Automated identification of uncertain cases in deep learning-based classification of dopamine transporter SPECT to improve clinical utility and acceptance.

PURPOSE: Deep convolutional neural networks (CNN) are promising for automatic classification of dopamine transporter (DAT)-SPECT images. Reporting the certainty of CNN-based decisions is highly desired to flag cases that might be misclassified and, therefore, require particularly careful inspection by the user. The aim of the current study was to design and validate a CNN-based system for the identification of uncertain cases. METHODS: A network ensemble (NE) combining five CNNs was trained for binary classification of [123I]FP-CIT DAT-SPECT images as "normal" or "neurodegeneration-typical reduction" with high accuracy (NE for classification, NEfC). An uncertainty detection module (UDM) was obtained by combining two additional NE, one trained for detection of "reduced" DAT-SPECT with high sensitivity, the other with high specificity. A case was considered "uncertain" if the "high sensitivity" NE and the "high specificity" NE disagreed. An internal "development" dataset of 1740 clinical DAT-SPECT images was used for training (n = 1250) and testing (n = 490). Two independent datasets with different image characteristics were used for testing only (n = 640, 645). Three established approaches for uncertainty detection were used for comparison (sigmoid, dropout, model averaging). RESULTS: In the test data from the development dataset, the NEfC achieved 98.0% accuracy. 4.3% of all test cases were flagged as "uncertain" by the UDM: 2.5% of the correctly classified cases and 90% of the misclassified cases. NEfC accuracy among "certain" cases was 99.8%. The three comparison methods were less effective in labelling misclassified cases as "uncertain" (40-80%). These findings were confirmed in both additional test datasets. CONCLUSION: The UDM allows reliable identification of uncertain [123I]FP-CIT SPECT with high risk of misclassification. We recommend that automatic classification of [123I]FP-CIT SPECT images is combined with an UDM to improve clinical utility and acceptance. The proposed UDM method ("high sensitivity versus high specificity") might be useful also for DAT imaging with other ligands and for other binary classification tasks.

Removing outliers from the normative database improves regional atrophy detection in single-subject voxel-based morphometry.

PURPOSE: Single-subject voxel-based morphometry (VBM) compares an individual T1-weighted MRI to a sample of normal MRI in a normative database (NDB) to detect regional atrophy. Outliers in the NDB might result in reduced sensitivity of VBM. The primary aim of the current study was to propose a method for outlier removal ("NDB cleaning") and to test its impact on the performance of VBM for detection of Alzheimer's disease (AD) and frontotemporal lobar degeneration (FTLD). METHODS: T1-weighted MRI of 81 patients with biomarker-confirmed AD (n = 51) or FTLD (n = 30) and 37 healthy subjects with simultaneous FDG-PET/MRI were included as test dataset. Two different NDBs were used: a scanner-specific NDB (37 healthy controls from the test dataset) and a non-scanner-specific NDB comprising 164 normal T1-weighted MRI from 164 different MRI scanners. Three different quality metrics based on leave-one-out testing of the scans in the NDB were implemented. A scan was removed if it was an outlier with respect to one or more quality metrics. VBM maps generated with and without NDB cleaning were assessed visually for the presence of AD or FTLD. RESULTS: Specificity of visual interpretation of the VBM maps for detection of AD or FTLD was 100% in all settings. Sensitivity was increased by NDB cleaning with both NDBs. The effect was statistically significant for the multiple-scanner NDB (from 0.47 [95%-CI 0.36-0.58] to 0.61 [0.49-0.71]). CONCLUSION: NDB cleaning has the potential to improve the sensitivity of VBM for the detection of AD or FTLD without increasing the risk of false positive findings.

Voxel-based morphometry in single subjects without a scanner-specific normal database using a convolutional neural network.

