Natalie Ullman

Minimally Invasive Surgery Plus Recombinant Tissue-type Plasminogen Activator for Intracerebral Hemorrhage Evacuation Decreases Perihematomal Edema

Background and Purpose— Perihematomal edema (PHE) can worsen outcomes after intracerebral hemorrhage (ICH). Reports suggest that blood degradation products lead to PHE. We hypothesized that hematoma evacuation will reduce PHE volume and that treatment with recombinant tissue-type plasminogen activator (rt-PA) will not exacerbate it. Methods— Minimally invasive surgery and rt-PA in ICH evacuation (MISTIE) phase II tested safety and efficacy of hematoma evacuation after ICH. We conducted a semiautomated, computerized volumetric analysis on computed tomography to assess impact of hematoma removal on PHE and effects of rt-PA on PHE. Volumetric analyses were performed on baseline stability and end of treatment scans. Results— Seventy-nine surgical and 39 medical patients from minimally invasive surgery and rt-PA in ICH evacuation phase II (MISTIE II) were analyzed. Mean hematoma volume at end of treatment was 19.6±14.5 cm3 for the surgical cohort and 40.7±13.9 cm3 for the medical cohort (P<0.001). Edema volume at end of treatment was lower for the surgical cohort: 27.7±13.3 cm3 than medical cohort: 41.7±14.6 cm3 (P<0.001). Graded effect of clot removal on PHE was observed when patients with >65%, 20% to 65%, and <20% ICH removed were analyzed (P<0.001). Positive correlation between PHE reduction and percent of ICH removed was identified (&rgr;=0.658; P<0.001). In the surgical cohort, 69 patients underwent surgical aspiration and rt-PA, whereas 10 underwent surgical aspiration only. Both cohorts achieved similar clot reduction: surgical aspiration and rt-PA, 18.9±14.5 cm3; and surgical aspiration only, 24.5±14.0 cm3 (P=0.26). Edema at end of treatment in surgical aspiration and rt-PA was 28.1±13.8 cm3 and 24.4±8.6 cm3 in surgical aspiration only (P=0.41). Conclusions— Hematoma evacuation is associated with significant reduction in PHE. Furthermore, PHE does not seem to be exacerbated by rt-PA, making such neurotoxic effects unlikely when the drug is delivered to intracranial clot.

Thrombolytic removal of intraventricular haemorrhage in treatment of severe stroke: results of the randomised, multicentre, multiregion, placebo-controlled CLEAR III trial

BACKGROUND Intraventricular haemorrhage is a subtype of intracerebral haemorrhage, with 50% mortality and serious disability for survivors. We aimed to test whether attempting to remove intraventricular haemorrhage with alteplase versus saline irrigation improved functional outcome. METHODS In this randomised, double-blinded, placebo-controlled, multiregional trial (CLEAR III), participants with a routinely placed extraventricular drain, in the intensive care unit with stable, non-traumatic intracerebral haemorrhage volume less than 30 mL, intraventricular haemorrhage obstructing the 3rd or 4th ventricles, and no underlying pathology were adaptively randomly assigned (1:1), via a web-based system to receive up to 12 doses, 8 h apart of 1 mg of alteplase or 0·9% saline via the extraventricular drain. The treating physician, clinical research staff, and participants were masked to treatment assignment. CT scans were obtained every 24 h throughout dosing. The primary efficacy outcome was good functional outcome, defined as a modified Rankin Scale score (mRS) of 3 or less at 180 days per central adjudication by blinded evaluators. This study is registered with ClinicalTrials.gov, NCT00784134. FINDINGS Between Sept 18, 2009, and Jan 13, 2015, 500 patients were randomised: 249 to the alteplase group and 251 to the saline group. 180-day follow-up data were available for analysis from 246 of 249 participants in the alteplase group and 245 of 251 participants in the placebo group. The primary efficacy outcome was similar in each group (good outcome in alteplase group 48% vs saline 45%; risk ratio [RR] 1·06 [95% CI 0·88-1·28; p=0·554]). A difference of 3·5% (RR 1·08 [95% CI 0·90-1·29], p=0·420) was found after adjustment for intraventricular haemorrhage size and thalamic intracerebral haemorrhage. At 180 days, the treatment group had lower case fatality (46 [18%] vs saline 73 [29%], hazard ratio 0·60 [95% CI 0·41-0·86], p=0·006), but a greater proportion with mRS 5 (42 [17%] vs 21 [9%]; RR 1·99 [95% CI 1·22-3·26], p=0·007). Ventriculitis (17 [7%] alteplase vs 31 [12%] saline; RR 0·55 [95% CI 0·31-0·97], p=0·048) and serious adverse events (114 [46%] alteplase vs 151 [60%] saline; RR 0·76 [95% CI 0·64-0·90], p=0·002) were less frequent with alteplase treatment. Symptomatic bleeding (six [2%] in the alteplase group vs five [2%] in the saline group; RR 1·21 [95% CI 0·37-3·91], p=0·771) was similar. INTERPRETATION In patients with intraventricular haemorrhage and a routine extraventricular drain, irrigation with alteplase did not substantially improve functional outcomes at the mRS 3 cutoff compared with irrigation with saline. Protocol-based use of alteplase with extraventricular drain seems safe. Future investigation is needed to determine whether a greater frequency of complete intraventricular haemorrhage removal via alteplase produces gains in functional status. FUNDING National Institute of Neurological Disorders and Stroke.

