Data-driven normative values based on generative manifold learning for quantitative MRI.

In medicine, abnormalities in quantitative metrics such as the volume reduction of one brain region of an individual versus a control group are often provided as deviations from so-called normal values. These normative reference values are traditionally calculated based on the quantitative values from a control group, which can be adjusted for relevant clinical co-variables, such as age or sex. However, these average normative values do not take into account the globality of the available quantitative information. For example, quantitative analysis of T1-weighted magnetic resonance images based on anatomical structure segmentation frequently includes over 100 cerebral structures in the quantitative reports, and these tend to be analyzed separately. In this study, we propose a global approach to personalized normative values for each brain structure using an unsupervised Artificial Intelligence technique known as generative manifold learning. We test the potential benefit of these personalized normative values in comparison with the more traditional average normative values on a population of patients with drug-resistant epilepsy operated for focal cortical dysplasia, as well as on a supplementary healthy group and on patients with Alzheimer's disease.

Individualised profiling of white matter organisation in moderate-to-severe traumatic brain injury patients.

BACKGROUND AND PURPOSE: Approximately 65% of moderate-to-severe traumatic brain injury (m-sTBI) patients present with poor long-term behavioural outcomes, which can significantly impair activities of daily living. Numerous diffusion-weighted MRI studies have linked these poor outcomes to decreased white matter integrity of several commissural tracts, association fibres and projection fibres in the brain. However, most studies have focused on group-based analyses, which are unable to deal with the substantial between-patient heterogeneity in m-sTBI. As a result, there is increasing interest and need in conducting individualised neuroimaging analyses. MATERIALS AND METHODS: Here, we generated a detailed subject-specific characterisation of microstructural organisation of white matter tracts in 5 chronic patients with m-sTBI (29 - 49y, 2 females), presented as a proof-of-concept. We developed an imaging analysis framework using fixel-based analysis and TractLearn to determine whether the values of fibre density of white matter tracts at the individual patient level deviate from the healthy control group (n = 12, 8F, Mage = 35.7y, age range 25 - 64y). RESULTS: Our individualised analysis revealed unique white matter profiles, confirming the heterogenous nature of m-sTBI and the need of individualised profiles to properly characterise the extent of injury. Future studies incorporating clinical data, as well as utilising larger reference samples and examining the test-retest reliability of the fixel-wise metrics are warranted. CONCLUSIONS: Individualised profiles may assist clinicians in tracking recovery and planning personalised training programs for chronic m-sTBI patients, which is necessary to achieve optimal behavioural outcomes and improved quality of life.

HEMO2 life® improves renal function independent of cold ischemia time in kidney recipients: A comparison with a large multicenter prospective cohort study.

BACKGROUND: M101 is an extracellular hemoglobin isolated from a marine lugworm and is present in the medical device HEMO2 life®. The clinical investigation OXYOP was a paired kidney analysis (n = 60) designed to evaluate the safety and performance of HEMO2 life® used as an additive to preservation solution in renal transplantation. The secondary efficacy endpoints showed less delayed graft function (DGF) and better renal function in the HEMO2 life® group but due to the study design cold ischemia time (CIT) was longer in the contralateral kidneys. METHODS: An additional analysis was conducted including OXYOP patients and patients from the ASTRE database (n = 6584) to verify that the decrease in DGF rates observed in the HEMO2 life® group may not be due solely to the shorter CIT but also to HEMO2 life® performance. Kaplan-Meier estimate curves of cumulative probability of achieving a creatinine level below 250 µmol/L were generated and compared in both groups. A Cox model was used to test the effect of the explanatory variables (use of HEMO2 life® and CIT). Finally, a bootstrap strategy was used to randomly select smaller samples of patients and test them for statistical comparison in the ASTRE database. RESULTS: Kaplan-Meier estimate curves confirmed the existence of a relation between DGF and CIT and Cox analysis showed a benefit in the HEMO2 life® group regardless of the associated CIT. Boostrap analysis confirmed these results. CONCLUSIONS: The present study suggested that the better recovery of renal function observed among kidneys preserved with HEMO2 life® in the OXYOP study is a therapeutic benefit of this breakthrough innovative medical device.

A covariate-constraint method to map brain feature space into lower dimensional manifolds.

