Automatic segmentation of the thalamus using a massively trained 3D convolutional neural network: higher sensitivity for the detection of reduced thalamus volume by improved inter-scanner stability.

OBJECTIVES: To develop an automatic method for accurate and robust thalamus segmentation in T1w-MRI for widespread clinical use without the need for strict harmonization of acquisition protocols and/or scanner-specific normal databases. METHODS: A three-dimensional convolutional neural network (3D-CNN) was trained on 1975 T1w volumes from 170 MRI scanners using thalamus masks generated with FSL-FIRST as ground truth. Accuracy was evaluated with 18 manually labeled expert masks. Intra- and inter-scanner test-retest stability were assessed with 477 T1w volumes of a single healthy subject scanned on 123 MRI scanners. The sensitivity of 3D-CNN-based volume estimates for the detection of thalamus atrophy was tested with 127 multiple sclerosis (MS) patients and a normal database comprising 4872 T1w volumes from 160 scanners. The 3D-CNN was compared with a publicly available 2D-CNN (FastSurfer) and FSL. RESULTS: The Dice similarity coefficient of the automatic thalamus segmentation with manual expert delineation was similar for all tested methods (3D-CNN and FastSurfer 0.86 ± 0.02, FSL 0.87 ± 0.02). The standard deviation of the single healthy subject's thalamus volume estimates was lowest with 3D-CNN for repeat scans on the same MRI scanner (0.08 mL, FastSurfer 0.09 mL, FSL 0.15 mL) and for repeat scans on different scanners (0.28 mL, FastSurfer 0.62 mL, FSL 0.63 mL). The proportion of MS patients with significantly reduced thalamus volume was highest for 3D-CNN (24%, FastSurfer 16%, FSL 11%). CONCLUSION: The novel 3D-CNN allows accurate thalamus segmentation, similar to state-of-the-art methods, with considerably improved robustness with respect to scanner-related variability of image characteristics. This might result in higher sensitivity for the detection of disease-related thalamus atrophy. KEY POINTS: • A three-dimensional convolutional neural network was trained for automatic segmentation of the thalamus with a heterogeneous sample of T1w-MRI from 1975 patients scanned on 170 different scanners. • The network provided high accuracy for thalamus segmentation with manual segmentation by experts as ground truth. • Inter-scanner variability of thalamus volume estimates across different MRI scanners was reduced by more than 50%, resulting in increased sensitivity for the detection of thalamus atrophy.

Infratentorial lesions in multiple sclerosis patients: intra- and inter-rater variability in comparison to a fully automated segmentation using 3D convolutional neural networks.

OBJECTIVE: Automated quantification of infratentorial multiple sclerosis lesions on magnetic resonance imaging is clinically relevant but challenging. To overcome some of these problems, we propose a fully automated lesion segmentation algorithm using 3D convolutional neural networks (CNNs). METHODS: The CNN was trained on a FLAIR image alone or on FLAIR and T1-weighted images from 1809 patients acquired on 156 different scanners. An additional training using an extra class for infratentorial lesions was implemented. Three experienced raters manually annotated three datasets from 123 MS patients from different scanners. RESULTS: The inter-rater sensitivity (SEN) was 80% for supratentorial lesions but only 62% for infratentorial lesions. There was no statistically significant difference between the inter-rater SEN and the SEN of the CNN with respect to the raters. For supratentorial lesions, the CNN featured an intra-rater intra-scanner SEN of 0.97 (R1 = 0.90, R2 = 0.84) and for infratentorial lesion a SEN of 0.93 (R1 = 0.61, R2 = 0.73). CONCLUSION: The performance of the CNN improved significantly for infratentorial lesions when specifically trained on infratentorial lesions using a T1 image as an additional input and matches the detection performance of experienced raters. Furthermore, for infratentorial lesions the CNN was more robust against repeated scans than experienced raters. KEY POINTS: • A 3D convolutional neural network was trained on MRI data from 1809 patients (156 different scanners) for the quantification of supratentorial and infratentorial multiple sclerosis lesions. • Inter-rater variability was higher for infratentorial lesions than for supratentorial lesions. The performance of the 3D convolutional neural network (CNN) improved significantly for infratentorial lesions when specifically trained on infratentorial lesions using a T1 image as an additional input. • The detection performance of the CNN matches the detection performance of experienced raters.