OBJECTIVES: Reliable detection of disease-specific atrophy in individual T1w-MRI by voxel-based morphometry (VBM) requires scanner-specific normal databases (NDB), which often are not available. The aim of this retrospective study was to design, train, and test a deep convolutional neural network (CNN) for single-subject VBM without the need for a NDB (CNN-VBM). MATERIALS AND METHODS: The training dataset comprised 8945 T1w scans from 65 different scanners. The gold standard VBM maps were obtained by conventional VBM with a scanner-specific NDB for each of the 65 scanners. CNN-VBM was tested in an independent dataset comprising healthy controls (n = 37) and subjects with Alzheimer's disease (AD, n = 51) or frontotemporal lobar degeneration (FTLD, n = 30). A scanner-specific NDB for the generation of the gold standard VBM maps was available also for the test set. The technical performance of CNN-VBM was characterized by the Dice coefficient of CNN-VBM maps relative to VBM maps from scanner-specific VBM. For clinical testing, VBM maps were categorized visually according to the clinical diagnoses in the test set by two independent readers, separately for both VBM methods. RESULTS: The VBM maps from CNN-VBM were similar to the scanner-specific VBM maps (median Dice coefficient 0.85, interquartile range [0.81, 0.90]). Overall accuracy of the visual categorization of the VBM maps for the detection of AD or FTLD was 89.8% for CNN-VBM and 89.0% for scanner-specific VBM. CONCLUSION: CNN-VBM without NDB provides a similar performance in the detection of AD- and FTLD-specific atrophy as conventional VBM. CLINICAL RELEVANCE STATEMENT: A deep convolutional neural network for voxel-based morphometry eliminates the need of scanner-specific normal databases without relevant performance loss and, therefore, could pave the way for the widespread clinical use of voxel-based morphometry to support the diagnosis of neurodegenerative diseases. KEY POINTS: • The need of normal databases is a barrier for widespread use of voxel-based brain morphometry. • A convolutional neural network achieved a similar performance for detection of atrophy than conventional voxel-based morphometry. • Convolutional neural networks can pave the way for widespread clinical use of voxel-based morphometry.

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.

Single-subject analysis of regional brain volumetric measures can be strongly influenced by the method for head size adjustment.

PURPOSE: Total intracranial volume (TIV) is often a nuisance covariate in MRI-based brain volumetry. This study compared two TIV adjustment methods with respect to their impact on z-scores in single subject analyses of regional brain volume estimates. METHODS: Brain parenchyma, hippocampus, thalamus, and TIV were segmented in a normal database comprising 5059 T1w images. Regional volume estimates were adjusted for TIV using the residual method or the proportion method. Age was taken into account by regression with both methods. TIV- and age-adjusted regional volumes were transformed to z-scores and then compared between the two adjustment methods. Their impact on the detection of thalamus atrophy was tested in 127 patients with multiple sclerosis. RESULTS: The residual method removed the association with TIV in all regions. The proportion method resulted in a switch of the direction without relevant change of the strength of the association. The reduction of physiological between-subject variability was larger with the residual method than with the proportion method. The difference between z-scores obtained with the residual method versus the proportion method was strongly correlated with TIV. It was larger than one z-score point in 5% of the subjects. The area under the ROC curve of the TIV- and age-adjusted thalamus volume for identification of multiple sclerosis patients was larger with the residual method than with the proportion method (0.84 versus 0.79). CONCLUSION: The residual method should be preferred for TIV and age adjustments of T1w-MRI-based brain volume estimates in single subject analyses.

BrainLossNet: a fast, accurate and robust method to estimate brain volume loss from longitudinal MRI.