Human brain connectome studies aim to both explore healthy brains, and extract and analyze relevant features associated with pathologies of interest. Usually this consists of modeling the brain connectome as a graph and using graph metrics as features. A fine brain description requires graph metrics computation at the node level. Given the relatively reduced number of patients in standard cohorts, such data analysis problems fall in the high-dimension, low-sample-size framework. In this context, our goal is to provide a machine learning technique that exhibits flexibility, gives the investigator an understanding of the features and covariates, allows visualization and exploration, and yields insight into the data and the biological phenomena at stake. The retained approach is dimension reduction in a manifold learning methodology; the originality is that the investigator chooses one (or several) reduced variables. The proposed method is illustrated in two studies. The first one addresses comatose patients; the second one compares young and elderly populations. The method sheds light on the differences between brain connectivity graphs using graph metrics and potential clinical interpretations of these differences.

TractLearn: A geodesic learning framework for quantitative analysis of brain bundles.

Deep learning-based convolutional neural networks have recently proved their efficiency in providing fast segmentation of major brain fascicles structures, based on diffusion-weighted imaging. The quantitative analysis of brain fascicles then relies on metrics either coming from the tractography process itself or from each voxel along the bundle. Statistical detection of abnormal voxels in the context of disease usually relies on univariate and multivariate statistics models, such as the General Linear Model (GLM). Yet in the case of high-dimensional low sample size data, the GLM often implies high standard deviation range in controls due to anatomical variability, despite the commonly used smoothing process. This can lead to difficulties to detect subtle quantitative alterations from a brain bundle at the voxel scale. Here we introduce TractLearn, a unified framework for brain fascicles quantitative analyses by using geodesic learning as a data-driven learning task. TractLearn allows a mapping between the image high-dimensional domain and the reduced latent space of brain fascicles using a Riemannian approach. We illustrate the robustness of this method on a healthy population with test-retest acquisition of multi-shell diffusion MRI data, demonstrating that it is possible to separately study the global effect due to different MRI sessions from the effect of local bundle alterations. We have then tested the efficiency of our algorithm on a sample of 5 age-matched subjects referred with mild traumatic brain injury. Our contributions are to propose: 1/ A manifold approach to capture controls variability as standard reference instead of an atlas approach based on a Euclidean mean. 2/ A tool to detect global variation of voxels' quantitative values, which accounts for voxels' interactions in a structure rather than analyzing voxels independently. 3/ A ready-to-plug algorithm to highlight nonlinear variation of diffusion MRI metrics. With this regard, TractLearn is a ready-to-use algorithm for precision medicine.

Birth weight is associated with adolescent brain development: A multimodal imaging study in monozygotic twins.

Previous research has shown that the prenatal environment, commonly indexed by birth weight (BW), is a predictor of morphological brain development. We previously showed in monozygotic (MZ) twins associations between BW and brain morphology that were independent of genetics. In the present study, we employed a longitudinal MZ twin design to investigate whether variations in prenatal environment (as indexed by discordance in BW) are associated with resting-state functional connectivity (rs-FC) and with structural connectivity. We focused on the limbic and default mode networks (DMNs), which are key regions for emotion regulation and internally generated thoughts, respectively. One hundred and six healthy adolescent MZ twins (53 pairs; 42% male pairs) followed longitudinally from birth underwent a magnetic resonance imaging session at age 15. Graph theoretical analysis was applied to rs-FC measures. TrackVis was used to determine track count as an indicator of structural connectivity strength. Lower BW twins had less efficient limbic network connectivity as compared to their higher BW co-twin, driven by differences in the efficiency of the right hippocampus and right amygdala. Lower BW male twins had fewer tracks connecting the right hippocampus and right amygdala as compared to their higher BW male co-twin. There were no associations between BW and the DMN. These findings highlight the possible role of unique prenatal environmental influences in the later development of efficient spontaneous limbic network connections within healthy individuals, irrespective of DNA sequence or shared environment.

Variability and reproducibility in deep learning for medical image segmentation.