Safety and efficacy of minimally invasive surgery plus alteplase in intracerebral haemorrhage evacuation (MISTIE): a randomised, controlled, open-label, phase 2 trial

BACKGROUND Craniotomy, according to the results from trials, does not improve functional outcome after intracerebral haemorrhage. Whether minimally invasive catheter evacuation followed by thrombolysis for clot removal is safe and can achieve a good functional outcome is not known. We investigated the safety and efficacy of alteplase, a recombinant tissue plasminogen activator, in combination with minimally invasive surgery (MIS) in patients with intracerebral haemorrhage. METHODS MISTIE was an open-label, phase 2 trial that was done in 26 hospitals in the USA, Canada, the UK, and Germany. We used a computer-generated allocation sequence with a block size of four to centrally randomise patients aged 18-80 years with a non-traumatic (spontaneous) intracerebral haemorrhage of 20 mL or higher to standard medical care or image-guided MIS plus alteplase (0·3 mg or 1·0 mg every 8 h for up to nine doses) to remove clots using surgical aspiration followed by alteplase clot irrigation. Primary outcomes were all safety outcomes: 30 day mortality, 7 day procedure-related mortality, 72 h symptomatic bleeding, and 30 day brain infections. This trial is registered with ClinicalTrials.gov, number NCT00224770. FINDINGS Between Feb 2, 2006, and April 8, 2013, 96 patients were randomly allocated and completed follow-up: 54 (56%) in the MIS plus alteplase group and 42 (44%) in the standard medical care group. The primary outcomes did not differ between the standard medical care and MIS plus alteplase groups: 30 day mortality (four [9·5%, 95% CI 2·7-22.6] vs eight [14·8%, 6·6-27·1], p=0·542), 7 day mortality (zero [0%, 0-8·4] vs one [1·9%, 0·1-9·9], p=0·562), symptomatic bleeding (one [2·4%, 0·1-12·6] vs five [9·3%, 3·1-20·3], p=0·226), and brain bacterial infections (one [2·4%, 0·1-12·6] vs zero [0%, 0-6·6], p=0·438). Asymptomatic haemorrhages were more common in the MIS plus alteplase group than in the standard medical care group (12 [22·2%; 95% CI 12·0-35·6] vs three [7·1%; 1·5-19·5]; p=0·051). INTERPRETATION MIS plus alteplase seems to be safe in patients with intracerebral haemorrhage, but increased asymptomatic bleeding is a major cautionary finding. These results, if replicable, could lead to the addition of surgical management as a therapeutic strategy for intracerebral haemorrhage. FUNDING National Institute of Neurological Disorders and Stroke, Genentech, and Codman.SUMMARY Background Craniotomy, when evaluated in trials, does not improve outcome after intracerebral haemorrhage (ICH). Whether minimally invasive catheter evacuation followed by thrombolysis is safe and can achieve a good functional outcome by removing clot is unknown. We investigated safety and efficacy of alteplase with minimally invasive surgery (MIS) in patients with intracerebral haemorrhage. Methods MISTIE was an international, randomized, open-label study and was done in 26 hospitals in the USA, Canada, the UK, and Germany. Patients (aged 18–80 years), with non-traumatic (spontaneous) ICH ≥20 mL were randomly allocated, centrally, to medical care or image-guided MIS plus rt-PA (0.3 mg or 1.0 mg every 8 hours for up to 9 doses) to remove clot using surgical aspiration followed with alteplase clot irrigation. The primary efficacy outcome was the adjusted dichotomized modified Rankin Scale (mRS) 0–3 vs 4–6 assessed at day 180 after symptom onset. Analysis was by intention to treat. (ClinicalTrials.gov number NCT00224770). Findings Between February 2, 2006 and April 8, 2013, 96 subjects were randomized and completed follow-up: 54 received treatment and 42 medical care. Primary safety outcomes: mortality, symptomatic bleeding, brain infections, as well as withdrawal of care, did not differ between groups. Asymptomatic hemorrhages were more common in the surgical group (3 (7%) vs. 12 (22%) p= 0.05) producing a difference of 15.1% (95% CI: 1.5% to 28.6%). The estimated absolute benefit, i.e., the unadjusted difference in observed proportions of all subjects with mRS 0–3 (33% vs 21%) at 180 days comparing MISPA vs. medical control, is 0.109 [95%CI: −0.088, 0.294; p=0.26], and is 0.162 [95%CI: 0.003, 0.323; p=0.05] after adjustment for potential imbalances in baseline severity between study arms (primary efficacy outcome). Interpretation MIS+rt-PA appears safe with an apparent advantage of better functional outcome at 180 days. Increased asymptomatic bleeding is a major cautionary finding. The MISTIE trial results, if replicable, could produce a meaningful functional benefit adding surgical management as a therapeutic strategy for ICH. Funding National Institute of Neurologic Disorders and Stroke, Genentech, and Codman.

