSM, T1 mapping, dynamic 2D SPGR.AssessmentPhilips Ingenia, Siemens Prisma, and Siemens Skyra at three different institutions were assessed. A QSM phantom with concentrations of gadolinium, corresponding to magnetic susceptibilities of 0, 0.1, 0.2, 0.4, and 0.8 ppm was assayed. DCEQP was assessed by measuring a MultiHance bolus as the consistency of the width ratio of the curves at the input and outputs over a range of flow ratios between outputs.Statistical TestsEach biomarker was assessed by measures of accuracy (Pearson correlation), precision (paired t‐test between repeated measurements), and reproducibility (analysis of covariance [ANCOVA] between instruments).ResultsQSM accuracy of r2 > 0.997 on all three platforms was measured. Precision (P = 0.66 Achieva, P = 0.76 Prisma, P = 0.69 Skyra) and reproducibility (P = 0.89) were good. T1 mapping of accuracy was r2 > 0.98. No significant difference between width ratio regression slopes at site 2 (P = 0.669) or site 3 (P = 0.305), and no significant difference between width ratio regression slopes between sites was detected by ANCOVA (P = 0.48).Data ConclusionThe phantom performed as expected and determined that MR measures of QSM and DCEQP are accurate and consistent across repeated measurements and between platforms.Level of Evidence: 1Technical Efficacy Stage: 2J. Magn. Reson. Imaging 2020;51:1192–1199.

Background: Cerebral cavernous malformation is a genetic neurovascular disorder causing aberrant angiogenesis and lesion development within capillary beds. Lesions exhibit impaired endothelial barrier function, causing blood-brain barrier leakage and hemorrhage. Quantitative MRI biomarkers (dynamic contrast-enhanced quantitative permeability (DCEQP) and quantitative susceptibility mapping (QSM)) have been shown to correlate with CCM disease progression as measures of barrier leakage and hemorrhagic iron deposition, and there is interest in utilizing these biomarkers as endpoints in clinical trials. However, their consistencies across multiple platforms and institutions has not yet been rigorously verified. Materials and Methods: Instruments used were a 3T Philips Ingenia at University of Chicago and a 3T Siemens Prisma at Northwestern University. QSM was assessed by five balloons filled with varying concentrations of MultiHance, corresponding to magnetic susceptibilities of 0, 0.1, 0.2, 0.4, and 0.8, surrounded by an agarose gel. Permeability was assessed by the injection of a bolus of MultiHance through a coiled hose inside a water-filled shell, and the consistency of the peak-to-peak ratio of the curves at the input and output was measured. Each biomarker was assessed by accuracy (to expected values, Pearson correlation), precision (between repeated measurements, paired t-test), and reproducibility (between instruments, paired t-test). Results: QSM measured with an accuracy of r 2 = 0.997 on the Philips Achieva and r 2 = 0.999 on the Siemens Prisma. Each scanner had adequate precision between measurements, showing no significant difference by a paired t-test (p = 0.66 Achieva, p = 0.45 Prisma), and the reproducibility between scanners also showed no significant difference (p = 0.89). The peak-to-peak ratio on DCEQP scans was accurate (r 2 = 0.998), precise between measurements (p = 0.54), and repeatable between instruments (p = 0.49). Conclusion: Quantitative imaging biomarkers DCEQP and QSM are accurate, consistent across multiple measurements, and between platforms and will be useful for future clinical trials in CCM.

Abstract BACKGROUND More than a million Americans harbor a cerebral cavernous angioma (CA), and those who suffer a prior symptomatic hemorrhage have an exceptionally high rebleeding risk. Preclinical studies show that atorvastatin blunts CA lesion development and hemorrhage through inhibiting RhoA kinase (ROCK), suggesting it may confer a therapeutic benefit. OBJECTIVE To evaluate whether atorvastatin produces a difference compared to placebo in lesional iron deposition as assessed by quantitative susceptibility mapping (QSM) on magnetic resonance imaging in CAs that have demonstrated a symptomatic hemorrhage in the prior year. Secondary aims shall assess effects on vascular permeability, ROCK activity in peripheral leukocytes, signal effects on clinical outcomes, adverse events, and prespecified subgroups. METHODS The phase I/IIa placebo-controlled, double-blinded, single-site clinical trial aims to enroll 80 subjects randomized 1-1 to atorvastatin (starting dose 80 mg PO daily) or placebo. Dosing shall continue for 24-mo or until reaching a safety endpoint. EXPECTED OUTCOMES The trial is powered to detect an absolute difference of 20% in the mean percent change in lesional QSM per year (2-tailed, power 0.9, alpha 0.05). A decrease in QSM change would be a signal of potential benefit, and an increase would signal a safety concern with the drug. DISCUSSION With firm mechanistic rationale, rigorous preclinical discoveries, and biomarker validations, the trial shall explore a proof of concept effect of a widely used repurposed drug in stabilizing CAs after a symptomatic hemorrhage. This will be the first clinical trial of a drug aimed at altering rebleeding in CA.

Abstract BACKGROUND Brain cavernous angiomas with symptomatic hemorrhage (CASH) are uncommon but exact a heavy burden of neurological disability from recurrent bleeding, for which there is no proven therapy. Candidate drugs to stabilize the CASH lesion and prevent rebleeding will ultimately require testing of safety and efficacy in multisite clinical trials. Much progress has been made in understanding the epidemiology of CASH, and novel biomarkers have been linked to the biological mechanisms and clinical activity in lesions. Yet, the ability to enroll and risk-stratify CASH subjects has never been assessed prospectively at multiple sites. Biomarkers and other outcomes have not been evaluated for their sensitivity and reliability, nor have they been harmonized across sites. OBJECTIVE To address knowledge gaps and establish a research network as infrastructure for future clinical trials, through the Trial Readiness grant mechanism, funded by National Institute of Neurological Disorders and Stroke/National Institutes of Health. METHODS This project includes an observational cohort study to assess (1) the feasibility of screening, enrollment rates, baseline disease categorization, and follow-up of CASH using common data elements at multiple sites, (2) the reliability of imaging biomarkers including quantitative susceptibility mapping and permeability measures that have been shown to correlate with lesion activity, and (3) the rates of recurrent hemorrhage and change in functional status and biomarker measurements during prospective follow-up. EXPECTED OUTCOMES We propose a harmonized multisite assessment of enrollment rates of CASH, baseline features relevant to stratification in clinical trials, and follow-up assessments of functional outcomes in relation to clinical bleeds. We introduce novel biomarkers of vascular leak and hemorrhage, with firm mechanistic foundations, which have been linked to clinical disease activity. We shall test their reliability and validity at multiple sites, and assess their changes over time, with and without clinical rebleeds, hence their fitness as outcome instruments in clinical trials. DISCUSSION The timing cannot be more opportune, with therapeutic targets identified, exceptional collaboration among researchers and the patient community, along with several drugs ready to benefit from development of a path to clinical testing using this network in the next 5 years.