Apparent Diffusion Coefficient Signal Intensity Ratio Predicts the Effect of Revascularization on Ischemic Cerebral Edema.

BACKGROUND: Apparent diffusion coefficient (ADC) imaging is a biomarker of cytotoxic injury that predicts edema formation and outcome after ischemic stroke. It therefore has the potential to serve as a "tissue clock" to describe the extent of ischemic injury and potentially predict response to therapy. The goal of this study was to determine the relationship between baseline ADC signal intensity, revascularization, and edema formation. METHODS: We examined the ADC signal intensity ratio (ADCr) of the stroke lesion (defined as the baseline DWI hyperintense region) compared to the contralateral normal hemisphere in 65 subjects from the Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy trial. The associations between ADCr, neurologic outcome, and cerebral edema were examined. Finally, we explored the interaction between baseline ADCr and vessel recanalization at day 7 on post-stroke edema. RESULTS: We found that lower initial ADCr was associated with a worse outcome on the modified Rankin Scale (mRS) at 90 days (52.2% of those with ADCr <64% were mRS 5-6 vs. 19.1% with ADCr ≥64%, p = 0.006). Those subjects with reconstitution of flow distal to the initial vessel occlusion showed greater normalization of ADCr on follow-up scan (increase in ADCr of 16.4 ± 2.07 vs. 1.99 ± 4.33%, p = 0.0039). In those patients with low baseline ADCr, successful revascularization was associated with reduced edema (median swelling volume 164 mL [interquartile range (IQR) 53.3-190 mL] vs. 20.7 mL [IQR 3.20-55.1 mL], p = 0.024). CONCLUSIONS: This study reaffirms the association of ADCr with outcome after stroke, supports the idea that reperfusion may attenuate rather than enhance post-stroke edema, and indicates that the degree of edema with and without revascularization may be predicted by ADCr.

Relationship Between Changes in the Temporal Dynamics of the Blood-Oxygen-Level-Dependent Signal and Hypoperfusion in Acute Ischemic Stroke.

BACKGROUND AND PURPOSE: Changes in the blood-oxygen-level-dependent (BOLD) signal provide a noninvasive measure of blood flow, but a detailed comparison with established perfusion parameters in acute stroke is lacking. We investigated the relationship between BOLD signal temporal delay and dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) in stroke patients. METHODS: In 30 patients with acute (<24 hours) ischemic stroke, we performed Pearson correlation and multiple linear regression between DSC-MRI parameters (time to maximum [Tmax], mean transit time, cerebral blood flow, and cerebral blood volume) and BOLD-based parameters (BOLD delay and coefficient of BOLD variation). Prediction of severe hypoperfusion (Tmax >6 seconds) was assessed using receiver-operator characteristic (ROC) analysis. RESULTS: Correlation was highest between Tmax and BOLD delay (venous sinus reference; time shift range 7; median r=0.60; interquartile range=0.49-0.71). Coefficient of BOLD variation correlated with cerebral blood volume (median r= 0.37; interquartile range=0.24-0.51). Mean R2 for predicting BOLD delay by DSC-MRI was 0.54 (SD=0.2) and for predicting coefficient of BOLD variation was 0.37 (SD=0.17). BOLD delay (whole-brain reference, time shift range 3) had an area under the curve of 0.76 for predicting severe hypoperfusion (sensitivity=69.2%; specificity=80%), whereas BOLD delay (venous sinus reference, time shift range 3) had an area under the curve of 0.76 (sensitivity=67.3%; specificity=83.5%). CONCLUSIONS: BOLD delay is related to macrovascular delay and microvascular hypoperfusion, can identify severely hypoperfused tissue in acute stroke, and is a promising alternative to gadolinium contrast agent-based perfusion assessment in acute stroke. CLINICAL TRIAL REGISTRATION: URL: http://www.clinicaltrials.gov. Unique identifier: NCT00715533 and NCT02077582.