Medical image segmentation is an important tool for current clinical applications. It is the backbone of numerous clinical diagnosis methods, oncological treatments and computer-integrated surgeries. A new class of machine learning algorithm, deep learning algorithms, outperforms the results of classical segmentation in terms of accuracy. However, these techniques are complex and can have a high range of variability, calling the reproducibility of the results into question. In this article, through a literature review, we propose an original overview of the sources of variability to better understand the challenges and issues of reproducibility related to deep learning for medical image segmentation. Finally, we propose 3 main recommendations to address these potential issues: (1) an adequate description of the framework of deep learning, (2) a suitable analysis of the different sources of variability in the framework of deep learning, and (3) an efficient system for evaluating the segmentation results.

First-in-human use of a marine oxygen carrier (M101) for organ preservation: A safety and proof-of-principle study.

The medical device M101 is an extracellular hemoglobin featuring high oxygen-carrying capabilities. Preclinical studies demonstrated its safety as an additive to organ preservation solutions and its beneficial effect on ischemia/reperfusion injuries. OXYgen carrier for Organ Preservation (OXYOP) is a multicenter open-label study evaluating for the first time the safety of M101 added (1 g/L) to the preservation solution of one of two kidneys from the same donor. All adverse events (AEs) were analyzed by an independent data and safety monitoring board. Among the 58 donors, 38% were extended criteria donors. Grafts were preserved in cold storage (64%) or machine perfusion (36%) with a mean cold ischemia time (CIT) of 740 minutes. At 3 months, 490 AEs (41 serious) were reported, including two graft losses and two acute rejections (3.4%). No immunological, allergic, or prothrombotic effects were reported. Preimplantation and 3-month biopsies did not show thrombosis or altered microcirculation. Secondary efficacy end points showed less delayed graft function (DGF) and better renal function in the M101 group than in the contralateral kidneys. In the subgroup of grafts preserved in cold storage, Kaplan-Meier survival and Cox regression analysis showed beneficial effects on DGF independent of CIT (P = .048). This study confirms that M101 is safe and shows promising efficacy data.

Author Correction: Association between social asymmetry and depression in older adults: A phone Call Detail Records analysis.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Motor tract integrity predicts walking recovery: A diffusion MRI study in subacute stroke.

OBJECTIVE: To identify candidate biomarkers of walking recovery with motor tract integrity measurements using fractional anisotropy (FA) from the corticospinal tract (CST) and alternative motor pathways in patients with moderate to severe subacute stroke. METHODS: Walking recovery was first assessed with generalized linear mixed model (GLMM) with repeated measures of walking scores (WS) over 2 years of follow-up in a longitudinal study of 29 patients with subacute ischemic stroke. Baseline FA measures from the ipsilesional and contralesional CST (i-CST and c-CST), cortico-reticulospinal pathway (i-CRP and c-CRP), and cerebellar peduncles were derived from a 60-direction diffusion MRI sequence on a 3T scanner. We performed correlation tests between WS and FA measures. Third, we investigated using GLMM whether motor tract integrity contributes to predict walking recovery. RESULTS: We observed significant improvements of WS over time with a plateau reached at ≈6 months after stroke. WS significantly correlated with FA measures from i-CST, c-CST, i-CRP, and bilateral cerebellar peduncles. Walking recovery was predicted by FA measures from 3 tracts: i-CST, i-CRP, and contralesional superior cerebellar peduncle (c-SCP). Diffusion tensor imaging (DTI) predictors captured 80.5% of the unexplained variance of the model without DTI. CONCLUSIONS: We identified i-CST and alternative motor-related tracts (namely i-CRP and c-SCP) as candidate biomarkers of walking recovery. The role of the SCP in walk recovery may rely on cerebellar nuclei projections to the thalamus, red nucleus, and reticular formation. Our findings suggest that a set of white matter tracts, part of subcortical motor networks, contribute to walking recovery in patients with moderate to severe stroke.

Hubs disruption in mesial temporal lobe epilepsy. A resting-state fMRI study on a language-and-memory network.