Resolution of Intraventricular Hemorrhage Varies by Ventricular Region and Dose of Intraventricular Thrombolytic The Clot Lysis: Evaluating Accelerated Resolution of IVH (CLEAR IVH) Program

Background and Purpose— The Clot Lysis: Evaluating Accelerated Resolution of IVH (CLEAR IVH) program is assessing the efficacy of intraventricular recombinant tissue-type plasminogen activator (rtPA) for spontaneous intraventricular hemorrhage (IVH). This subanalysis assesses the effect of dose of rtPA by region on clearance of IVH. Methods— Sixty-four patients within 12 to 24 hours of spontaneous IVH were randomized to placebo or 0.3 mg, 1 mg, or 3 mg of rtPA twice daily through an extraventricular drain. Twelve subregions of the ventricles were scored from 0 to 4. Effect of dose on IVH clearance to 50% of baseline score was compared by survival analysis for all regions combined and by subregion. Models including ventricular region, dose, and baseline score were compared by Cox proportional hazards. Results— IVH score reduced faster across all regions with increasing rtPA dose (clearance to 50%: log-rank P<0.0001; placebo—11.43 days, 95% CI, 5.68–17.18; 0.3 mg—3.19 days, 1.00–5.38; 1 mg—3.54 days, 0.45–6.64; 3 mg—2.59 days, 1.72–3.46). In the combined models, dose and baseline score were independently associated with reduction in IVH score, which was quickest in the midline ventricles, then the anterior half of the lateral ventricles and slowest in the posterior half of the lateral ventricles (clearance to 50%: P<0.0001; rtPA dose: hazard ratio, 1.47, 1.30–1.67; midline versus anterolateral hazard ratio, 1.71, 1.08–2.71; midline versus posterolateral hazard ratio, 4.05, 2.46–6.65; baseline score hazard ratio, 0.96, 0.91–1.01) with a significant interaction between dose and ventricular region (P=0.005). Conclusions— rtPA accelerates resolution of IVH. This effect is dose-dependent, is greatest in the midline ventricles, and least in the posterolateral ventricles. Clinical Trial Registration— URL: www.clinicaltrials.gov. Unique identifier: NCT00650858.