Subtracted Dynamic MR Perfusion Source Images (sMRP-SI) provide Collateral Blood Flow Assessment in MCA Occlusions and Predict Tissue Fate.

OBJECTIVES: Collateral blood flow is accepted as a predictive factor of tissue fate in ischemic stroke. Thus, we aimed to evaluate a new method derived from MR perfusion source images to assess collateral flow in patients with ICA/MCA occlusions. METHODS: A total of 132 patients of the prospective 1000+ study were examined. MR perfusion source images were assessed according to Δimg_n = img_n + 1 - img_n - 1 using the five-grade Higashida collateral flow rating system. Higashida scores were correlated to mismatch (MM) volume, mismatch ratio, day 6 FLAIR lesion volumes and day 90 mRS. RESULTS: Patients with Higashida scores 3 and 4 had significantly lower admission NIHSS, smaller FLAIR day 6 lesion volumes (p < 0.001) and higher rates of better long-term outcome (mRS 0-2, p = 0.002). There was a linear trend for the association of Higashida grade 1 (p = 0.002) and 2 (p = 0.001) with unfavourable outcome (day 90 mRS 3-6), but no significant association was found for MM volume, MM ratio and day 90 mRS. Inter-rater agreement was 0.58 (95% CI 0.43-0.73) on day 1, 0.70 (95% CI 0.58-0.81) on day 2. CONCLUSION: sMRP-SI Higashida score offers a non-invasive collateral vessel and tissue perfusion assessment of ischemic tissue. The predictive value of Higashida rating proved superior to MM with regard to day 90 mRS. KEY POINTS: • Assessment of collateral flow using subtracted dynamic MR perfusion source imaging (sMRP-SI). • sMRP-SI offers additional information about morphological characteristics of ischemic brain tissue. • sMRP-SI collateral flow assessment proves superior to mismatch volume. • Better collateral flow was significantly associated with better outcome (day 90 mRS).

Hyperintense acute reperfusion marker is associated with higher contrast agent dosage in acute ischaemic stroke.

OBJECTIVE: The hyperintense acute reperfusion marker (HARM) on fluid-attenuated inversion recovery (FLAIR) images is associated with blood-brain barrier (BBB) permeability changes. The aim of this study was to examine the influence of contrast agent dosage on HARM incidence in acute ischaemic stroke patients. METHODS: We prospectively included 529 acute ischaemic stroke patients (204 females, median age 71 years). Patients underwent a first stroke-MRI within 24 hours from symptom onset and had a follow-up on day 2. The contrast agent Gadobutrol was administered to the patients for perfusion imaging or MR angiography. The total dosage was calculated as ml/kg body weight and ranged between 0.04 and 0.31 mmol/kg on the first examination. The incidence of HARM was evaluated on day 2 FLAIR images. RESULTS: HARM was detected in 97 patients (18.3%). HARM incidence increased significantly with increasing dosages of Gadobutrol. Also, HARM positive patients were significantly older. HARM was not an independent predictor of worse clinical outcome, and we did not find an association with increase risk of haemorrhagic transformation. CONCLUSIONS: A higher dosage of Gadobutrol in acute stroke patients on initial MRI is associated with increased HARM incidence on follow-up. MRI studies on BBB should therefore standardize contrast agent dosages. KEY POINTS: • Hyperintense acute reperfusion marker on MRI indicates blood-brain barrier disruption. • This observational study on stroke patients characterizes HARM. • Incidence depends on contrast agent dosage on the previous day. • HARM is also associated with older age and poor kidney function. • Interpretation of HARM must take dosage into consideration.

Case report of a young stroke patient showing interim normalization of the MRI diffusion-weighted imaging lesion.