PURPOSE: MRI-derived brain volume loss (BVL) is widely used as neurodegeneration marker. SIENA is state-of-the-art for BVL measurement, but limited by long computation time. Here we propose "BrainLossNet", a convolutional neural network (CNN)-based method for BVL-estimation. METHODS: BrainLossNet uses CNN-based non-linear registration of baseline(BL)/follow-up(FU) 3D-T1w-MRI pairs. BVL is computed by non-linear registration of brain parenchyma masks segmented in the BL/FU scans. The BVL estimate is corrected for image distortions using the apparent volume change of the total intracranial volume. BrainLossNet was trained on 1525 BL/FU pairs from 83 scanners. Agreement between BrainLossNet and SIENA was assessed in 225 BL/FU pairs from 94 MS patients acquired with a single scanner and 268 BL/FU pairs from 52 scanners acquired for various indications. Robustness to short-term variability of 3D-T1w-MRI was compared in 354 BL/FU pairs from a single healthy men acquired in the same session without repositioning with 116 scanners (Frequently-Traveling-Human-Phantom dataset, FTHP). RESULTS: Processing time of BrainLossNet was 2-3 min. The median [interquartile range] of the SIENA-BrainLossNet BVL difference was 0.10% [- 0.18%, 0.35%] in the MS dataset, 0.08% [- 0.14%, 0.28%] in the various indications dataset. The distribution of apparent BVL in the FTHP dataset was narrower with BrainLossNet (p = 0.036; 95th percentile: 0.20% vs 0.32%). CONCLUSION: BrainLossNet on average provides the same BVL estimates as SIENA, but it is significantly more robust, probably due to its built-in distortion correction. Processing time of 2-3 min makes BrainLossNet suitable for clinical routine. This can pave the way for widespread clinical use of BVL estimation from intra-scanner BL/FU pairs.

A systematic approach to deep learning-based nodule detection in chest radiographs.

Lung cancer is a serious disease responsible for millions of deaths every year. Early stages of lung cancer can be manifested in pulmonary lung nodules. To assist radiologists in reducing the number of overseen nodules and to increase the detection accuracy in general, automatic detection algorithms have been proposed. Particularly, deep learning methods are promising. However, obtaining clinically relevant results remains challenging. While a variety of approaches have been proposed for general purpose object detection, these are typically evaluated on benchmark data sets. Achieving competitive performance for specific real-world problems like lung nodule detection typically requires careful analysis of the problem at hand and the selection and tuning of suitable deep learning models. We present a systematic comparison of state-of-the-art object detection algorithms for the task of lung nodule detection. In this regard, we address the critical aspect of class imbalance and and demonstrate a data augmentation approach as well as transfer learning to boost performance. We illustrate how this analysis and a combination of multiple architectures results in state-of-the-art performance for lung nodule detection, which is demonstrated by the proposed model winning the detection track of the Node21 competition. The code for our approach is available at https://github.com/FinnBehrendt/node21-submit.

AI-based detection of contrast-enhancing MRI lesions in patients with multiple sclerosis.

BACKGROUND: Contrast-enhancing (CE) lesions are an important finding on brain magnetic resonance imaging (MRI) in patients with multiple sclerosis (MS) but can be missed easily. Automated solutions for reliable CE lesion detection are emerging; however, independent validation of artificial intelligence (AI) tools in the clinical routine is still rare. METHODS: A three-dimensional convolutional neural network for CE lesion segmentation was trained externally on 1488 datasets of 934 MS patients from 81 scanners using concatenated information from FLAIR and T1-weighted post-contrast imaging. This externally trained model was tested on an independent dataset comprising 504 T1-weighted post-contrast and FLAIR image datasets of MS patients from clinical routine. Two neuroradiologists (R1, R2) labeled CE lesions for gold standard definition in the clinical test dataset. The algorithmic output was evaluated on both patient- and lesion-level. RESULTS: On a patient-level, recall, specificity, precision, and accuracy of the AI tool to predict patients with CE lesions were 0.75, 0.99, 0.91, and 0.96. The agreement between the AI tool and both readers was within the range of inter-rater agreement (Cohen's kappa; AI vs. R1: 0.69; AI vs. R2: 0.76; R1 vs. R2: 0.76). On a lesion-level, false negative lesions were predominately found in infratentorial location, significantly smaller, and at lower contrast than true positive lesions (p < 0.05). CONCLUSIONS: AI-based identification of CE lesions on brain MRI is feasible, approaching human reader performance in independent clinical data and might be of help as a second reader in the neuroradiological assessment of active inflammation in MS patients. CRITICAL RELEVANCE STATEMENT: Al-based detection of contrast-enhancing multiple sclerosis lesions approaches human reader performance, but careful visual inspection is still needed, especially for infratentorial, small and low-contrast lesions.