Accuracy of the ABC/2 Score for Intracerebral Hemorrhage: Systematic Review and Analysis of MISTIE, CLEAR-IVH, and CLEAR III

Background and Purpose— The ABC/2 score estimates intracerebral hemorrhage (ICH) volume, yet validations have been limited by small samples and inappropriate outcome measures. We determined accuracy of the ABC/2 score calculated at a specialized reading center (RC-ABC) or local site (site-ABC) versus the reference-standard computed tomography–based planimetry (CTP). Methods— In Minimally Invasive Surgery Plus Recombinant Tissue-Type Plasminogen Activator for Intracerebral Hemorrhage Evacuation-II (MISTIE-II), Clot Lysis Evaluation of Accelerated Resolution of Intraventricular Hemorrhage (CLEAR-IVH) and CLEAR-III trials. ICH volume was prospectively calculated by CTP, RC-ABC, and site-ABC. Agreement between CTP and ABC/2 was defined as an absolute difference up to 5 mL and relative difference within 20%. Determinants of ABC/2 accuracy were assessed by logistic regression. Results— In 4369 scans from 507 patients, CTP was more strongly correlated with RC-ABC (r2=0.93) than with site-ABC (r2=0.87). Although RC-ABC overestimated CTP-based volume on average (RC-ABC, 15.2 cm3; CTP, 12.7 cm3), agreement was reasonable when categorized into mild, moderate, and severe ICH (&kgr;=0.75; P<0.001). This was consistent with overestimation of ICH volume in 6 of 8 previous studies. Agreement with CTP was greater for RC-ABC (84% within 5 mL; 48% of scans within 20%) than for site-ABC (81% within 5 mL; 41% within 20%). RC-ABC had moderate accuracy for detecting ≥5 mL change in CTP volume between consecutive scans (sensitivity, 0.76; specificity, 0.86) and was more accurate with smaller ICH, thalamic hemorrhage, and homogeneous clots. Conclusions— ABC/2 scores at local or central sites are sufficiently accurate to categorize ICH volume and assess eligibility for the CLEAR-III and MISTIE III studies and moderately accurate for change in ICH volume. However, accuracy decreases with large, irregular, or lobar clots. Clinical Trial Registration— URL: http://www.clinicaltrials.gov. Unique identifier: MISTIE-II NCT00224770; CLEAR-III NCT00784134.

Validated automatic brain extraction of head CT images.

BACKGROUND X-ray computed tomography (CT) imaging of the brain is commonly used in diagnostic settings. Although CT scans are primarily used in clinical practice, they are increasingly used in research. A fundamental processing step in brain imaging research is brain extraction - the process of separating the brain tissue from all other tissues. Methods for brain extraction have either been 1) validated but not fully automated, or 2) fully automated and informally proposed, but never formally validated. AIM To systematically analyze and validate the performance of FSLs brain extraction tool (BET) on head CT images of patients with intracranial hemorrhage. This was done by comparing the manual gold standard with the results of several versions of automatic brain extraction and by estimating the reliability of automated segmentation of longitudinal scans. The effects of the choice of BET parameters and data smoothing is studied and reported. METHODS All images were thresholded using a 0-100 Hounsfield unit (HU) range. In one variant of the pipeline, data were smoothed using a 3-dimensional Gaussian kernel (σ=1mm(3)) and re-thresholded to 0-100HU; in the other, data were not smoothed. BET was applied using 1 of 3 fractional intensity (FI) thresholds: 0.01, 0.1, or 0.35 and any holes in the brain mask were filled. For validation against a manual segmentation, 36 images from patients with intracranial hemorrhage were selected from 19 different centers from the MISTIE (Minimally Invasive Surgery plus recombinant-tissue plasminogen activator for Intracerebral Evacuation) stroke trial. Intracranial masks of the brain were manually created by one expert CT reader. The resulting brain tissue masks were quantitatively compared to the manual segmentations using sensitivity, specificity, accuracy, and the Dice Similarity Index (DSI). Brain extraction performance across smoothing and FI thresholds was compared using the Wilcoxon signed-rank test. The intracranial volume (ICV) of each scan was estimated by multiplying the number of voxels in the brain mask by the dimensions of each voxel for that scan. From this, we calculated the ICV ratio comparing manual and automated segmentation: ICVautomated/ICVmanual. To estimate the performance in a large number of scans, brain masks were generated from the 6 BET pipelines for 1095 longitudinal scans from 129 patients. Failure rates were estimated from visual inspection. ICV of each scan was estimated and an intraclass correlation (ICC) was estimated using a one-way ANOVA. RESULTS Smoothing images improves brain extraction results using BET for all measures except specificity (all p<0.01, uncorrected), irrespective of the FI threshold. Using an FI of 0.01 or 0.1 performed better than 0.35. Thus, all reported results refer only to smoothed data using an FI of 0.01 or 0.1. Using an FI of 0.01 had a higher median sensitivity (0.9901) than an FI of 0.1 (0.9884, median difference: 0.0014, p<0.001), accuracy (0.9971 vs. 0.9971; median difference: 0.0001, p<0.001), and DSI (0.9895 vs. 0.9894; median difference: 0.0004, p<0.001) and lower specificity (0.9981 vs. 0.9982; median difference: -0.0001, p<0.001). These measures are all very high indicating that a range of FI values may produce visually indistinguishable brain extractions. Using smoothed data and an FI of 0.01, the mean (SD) ICV ratio was 1.002 (0.008); the mean being close to 1 indicates the ICV estimates are similar for automated and manual segmentation. In the 1095 longitudinal scans, this pipeline had a low failure rate (5.2%) and the ICC estimate was high (0.929, 95% CI: 0.91, 0.945) for successfully extracted brains. CONCLUSION BET performs well at brain extraction on thresholded, 1mm(3) smoothed CT images with an FI of 0.01 or 0.1. Smoothing before applying BET is an important step not previously discussed in the literature. Analysis code is provided.