Mesial temporal lobe epilepsy (mTLE) affects the brain networks at several levels and patients suffering from mTLE experience cognitive impairment for language and memory. Considering the importance of language and memory reorganization in this condition, the present study explores changes of the embedded language-and-memory network (LMN) in terms of functional connectivity (FC) at rest, as measured with functional MRI. We also evaluate the cognitive efficiency of the reorganization, that is, whether or not the reorganizations support or allow the maintenance of optimal cognitive functioning despite the seizure-related damage. Data from 37 patients presenting unifocal mTLE were analyzed and compared to 48 healthy volunteers in terms of LMN-FC using two methods: pairwise correlations (region of interest [ROI]-to-ROI) and graph theory. The cognitive efficiency of the LMN-FC reorganization was measured using correlations between FC parameters and language and memory scores. Our findings revealed a large perturbation of the LMN hubs in patients. We observed a hyperconnectivity of limbic areas near the dysfunctional hippocampus and mainly a hypoconnectivity for several cortical regions remote from the dysfunctional hippocampus. The loss of FC was more important in left mTLE (L-mTLE) than in right (R-mTLE) patients. The LMN-FC reorganization may not be always compensatory and not always useful for patients as it may be associated with lower cognitive performance. We discuss the different connectivity patterns obtained and conclude that interpretation of FC changes in relation to neuropsychological scores is important to determine cognitive efficiency, suggesting the concept of "connectome" would gain to be associated with a "cognitome" concept.

Multimodal assessment of language and memory reorganization: a proof of concept in two patients with drug-resistant temporal lobe epilepsy.

We report two patients suffering from drug-resistant temporal lobe epilepsy to show how their neuroplasticity can be apprehended using a multimodal, integrative and clinically relevant approach. This is a proof of concept based on using multimodal data including: (1) white matter structural connectivity (DTI) of the main tracts involved in language and memory; (2) neurophysiological biomarkers (fMRI-BOLD signal and LI lateralization indices); and (3) cognitive scores as measured during the neuropsychological assessment. We characterized tri-modal data for each patient using a descriptive integrative approach, in terms of reorganization and by comparing with a group of healthy participants. This proof of concept suggests that the inclusion of multimodal data in clinical studies is currently a major challenge. Since the various datasets obtained from MRI neuroimaging and cognitive scores are probably interrelated, it is important to go beyond the mono-modal approach and move towards greater integration of several multimodal data. Multimodal integration of anatomical, functional, and cognitive data facilitates the identification of comprehensive neurocognitive patterns in epileptic patients, thus enabling clinicians to differentiate between reorganization profiles and help to predict post-surgical outcomes for curative neurosurgery.

Association between social asymmetry and depression in older adults: A phone Call Detail Records analysis.

Analyzing social interactions on a passive and non-invasive way through the use of phone call detail records (CDRs) is now recognized as a promising approach in health monitoring. However, deeper investigations are required to confirm its relevance in social interaction modeling. Particularly, no clear consensus exists in the use of the direction parameter characterizing the directed nature of interactions in CDRs. In the present work, we specifically investigate, in a 26-older-adults population over 12 months, whether and how this parameter could be used in CDRs analysis. We then evaluate its added-value for depression assessment regarding the Geriatric Depression Scale score assessed within our population during the study. The results show the existence of three clusters of phone call activity named (1) proactive, (2) interactive, and (3) reactive. Then, we introduce the notion of asymmetry that synthesizes these activities. We find significant correlations between asymmetry and the depressive state assessed in the older individual. Particularly, (1) reactive users are more depressed than the others, and (2) not depressed older adults tend to be proactive. Taken together, the present findings suggest the phone's potential to be used as a social sensor containing relevant health-related insights when the direction parameter is considered.

How to Measure Circadian Rhythms of Activity and Their Disruptions in Humans Using Passive and Unobtrusive Capture of Phone Call Activity.

Monitoring circadian rhythms of social activity is crucial for preserving the health and wellness of ill elderly people. In this paper, we assess the ability of phones to be used as a temporal and social daily activity sensor from a passive and unobstructive measure of phone call activity. To this end, we introduce a methodology specifically designed to automatically measure both persistence and disruptions in circadian rhythms of phone call activity with 26 adults older than 65 years.

Diffusion tensor imaging reveals diffuse white matter injuries in locked-in syndrome patients.