BACKGROUND: In acute ischemic stroke, diffusion weighted imaging (DWI) shows hyperintensities and is considered to indicate irreversibly damaged tissue. We present the case of a young stroke patient with unusual variability in the development of signal intensities within the same vessel territory. CASE PRESENTATION: A 35-year-old patient presented with symptoms of global aphasia and hypesthesia of the left hand. MRI demonstrated a scattered lesion in the MCA territory. After rtPA therapy the patient received further MRI examination, three times on day 1, and once on day 2, 3, 5 and 43. The posterior part of the lesion showed the usual pattern with increasing DWI hyperintensity and decreased ADC, as well as delayed FLAIR positivity. However, the anterior part of the lesion, which was clearly visible in the first examination completely normalized on the first day and only reappeared on day 2. This was accompanied by a normalization of the ADC as well as an even further delayed FLAIR positivity. CONCLUSION: We showed that interim normalization of DWI and ADC in the acute phase can not only be found in rodent models of stroke, but also in humans. We propose that DWI lesion development might be more variable during the first 24 h after stroke than previously assumed.

Early time course of FLAIR signal intensity differs between acute ischemic stroke patients with and without hyperintense acute reperfusion marker.

BACKGROUND: In animal models of stroke, the time course of blood-brain barrier (BBB) disruptions has been elaborately studied. In human patients, leakage of gadolinium into cerebrospinal fluid (CSF) space, visualized on MRI fluid attenuated inversion recovery (FLAIR) images, is considered a sign of BBB disruptions. It was termed 'hyperintense acute reperfusion marker' (HARM) and was associated with hemorrhages. However, the time course of the leakage is unknown and difficult to study in human patients. Also, the association of HARM with signal intensities and enhancement in the parenchyma on FLAIR images has not been thoroughly researched. METHODS: We analyzed imaging data of acute ischemic stroke patients who underwent repetitive MRI examinations within the first 36 h after the time of symptom onset. HARM was evaluated on FLAIR images. Regions of interest (ROI) of the hyperintensities on diffusion-weighted imaging (DWI) were determined for each time point and mirrored to the contralateral side. The ROI were furthermore corrected for CSF-filled space, using apparent diffusion coefficient (ADC) images. The corrected ROI were used to determine mean signal intensities of the lesions relative to the contralateral side on FLAIR, ADC and B0 images for each time point. RESULTS: The 18 included patients (5 females; median age: 69 years; median NIHSS score: 5) received 3-5 MRI examinations on the first day and 1-2 examinations on day 2 after stroke. Eight of the patients (44.4%) showed HARM on at least 1 examination. In 6 of these patients, HARM was already seen at the second examination, at the earliest 3.5 h after symptom onset. The HARM-positive patients had higher relative signal intensities (rSI) on FLAIR images in the parenchyma corresponding to the DWI-positive tissue compared with the HARM-negative patients. This difference between groups was statistically significant for the 2nd and 3rd examination (medians of 4.31 and 6.37 h from symptom onset, p < 0.001 and p = 0.005, respectively). No significant difference in rSI between groups was seen for ADC or B0 images. CONCLUSION: HARM does not only represent a contrast medium leakage from the pial system into the CSF space. It is accompanied by a markedly increased rSI in the early ischemic lesion on FLAIR images, which is likely due to parenchymal enhancement. The lack of differences on B0 images excludes a pure T2 effect.

The potential of microvessel density in prediction of infarct growth: a two-month experimental study in vessel size imaging.

OBJECTIVES: Vessel size imaging is a novel technique to evaluate pathological changes of the microvessel density quantity Q and the mean vessel size index (VSI). As a follow-up study, we assessed these parameters for microscopic description of ischemic penumbra and their potentials in predicting lesion growth. METHODS: Seventy-five patients with a perfusion-diffusion mismatch were examined within 24 h from symptom onset. We defined three regions of interest: the initial infarct (INF), the ischemic penumbra (IPE), and the healthy region (HEA) symmetric to the IPE. For 23 patients with a 6th-day follow-up, IPE regions were divided into areas of infarct growth and areas of oligemia. RESULT: The median values of Q and VSI were: for INF 0.29 s(-1/3) and 15.8 µm, for IPE 0.33 s(-1/3) and 20.6 µm and for HEA 0.36 s(-1/3) and 17.4 µm. The Q in the IPE was significantly smaller than in HEA, and VSI was significantly larger. The Q with a threshold of 0.32 s(-1/3) predicted the final infarction with a sensitivity of 69% and a specificity of 64%. CONCLUSIONS: The reduced Q and increased VSI in the IPE confirmed our previous pilot results. Although Q showed a trend to identify the severity of ischemia in an overall voxel population, its potential in predicting infarct growth needs to be further tested in a larger cohort including a clear status of reperfusion and recanalization.