Three-dimensional deep learning with spatial erasing for unsupervised anomaly segmentation in brain MRI.

PURPOSE: Brain Magnetic Resonance Images (MRIs) are essential for the diagnosis of neurological diseases. Recently, deep learning methods for unsupervised anomaly detection (UAD) have been proposed for the analysis of brain MRI. These methods rely on healthy brain MRIs and eliminate the requirement of pixel-wise annotated data compared to supervised deep learning. While a wide range of methods for UAD have been proposed, these methods are mostly 2D and only learn from MRI slices, disregarding that brain lesions are inherently 3D and the spatial context of MRI volumes remains unexploited. METHODS: We investigate whether using increased spatial context by using MRI volumes combined with spatial erasing leads to improved unsupervised anomaly segmentation performance compared to learning from slices. We evaluate and compare 2D variational autoencoder (VAE) to their 3D counterpart, propose 3D input erasing, and systemically study the impact of the data set size on the performance. RESULTS: Using two publicly available segmentation data sets for evaluation, 3D VAEs outperform their 2D counterpart, highlighting the advantage of volumetric context. Also, our 3D erasing methods allow for further performance improvements. Our best performing 3D VAE with input erasing leads to an average DICE score of 31.40% compared to 25.76% for the 2D VAE. CONCLUSIONS: We propose 3D deep learning methods for UAD in brain MRI combined with 3D erasing and demonstrate that 3D methods clearly outperform their 2D counterpart for anomaly segmentation. Also, our spatial erasing method allows for further performance improvements and reduces the requirement for large data sets.

T-Cell Specificity Influences Disease Heterogeneity in Multiple Sclerosis.

BACKGROUND AND OBJECTIVES: Encouraged by the enormous progress that the identification of specific autoantigens added to the understanding of neurologic autoimmune diseases, we undertook here an in-depth study of T-cell specificities in the autoimmune disease multiple sclerosis (MS), for which the spectrum of responsible autoantigens is not fully defined yet. The identification of target antigens in MS is crucial for therapeutic strategies aimed to induce antigen-specific tolerance. In addition, knowledge of relevant T-cell targets can improve our understanding of disease heterogeneity, a hallmark of MS that complicates clinical management. METHODS: The proliferative response and interferon gamma (IFN-γ) release of CSF-infiltrating CD4+ T cells from patients with MS against several autoantigens was used to identify patients with different intrathecal T-cell specificities. Fresh CSF-infiltrating and paired circulating lymphocytes in these patients were characterized in depth by ex vivo immunophenotyping and transcriptome analysis of relevant T-cell subsets. Further examination of these patients included CSF markers of inflammation and neurodegeneration and a detailed characterization with respect to demographic, clinical, and MRI features. RESULTS: By testing CSF-infiltrating CD4+ T cells from 105 patients with MS against seven long-known myelin and five recently described GDP-l-fucose synthase peptides, we identified GDP-l-fucose synthase and myelin oligodendrocyte glycoprotein (35-55) responder patients. Immunophenotyping of CSF and paired blood samples in these patients revealed a significant expansion of an effector memory (CCR7- CD45RA-) CD27- Th1 CD4+ cell subset in GDP-l-fucose synthase responders. Subsequent transcriptome analysis of this subset demonstrated expression of Th1 and cytotoxicity-associated genes. Patients with different intrathecal T-cell specificities also differ regarding inflammation- and neurodegeneration-associated biomarkers, imaging findings, expression of HLA class II alleles, and seasonal distribution of the time of the lumbar puncture. DISCUSSION: Our observations reveal an association between autoantigen reactivity and features of disease heterogeneity that strongly supports an important role of T-cell specificity in MS pathogenesis. These data have the potential to improve patient classification in clinical practice and to guide the development of antigen-specific tolerization strategies.

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.

Age-dependent cut-offs for pathological deep gray matter and thalamic volume loss using Jacobian integration.