Locked-in syndrome (LIS) is a state of quadriplegia and anarthria with preserved consciousness, which is generally triggered by a disruption of specific white matter fiber tracts, following a lesion in the ventral part of the pons. However, the impact of focal lesions on the whole brain white matter microstructure and structural connectivity pathways remains unknown. We used diffusion tensor magnetic resonance imaging (DT-MRI) and tract-based statistics to characterise the whole white matter tracts in seven consecutive LIS patients, with ventral pontine injuries but no significant supratentorial lesions detected with morphological MRI. The imaging was performed in the acute phase of the disease (26 ± 13 days after the accident). DT-MRI-derived metrics were used to quantitatively assess global white matter alterations. All diffusion coefficient Z-scores were decreased for almost all fiber tracts in all LIS patients, with diffuse white matter alterations in both infratentorial and supratentorial areas. A mixture model of two multidimensional Gaussian distributions was fitted to cluster the white matter fiber tracts studied in two groups: the least (group 1) and most injured white matter fiber tracts (group 2). The greatest injuries were revealed along pathways crossing the lesion responsible for the LIS: left and right medial lemniscus (98.4% and 97.9% probability of belonging to group 2, respectively), left and right superior cerebellar peduncles (69.3% and 45.7% probability) and left and right corticospinal tract (20.6% and 46.5% probability). This approach demonstrated globally compromised white matter tracts in the acute phase of LIS, potentially underlying cognitive deficits.

Gray Matter Volume and Cognitive Performance During Normal Aging. A Voxel-Based Morphometry Study.

Normal aging is characterized by decline in cognitive functioning in conjunction with extensive gray matter (GM) atrophy. A first aim of this study was to determine GM volume differences related to aging by comparing two groups of participants, middle-aged group (MAG, mean age 41 years, N = 16) and older adults (OG, mean age 71 years, N = 14) who underwent an magnetic resonance images (MRI) voxel-based morphometry (VBM) evaluation. The VBM analyses included two optimized pipelines, for the cortex and for the cerebellum. Participants were also evaluated on a wide range of cognitive tests assessing both domain-general and language-specific processes, in order to examine how GM volume differences between OG and MAG relate to cognitive performance. Our results show smaller bilateral GM volume in the OG relative to the MAG, in several cerebral and right cerebellar regions involved in language and executive functions. Importantly, our results also revealed smaller GM volume in the right cerebellum in OG relative to MAG, supporting the idea of a complex cognitive role for this structure. This study provides a broad picture of cerebral, but also cerebellar and cognitive changes associated with normal aging.

Facial nerve tractography: A new tool for the detection of perineural spread in parotid cancers.

OBJECTIVES: To determine whether facial nerve MR tractography is useful in detecting PeriNeural Spread in parotid cancers. METHODS: Forty-five participants were enrolled. Thirty patients with surgically managed parotid tumors (15 malignant, 15 benign) were compared with 15 healthy volunteers. All of them had undergone 3T-MRI with diffusion acquisition and post-processing constrained spherical deconvolution-based tractography. Parameters of diffusion-weighted sequences were b-value 1,000 s/mm2, 32 directions. Two radiologists performed a blinded visual reading of tractographic maps and graded the facial nerve average pathlength and fractional anisotropy (FA). We also compared diagnostic accuracy of tractography with morphological MRI sequences to detect PeriNeural Spread. Non-parametric methods were used. RESULTS: Average pathlength was significantly higher in cases with PeriNeural Spread (39.86 mm [Quartile1: 36.27; Quartile3: 51.19]) versus cases without (16.23 mm [12.90; 24.90]), p<0.001. The threshold above which there was a significant association with PeriNeural Spread was set at 27.36 mm (Se: 100%; Sp: 84%; AUC: 0.96, 95% CI 0.904-1). There were no significant differences in FA between groups. Tractography map visual analyses directly displayed PeriNeural Spread in distal neural ramifications with sensitivity of 75%, versus 50% using morphological sequences. CONCLUSIONS: Tractography could be used to identify facial nerve PeriNeural Spread by parotid cancers. KEY POINTS: • Tractography could detect facial nerve PeriNeural Spread in parotid cancers. • The average pathlength parameter is increased in case of PeriNeural Spread. • Tractography could map PeriNeural Spread more precisely than conventional imaging.

Parietal operculum and motor cortex activities predict motor recovery in moderate to severe stroke.