Automated vs manual delineations of regions of interest- a comparison in commercially available perfusion MRI software.

BACKGROUND: In perfusion magnetic resonance imaging a manual approach to delineation of regions of interest is, due to rater bias and time intensive operator input, clinically less favorable than an automated approach would be. The goal of our study was to compare the performances of these approaches. METHODS: Using Stroketool, PMA and Perfscape/Neuroscape perfusion maps of cerebral blood flow, mean transit time and Tmax were created for 145 patients with acute ischemic stroke. Volumes of hypoperfused tissue were calculated using both a manual and an automated protocol, and the results compared between methods. RESULTS: The median difference between the automatically and manually derived volumes was up to 210 ml in Perfscape/Neuroscape, 123 ml in PMA and 135 ml in Stroketool. Correlation coefficients between perfusion volumes and radiological and clinical outcome were much lower for the automatic volumes than for the manually derived ones. CONCLUSIONS: The agreement of the two methods was very poor, with the automated use producing falsely exaggerated volumes of hypoperfused tissue. Software improvements are necessary to enable highly automated protocols to credibly assess perfusion deficits.

Fully automated postprocessing carries a risk of substantial overestimation of perfusion deficits in acute stroke magnetic resonance imaging.

BACKGROUND AND PURPOSE: Due to the risk of rater bias and time restrictions in clinical practice, an automated approach to delineation of hypoperfused tissue in patients with acute ischemic stroke would be preferred to a manual one. We tested the hypothesis that existing software solutions, on account of numerous artifacts, produce hypoperfused tissue even in a cohort of patients with no ischemia. METHODS: Thirty-nine patients, all admitted for exclusion of cerebral ischemia or hemorrhage and without a final diagnosis of stroke imaged between September 2008 and May 2009 were included in the study. Using 3 different software packages (PerfScape/NeuroScape, PMA and Stroketool), perfusion maps of mean transit time, cerebral blood flow and T(max) were created for each patient. Three different thresholds were applied to each parameter map, and subsequent volumes of hypoperfused tissue were calculated. RESULTS: The median volume of hypoperfused tissue for all the subjects was 92.9 ml (interquartile range, IQR: 13.3-323.4 ml) when calculated by PerfScape/NeuroScape, 30.42 ml (IQR: 13.9-71.4 ml) when calculated by PMA and 78.71 ml (IQR: 40.3-140.8 ml) when calculated by Stroketool. The volumes derived via the different software applications mostly showed only a weak-to-moderate association with each other (Spearman's correlation coefficient between 0.02 and 0.76). CONCLUSIONS: Although automated protocols show promise, the programs Stroketool, PerfScape and PMA require substantial improvement in order to be able to automatically and reliably differentiate between patients with a credible region of ischemia-related hypoperfusion and those without.

Search for a map and threshold in perfusion MRI to accurately predict tissue fate: a protocol for assessing lesion growth in patients with persistent vessel occlusion.