INTRODUCTION: Several recent studies indicate that deep gray matter or thalamic volume loss (VL) might be promising surrogate markers of disease activity in multiple sclerosis (MS) patients. To allow applying these markers to individual MS patients in clinical routine, age-dependent cut-offs distinguishing physiological from pathological VL and an estimation of the measurement error, which provides the confidence of the result, are to be defined. METHODS: Longitudinal MRI scans of the following cohorts were analyzed in this study: 189 healthy controls (HC) (mean age 54 years, 22% female), 98 MS patients from Zurich university hospital (mean age 34 years, 62% female), 33 MS patients from Dresden university hospital (mean age 38 years, 60% female), and publicly available reliability data sets consisting of 162 short-term MRI scan-rescan pairs with scan intervals of days or few weeks. Percentage annualized whole brain volume loss (BVL), gray matter (GM) volume loss (GMVL), deep gray matter volume loss (deep GMVL), and thalamic volume loss (ThalaVL) were computed deploying the Jacobian integration (JI) method. BVL was additionally computed using Siena, an established method used in many Phase III drug trials. A linear mixed effect model was used to estimate the measurement error as the standard deviation (SD) of model residuals of all 162 scan-rescan pairs For estimation of age-dependent cut-offs, a quadratic regression function between age and the corresponding annualized VL values of the HC was computed. The 5th percentile was defined as the threshold for pathological VL per year since 95% of HC subjects exhibit a less pronounced VL for a given age. For the MS patients BVL, GMVL, deep GMVL, and ThalaVL were mutually compared and a paired t-test was used to test whether there are systematic differences in VL between these brain regions. RESULTS: Siena and JI showed a high agreement for BVL measures, with a median absolute difference of 0.1% and a correlation coefficient of r = 0.78. Siena and GMVL showed a similar standard deviation (SD) of the scan-rescan error of 0.28% and 0.29%, respectively. For deep GMVL, ThalaVL the SD of the scan-rescan error was slightly higher (0.43% and 0.5%, respectively). Among the HC the thalamus showed the highest mean VL (-0.16%, -0.39%, and -0.59% at ages 35, 55, and 75, respectively). Corresponding cut-offs for a pathological VL/year were -0.68%, -0.91%, and -1.11%. The MS cohorts did not differ in BVL and GMVL. However, both MS cohorts showed a significantly (p = 0.05) stronger deep GMVL than BVL per year. CONCLUSION: It might be methodologically feasible to assess deep GMVL using JI in individual MS patients. However, age and the measurement error need to be taken into account. Furthermore, deep GMVL may be used as a complementary marker to BVL since MS patients exhibit a significantly stronger deep GMVL than BVL.

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.

MS optic neuritis-induced long-term structural changes within the visual pathway.

BACKGROUND: The visual pathway is commonly involved in multiple sclerosis (MS), even in its early stages, including clinical episodes of optic neuritis (ON). The long-term structural damage within the visual compartment in patients with ON, however, is yet to be elucidated. OBJECTIVE: Our aim was to characterize visual system structure abnormalities using MRI along with optical coherence tomography (OCT) and pattern-reversal visual evoked potentials (VEPs) depending on a single history of ON. METHODS: Twenty-eight patients with clinically definitive MS, either with a history of a single ON (HON) or without such history and normal VEP findings (NON), were included. OCT measures comprised OCT-derived peripapillary retinal nerve fiber layer (RNFL) and macular ganglion cell/inner plexiform layer (GCIPL) thickness. Cortical and global gray and white matter, thalamic, and T2 lesion volumes were assessed using structural MRI. Diffusion-weighted MRI-derived measures included fractional anisotropy (FA), mean (MD), radial (RD), and axial (AD) diffusivity within the optic radiation (OR). RESULTS: Mean (SD) duration after ON was 8.3 (3.7) years. Compared with the NON group, HON patients showed significant RNFL (p = 0.01) and GCIPL thinning (p = 0.002). OR FA (p = 0.014), MD (p = 0.005), RD (p = 0.007), and AD (p = 0.004) were altered compared with NON. Global gray and white as well as other regional gray matter structures did not differ between the 2 groups. CONCLUSION: A single history of ON induces long-term structural damage within the retina and OR suggestive of both retrograde and anterograde neuroaxonal degeneration.