Characterization of Intraventricular and Intracerebral Hematomas in Non-Contrast CT

Characterization of hematomas is essential in scan reading, manual delineation, and designing automatic segmentation algorithms. Our purpose is to characterize the distribution of intraventricular (IVH) and intracerebral hematomas (ICH) in NCCT scans, study their relationship to gray matter (GM), and to introduce a new tool for quantitative hematoma delineation. We used 289 serial retrospective scans of 51 patients. Hematomas were manually delineated in a two-stage process. Hematoma contours generated in the first stage were quantified and enhanced in the second stage. Delineation was based on new quantitative rules and hematoma profiling, and assisted by a dedicated tool superimposing quantitative information on scans with 3D hematoma display. The tool provides: density maps (40–85HU), contrast maps (8/15HU), mean horizontal/vertical contrasts for hematoma contours, and hematoma contours below a specified mean contrast (8HU). White matter (WM) and GM were segmented automatically. IVH/ICH on serial NCCT is characterized by 59.0HU mean, 60.0HU median, 11.6HU standard deviation, 23.9HU mean contrast, 0.99HV/day slope, and 0.24 skewness (changing over time from negative to positive). Its 0.1st −99.9th percentile range corresponds to 25–88HU range. WM and GM are highly correlated (R2=0.88; p<10−10) whereas the GM-GS correlation is weak (R2=0.14; p<10−10). The intersection point of mean GM-hematoma density distributions is at 55.6±5.8HU with the corresponding GM/hematoma percentiles of 88th/40th. Objective characterization of IVH/ICH and stating the rules quantitatively will aid raters to delineate hematomas more robustly and facilitate designing algorithms for automatic hematoma segmentation. Our two-stage process is general and potentially applicable to delineate other pathologies on various modalities more robustly and quantitatively.The Pipeline embolization device (PED) is one of the flow-diverting stents approved for the treatment of unruptured large or wide-necked cerebral aneurysms in 2011(1). Its use has now been extended to the treatment of recently ruptured dissecting cerebral aneurysm, carotid pseudoaneurysm from radiation injury, and blister aneurysms(2,3). We aimed to evaluate the effectiveness of utilizing the PED as a primary treatment for ruptured dissecting intracranial aneurysms. A single center retrospective review was conducted for all patients primarily treated with PED for acute subarachnoid hemorrhage (SAH) from ruptured dissecting cerebral aneurysms between December 2010 and February 2013. Patients were followed up with CT angiogram (CTA) or digital subtraction angiogram (DSA). Eight patients with a total of eight dissecting aneurysms were identified. The mean duration from SAH to treatment was 2.5 days. Six of the aneurysms arose from vertebral arteries and two from the basilar artery. Immediate check-DSA confirmed satisfactory contrast stasis in all eight cases, and complete aneurysmal obliteration was achieved at six months. There were two (25%) procedure-related complications, but no major procedure-related complications, such as thromboembolic events or rebleeding from aneurysm were encountered. The PED is a feasible treatment option for ruptured dissecting cerebral aneurysms in acute phase. According to our experience, using PED as flow-diverters in acute SAH does not significantly increase the complication risks or mortality rate if the antiplatelet regime is carefully monitored. Future studies shall evaluate the optimal antiplatelet regimen for using the PED in the acute phase.