BACKGROUND: The MRI-based mismatch concept has been used to estimate the risk of infarction in ischemic stroke. Based on multiple studies on magnetic resonance perfusion imaging, it seems unlikely that any perfusion parameter threshold will provide a reliable prediction of radiological or clinical outcome for all patients. The goal of our study was to find a minimally biased yet maximally useful perfusion postprocessing protocol which would offer the treating physician a useful estimate of tissue fate. METHODS: One hundred and forty-five acute ischemic stroke patients, admitted within 24 h after stroke to the Charité-University Medicine Hospital in Berlin between March 2008 and November 2009, were included in this study. Using three different software packages (Perfscape/Neuroscape, PMA and Stroketool), maps of mean transit time, cerebral blood flow (CBF) and T(max) were created. Three different thresholds were applied on each parameter map and subsequent volumes of hypoperfused tissue were calculated. RESULTS: Overall, the maps and thresholds giving the least amount of overestimation of the final infarct volume were T(max) 8 s in Perfscape/Neuroscape, CBF 20 ml/100 g/min in PMA and CBF 15% (of the highest value on the scale for a given patient) in Stroketool. In patients with persistent vessel occlusion, a CBF map with a restrictive threshold showed volumes of tissue at definite risk of infarction in up to 100% of patients. The additional use of a CBF map with a high threshold enabled identification of patients without penumbras. CONCLUSIONS: No combination of software, map and threshold was able to give a reliable estimate of tissue fate for either all patients or any subgroup of patients. However, in patients with vessel occlusion, combination of a CBF map with a low and a high threshold can enable calculation of the minimum volume of brain tissue that will inevitably be lost if the occlusion persists.

Clinical and radiological courses do not differ between fluid-attenuated inversion recovery-positive and negative patients with stroke after thrombolysis.

BACKGROUND AND PURPOSE: After acute ischemic stroke, the proportion of patients with detectable lesions on fluid-attenuated inversion recovery (FLAIR) MRI sequences increases over time. We investigated whether thrombolysis was less effective in FLAIR-positive versus -negative patients. METHODS: In this single-center hospital-based study, all consecutive patients with ischemic stroke who underwent an MRI before and 24 hours after thrombolysis between May 2008 and October 2009 were included. Patients were included if exact time of onset was known and thrombolysis was performed within 3 hours up until August 2008 and within 4.5 hours from September 2008 on. Blinded to time of symptom onset, 3 raters independently judged the visibility of lesions on FLAIR. Lesion volumes on diffusion-weighted imaging as well as National Institutes of Health Stroke Scale before and 1 day after thrombolysis were determined. RESULTS: Of 51 patients (25 females, mean age 71, median National Institutes of Health Stroke Scale 6), 26 were FLAIR-positive. Neither lesion growth nor change in National Institutes of Health Stroke Scale differed significantly between FLAIR-positive versus -negative patients: median growth 2.6 mL (interquartile growth, -0.1 to 17.6) versus 0.8 mL (interquartile range, 0.1 to 17.8) and change in National Institutes of Health Stroke Scale -2.5 (interquartile range, -5 to 0) versus -2.0 (interquartile range, -5 to 0.5), respectively (P>0.5, Mann-Whitney rank sum test). CONCLUSIONS: Visibility of lesions on FLAIR in areas of diffusion restriction was not predictive of the response to thrombolysis.

Multiple sclerosis lesion activity segmentation with attention-guided two-path CNNs.

Multiple sclerosis is an inflammatory autoimmune demyelinating disease that is characterized by lesions in the central nervous system. Typically, magnetic resonance imaging (MRI) is used for tracking disease progression. Automatic image processing methods can be used to segment lesions and derive quantitative lesion parameters. So far, methods have focused on lesion segmentation for individual MRI scans. However, for monitoring disease progression, lesion activity in terms of new and enlarging lesions between two time points is a crucial biomarker. For this problem, several classic methods have been proposed, e.g., using difference volumes. Despite their success for single-volume lesion segmentation, deep learning approaches are still rare for lesion activity segmentation. In this work, convolutional neural networks (CNNs) are studied for lesion activity segmentation from two time points. For this task, CNNs are designed and evaluated that combine the information from two points in different ways. In particular, two-path architectures with attention-guided interactions are proposed that enable effective information exchange between the two time point's processing paths. It is demonstrated that deep learning-based methods outperform classic approaches and it is shown that attention-guided interactions significantly improve performance. Furthermore, the attention modules produce plausible attention maps that have a masking effect that suppresses old, irrelevant lesions. A lesion-wise false positive rate of 26.4% is achieved at a true positive rate of 74.2%, which is not significantly different from the interrater performance.

Fully automated longitudinal segmentation of new or enlarged multiple sclerosis lesions using 3D convolutional neural networks.