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.

MRI FLAIR lesion segmentation in multiple sclerosis: Does automated segmentation hold up with manual annotation?

INTRODUCTION: Magnetic resonance imaging (MRI) has become key in the diagnosis and disease monitoring of patients with multiple sclerosis (MS). Both, T2 lesion load and Gadolinium (Gd) enhancing T1 lesions represent important endpoints in MS clinical trials by serving as a surrogate of clinical disease activity. T2- and fluid-attenuated inversion recovery (FLAIR) lesion quantification - largely due to methodological constraints - is still being performed manually or in a semi-automated fashion, although strong efforts have been made to allow automated quantitative lesion segmentation. In 2012, Schmidt and co-workers published an algorithm to be applied on FLAIR sequences. The aim of this study was to apply the Schmidt algorithm on an independent data set and compare automated segmentation to inter-rater variability of three independent, experienced raters. METHODS: MRI data of 50 patients with RRMS were randomly selected from a larger pool of MS patients attending the MS Clinic at the Brain and Mind Centre, University of Sydney, Australia. MRIs were acquired on a 3.0T GE scanner (Discovery MR750, GE Medical Systems, Milwaukee, WI) using an 8 channel head coil. We determined T2-lesion load (total lesion volume and total lesion number) using three versions of an automated segmentation algorithm (Lesion growth algorithm (LGA) based on SPM8 or SPM12 and lesion prediction algorithm (LPA) based on SPM12) as first described by Schmidt et al. (2012). Additionally, manual segmentation was performed by three independent raters. We calculated inter-rater correlation coefficients (ICC) and dice coefficients (DC) for all possible pairwise comparisons. RESULTS: We found a strong correlation between manual and automated lesion segmentation based on LGA SPM8, regarding lesion volume (ICC = 0.958 and DC = 0.60) that was not statistically different from the inter-rater correlation (ICC = 0.97 and DC = 0.66). Correlation between the two other algorithms (LGA SPM12 and LPA SPM12) and manual raters was weaker but still adequate (ICC = 0.927 and DC = 0.53 for LGA SPM12 and ICC = 0.949 and DC = 0.57 for LPA SPM12). Variability of both manual and automated segmentation was significantly higher regarding lesion numbers. CONCLUSION: Automated lesion volume quantification can be applied reliably on FLAIR data sets using the SPM based algorithm of Schmidt et al. and shows good agreement with manual segmentation.

Global and regional annual brain volume loss rates in physiological aging.

The objective is to estimate average global and regional percentage brain volume loss per year (BVL/year) of the physiologically ageing brain. Two independent, cross-sectional single scanner cohorts of healthy subjects were included. The first cohort (n = 248) was acquired at the Medical Prevention Center (MPCH) in Hamburg, Germany. The second cohort (n = 316) was taken from the Open Access Series of Imaging Studies (OASIS). Brain parenchyma (BP), grey matter (GM), white matter (WM), corpus callosum (CC), and thalamus volumes were calculated. A non-parametric technique was applied to fit the resulting age-volume data. For each age, the BVL/year was derived from the age-volume curves. The resulting BVL/year curves were compared between the two cohorts. For the MPCH cohort, the BVL/year curve of the BP was an increasing function starting from 0.20% at the age of 35 years increasing to 0.52% at 70 years (corresponding values for GM ranged from 0.32 to 0.55%, WM from 0.02 to 0.47%, CC from 0.07 to 0.48%, and thalamus from 0.25 to 0.54%). Mean absolute difference between BVL/year trajectories across the age range of 35-70 years was 0.02% for BP, 0.04% for GM, 0.04% for WM, 0.11% for CC, and 0.02% for the thalamus. Physiological BVL/year rates were remarkably consistent between the two cohorts and independent from the scanner applied. Average BVL/year was clearly age and compartment dependent. These results need to be taken into account when defining cut-off values for pathological annual brain volume loss in disease models, such as multiple sclerosis.