While motor recovery following mild stroke has been extensively studied with neuroimaging, mechanisms of recovery after moderate to severe strokes of the types that are often the focus for novel restorative therapies remain obscure. We used fMRI to: 1) characterize reorganization occurring after moderate to severe subacute stroke, 2) identify brain regions associated with motor recovery and 3) to test whether brain activity associated with passive movement measured in the subacute period could predict motor outcome six months later. Because many patients with large strokes involving sensorimotor regions cannot engage in voluntary movement, we used passive flexion-extension of the paretic wrist to compare 21 patients with subacute ischemic stroke to 24 healthy controls one month after stroke. Clinical motor outcome was assessed with Fugl-Meyer motor scores (motor-FMS) six months later. Multiple regression, with predictors including baseline (one-month) motor-FMS and sensorimotor network regional activity (ROI) measures, was used to determine optimal variable selection for motor outcome prediction. Sensorimotor network ROIs were derived from a meta-analysis of arm voluntary movement tasks. Bootstrapping with 1000 replications was used for internal model validation. During passive movement, both control and patient groups exhibited activity increases in multiple bilateral sensorimotor network regions, including the primary motor (MI), premotor and supplementary motor areas (SMA), cerebellar cortex, putamen, thalamus, insula, Brodmann area (BA) 44 and parietal operculum (OP1-OP4). Compared to controls, patients showed: 1) lower task-related activity in ipsilesional MI, SMA and contralesional cerebellum (lobules V-VI) and 2) higher activity in contralesional MI, superior temporal gyrus and OP1-OP4. Using multiple regression, we found that the combination of baseline motor-FMS, activity in ipsilesional MI (BA4a), putamen and ipsilesional OP1 predicted motor outcome measured 6 months later (adjusted-R2 = 0.85; bootstrap p < 0.001). Baseline motor-FMS alone predicted only 54% of the variance. When baseline motor-FMS was removed, the combination of increased activity in ipsilesional MI-BA4a, ipsilesional thalamus, contralesional mid-cingulum, contralesional OP4 and decreased activity in ipsilesional OP1, predicted better motor outcome (djusted-R2 = 0.96; bootstrap p < 0.001). In subacute stroke, fMRI brain activity related to passive movement measured in a sensorimotor network defined by activity during voluntary movement predicted motor recovery better than baseline motor-FMS alone. Furthermore, fMRI sensorimotor network activity measures considered alone allowed excellent clinical recovery prediction and may provide reliable biomarkers for assessing new therapies in clinical trial contexts. Our findings suggest that neural reorganization related to motor recovery from moderate to severe stroke results from balanced changes in ipsilesional MI (BA4a) and a set of phylogenetically more archaic sensorimotor regions in the ventral sensorimotor trend, in which OP1 and OP4 processes may complement the ipsilesional dorsal motor cortex in achieving compensatory sensorimotor recovery.

Resting-state networks distinguish locked-in from vegetative state patients.

PURPOSE: Locked-in syndrome and vegetative state are distinct outcomes from coma. Despite their differences, they are clinically difficult to distinguish at the early stage and current diagnostic tools remain insufficient. Since some brain functions are preserved in locked-in syndrome, we postulated that networks of spontaneously co-activated brain areas might be present in locked-in patients, similar to healthy controls, but not in patients in a vegetative state. METHODS: Five patients with locked-in syndrome, 12 patients in a vegetative state and 19 healthy controls underwent a resting-state fMRI scan. Individual spatial independent component analysis was used to separate spontaneous brain co-activations from noise. These co-activity maps were selected and then classified by two raters as either one of eight resting-state networks commonly shared across subjects or as specific to a subject. RESULTS: The numbers of spontaneous co-activity maps, total resting-state networks, and resting-state networks underlying high-level cognitive activity were shown to differentiate controls and locked-in patients from patients in a vegetative state. Analyses of each common resting-state network revealed that the default mode network accurately distinguished locked-in from vegetative-state patients. The frontoparietal network also had maximum specificity but more limited sensitivity. CONCLUSIONS: This study reinforces previous reports on the preservation of the default mode network in locked-in syndrome in contrast to vegetative state but extends them by suggesting that other networks might be relevant to the diagnosis of locked-in syndrome. The aforementioned analysis of fMRI brain activity at rest might be a step in the development of a diagnostic biomarker to distinguish locked-in syndrome from vegetative state.