The quantification of new or enlarged lesions from follow-up MRI scans is an important surrogate of clinical disease activity in patients with multiple sclerosis (MS). Not only is manual segmentation time consuming, but inter-rater variability is high. Currently, only a few fully automated methods are available. We address this gap in the field by employing a 3D convolutional neural network (CNN) with encoder-decoder architecture for fully automatic longitudinal lesion segmentation. Input data consist of two fluid attenuated inversion recovery (FLAIR) images (baseline and follow-up) per patient. Each image is entered into the encoder and the feature maps are concatenated and then fed into the decoder. The output is a 3D mask indicating new or enlarged lesions (compared to the baseline scan). The proposed method was trained on 1809 single point and 1444 longitudinal patient data sets and then validated on 185 independent longitudinal data sets from two different scanners. From the two validation data sets, manual segmentations were available from three experienced raters, respectively. The performance of the proposed method was compared to the open source Lesion Segmentation Toolbox (LST), which is a current state-of-art longitudinal lesion segmentation method. The mean lesion-wise inter-rater sensitivity was 62%, while the mean inter-rater number of false positive (FP) findings was 0.41 lesions per case. The two validated algorithms showed a mean sensitivity of 60% (CNN), 46% (LST) and a mean FP of 0.48 (CNN), 1.86 (LST) per case. Sensitivity and number of FP were not significantly different (p < 0.05) between the CNN and manual raters. New or enlarged lesions counted by the CNN algorithm appeared to be comparable with manual expert ratings. The proposed algorithm seems to outperform currently available approaches, particularly LST. The high inter-rater variability in case of manual segmentation indicates the complexity of identifying new or enlarged lesions. An automated CNN-based approach can quickly provide an independent and deterministic assessment of new or enlarged lesions from baseline to follow-up scans with acceptable reliability.

MRI-Based Brain Volumetry at a Single Time Point Complements Clinical Evaluation of Patients With Multiple Sclerosis in an Outpatient Setting.

Purpose: Thalamic atrophy and whole brain atrophy in multiple sclerosis (MS) are associated with disease progression. The motivation of this study was to propose and evaluate a new grouping scheme which is based on MS patients' whole brain and thalamus volumes measured on MRI at a single time point. Methods: In total, 185 MS patients (128 relapsing-remitting (RRMS) and 57 secondary-progressive MS (SPMS) patients) were included from an outpatient facility. Whole brain parenchyma (BP) and regional brain volumes were derived from single time point MRI T1 images. Standard scores (z-scores) were computed by comparing individual brain volumes against corresponding volumes from healthy controls. A z-score cut-off of -1.96 was applied to separate pathologically atrophic from normal brain volumes for thalamus and whole BP (accepting a 2.5% error probability). Subgroup differences with respect to the Symbol Digit Modalities Test (SDMT) and the Expanded Disability Status Scale (EDSS) were assessed. Results: Except for two, all MS patients showed either no atrophy (group 0: 61 RRMS patients, 10 SPMS patients); thalamic but no BP atrophy (group 1: 37 RRMS patients; 18 SPMS patients) or thalamic and BP atrophy (group 2: 28 RRMS patients; 29 SPMS patients). RRMS patients without atrophy and RRMS patients with thalamic atrophy did not differ in EDSS, however, patients with thalamus and BP atrophy showed significantly higher EDSS scores than patients in the other groups. Conclusion: MRI-based brain volumetry at a single time point is able to reliably distinguish MS patients with isolated thalamus atrophy (group 1) from those without brain atrophy (group 0). MS patients with isolated thalamus atrophy might be at risk for the development of widespread atrophy and disease progression. Since RRMS patients in group 0 and 1 are clinically not distinguishable, the proposed grouping may aid identification of RRMS patients at risk of disease progression and thus complement clinical evaluation in the routine patient care.

Within-patient fluctuation of brain volume estimates from short-term repeated MRI measurements using SIENA/FSL.