Applicability of Clinical Trials in an Unselected Cohort of Patients With Intracerebral Hemorrhage

Background and Purpose— Patient selection in clinical trials on intracerebral hemorrhage (ICH) affects overall applicability of results. We estimated eligibility for completed, ongoing, and planned clinical trials in an unselected cohort of patients with ICH. Methods— Large clinical ICH trials were identified using trial registration databases. Each trial’s inclusion criteria were applied to a consecutive group of patients with ICH from the prospective hospital-based Lund Stroke Register. Survival status was obtained from the National Census Office and 90-day poor functional outcome (modified Rankin Scale ≥4) from the Swedish Stroke Register or medical files. Results— Among 253 patients with ICH, estimated eligibility proportions ranged between 2% and 36% for the 11 identified clinical trials. Patients not eligible for any trial (n=96) had more intraventricular hemorrhage, lower baseline level of consciousness, higher rates of cerebellar ICH, and lower rates of lobar ICH (P⩽0.001). Thirty-day case fatality for noneligible patients was 54% versus 18% among patients eligible in ≥1 trial (95% confidence interval, 44%–64% versus 13%–25%; P<0.001). Noneligible ICH patients more frequently had poor functional outcome (75% versus 48%; 95% confidence interval, 65%–83% versus 40%–56%; P<0.001). Conclusions— There is large variation in proportions of patients with ICH eligible for inclusion in clinical trials and over a third of patients with ICH are not eligible for any trial.

Early Quantification of Hematoma Hounsfield Units on Noncontrast CT in Acute Intraventricular Hemorrhage Predicts Ventricular Clearance after Intraventricular Thrombolysis.

BACKGROUND AND PURPOSE: Thrombolytic efficacy of intraventricular rtPA for acute intraventricular hemorrhage may depend on hematoma composition. We assessed whether hematoma Hounsfield unit quantification informs intraventricular hemorrhage clearance after intraventricular rtPA. MATERIALS AND METHODS: Serial NCCT was performed on 52 patients who received intraventricular rtPA as part of the Clot Lysis Evaluation of Accelerated Resolution of Intraventricular Hemorrhage trial and 12 controls with intraventricular hemorrhage, but no rtPA treatment. A blinded investigator calculated Hounsfield unit values for intraventricular hemorrhage volumes on admission (t0), days 3–4 (t1), and days 6–9 (t2). Controls were matched uniquely to 12 rtPA-treated patients for comparison. RESULTS: Median intraventricular hemorrhage volume on admission for patients treated with intraventricular rtPA was 31.9 mL (interquartile range, 34.1 mL), and it decreased to 4.9 mL (interquartile range, 14.5 mL) (t2). Mean (±standard error of the mean) Hounsfield unit for intraventricular hemorrhage was 52.1 (0.59) at t0 and decreased significantly to 50.1 (0.63) (t1), and to 45.1 (0.71) (t2). Total intraventricular hemorrhage Hounsfield unit count was significantly correlated with intraventricular hemorrhage volume at all time points (t0: P = .002; t1: P < .001; t2: P < .001). On serologic and CSF analysis at t0, only higher CSF protein was positively correlated with intraventricular hemorrhage Hounsfield units (P = .03). In 24 matched patients treated with rtPA and controls, total intraventricular hemorrhage Hounsfield units were significantly lower in patients treated with rtPA at t2 (P = .02). Higher Hounsfield unit quantification of fourth ventricle hematomas independently predicted slower clearance of this ventricle (95% CI, 0.02–0.14; P = .02), along with higher intraventricular hemorrhage volume (95% CI, 0.02–0.41; P = .03) and lower CSF protein levels (95% CI, −0.003 to −0.002; P < .001). CONCLUSIONS: Intraventricular hemorrhage Hounsfield unit counts decrease significantly in the acute phase and to a greater extent with intraventricular rtPA treatment. Intraventricular hemorrhage Hounsfield units are correlated significantly with CSF protein and not with serum erythrocyte or platelet concentrations. Hounsfield unit counts may reflect intraventricular hemorrhage clot composition and rtPA sensitivity.

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.