BACKGROUND: Measurements of brain volume loss (BVL) in individual patients are currently discussed controversially. One concern is the impact of short-term biological noise, like hydration status. METHODS: Three publicly available reliability MRI datasets with scan intervals of days to weeks were used. An additional cohort of 60 early relapsing multiple sclerosis (MS) patients with MRI follow-ups was analyzed to test whether after 1 year pathological BVL is detectable in a relevant fraction of MS patients. BVL was determined using SIENA/FSL. Results deviating from zero in the reliability datasets were considered as within-patient fluctuation (WPF) consisting of the intrinsic measurement error as well as the short-term biological fluctuations of brain volumes. We provide an approach to interpret BVL measurements in individual patients taking the WPF into account. RESULTS: The estimated standard deviation of BVL measurements from the pooled reliability datasets was 0.28%. For a BVL measurement of x% per year in an individual patient, the true BVL lies with an error probability of 5% in the interval x% ± (1.96 × 0.28)/(scan interval in years)%. To allow a BVL per year of at least 0.4% to be identified after 1 year, the measured BVL needs to exceed 0.94%. The median BVL per year in the MS patient cohort was 0.44%. In 11 out of 60 MS patients (18%) we found a BVL per year equal or greater than 0.94%. CONCLUSION: The estimated WPF may be helpful when interpreting BVL results on an individual patient level in diseases such as MS.

Estimates of age-dependent cutoffs for pathological brain volume loss using SIENA/FSL-a longitudinal brain volumetry study in healthy adults.

Brain volume loss (BVL) has gained increasing interest for monitoring tissue damage in neurodegenerative diseases including multiple sclerosis (MS). In this longitudinal study, 117 healthy participants (age range 37.3-82.6 years) received at least 2 magnetic resonance imaging examinations. BVL (in %) was determined with the Structural Image Evaluation using Normalisation of Atrophy/FMRIB Software Library and annualized. Mean BVL per year was 0.15%, 0.30%, 0.46%, and 0.61% at ages 45, 55, 65, and 75 years, respectively. The corresponding BVL per year values of the age-dependent 95th percentiles were 0.52%, 0.77%, 1.05% and 1.45%. Pathological BVL can be assumed if an individual BVL per year exceeds these thresholds for a given age. The mean BVL per year determined in this longitudinal study was consistent with results from a cross-sectional study that was published recently. The cut-off for a pathological BVL per year at the age of 45 years (0.52%) was consistent with the cut-off suggested previously to distinguish between physiological and pathological BVL in MS patients. Different cut-off values, however, need to be considered when interpreting BVL assessed in cohorts of higher ages.

Comparative Analysis of Markers of Mass Effect after Ischemic Stroke.

BACKGROUND AND PURPOSE: Midline shift determined on magnetic resonance imaging (MRI) or computed tomography (CT) images is a well-validated marker of mass effect after large hemispheric infarction and associated with mortality. In this study, we targeted a population with moderately sized strokes. We compared midline shift to other imaging markers and determined their ability to predict long-term outcome. METHODS: MRI scans were studied from the Echoplanar Imaging Thrombolysis Evaluation Trial (EPITHET) cohort. Midline shift, acute stroke lesion volume, lesional swelling volume, change in ipsilateral hemisphere volume, the ratio of ipsilateral to contralateral hemisphere volume, and the reduction in lateral ventricle volume were measured. The relationships of these markers with poor outcome (modified Rankin scale score 3-6 at day 90) were assessed. Receiver-operating characteristic (ROC) curves were generated to compare the performance of each metric. RESULTS: Of the 71 included patients, 59.2% had a poor outcome that was associated with significantly larger values for midline shift, lesional swelling volume, and ratio of hemisphere volumes. Lesional swelling volume, change in hemisphere volume, ratio of hemisphere volumes, and lateral ventricle displacement were each correlated with midline shift (Spearman r = .60, .49, .61, and -.56, respectively; all P < .0001). ROC curve analysis showed that lesional swelling volume (area under the curve [AUC] = .791) predicted poor outcome better than midline shift (AUC = .682). For predicting mortality, ROC curve analysis showed that these three markers were equivalent. CONCLUSION: The ratio of ipsilateral to contralateral hemisphere volume, baseline lesion volume and lesional swelling volume best predicted poor outcome across a spectrum of stroke sizes.