Projected pooling loss for red nucleus segmentation with soft topology constraints.

Purpose: Deep learning is the standard for medical image segmentation. However, it may encounter difficulties when the training set is small. Also, it may generate anatomically aberrant segmentations. Anatomical knowledge can be potentially useful as a constraint in deep learning segmentation methods. We propose a loss function based on projected pooling to introduce soft topological constraints. Our main application is the segmentation of the red nucleus from quantitative susceptibility mapping (QSM) which is of interest in parkinsonian syndromes. Approach: This new loss function introduces soft constraints on the topology by magnifying small parts of the structure to segment to avoid that they are discarded in the segmentation process. To that purpose, we use projection of the structure onto the three planes and then use a series of MaxPooling operations with increasing kernel sizes. These operations are performed both for the ground truth and the prediction and the difference is computed to obtain the loss function. As a result, it can reduce topological errors as well as defects in the structure boundary. The approach is easy to implement and computationally efficient. Results: When applied to the segmentation of the red nucleus from QSM data, the approach led to a very high accuracy (Dice 89.9%) and no topological errors. Moreover, the proposed loss function improved the Dice accuracy over the baseline when the training set was small. We also studied three tasks from the medical segmentation decathlon challenge (MSD) (heart, spleen, and hippocampus). For the MSD tasks, the Dice accuracies were similar for both approaches but the topological errors were reduced. Conclusions: We propose an effective method to automatically segment the red nucleus which is based on a new loss for introducing topology constraints in deep learning segmentation.

Incorporation of edited MRS into clinical practice may improve care of patients with IDH-mutant glioma.

BACKGROUND AND PURPOSE: Isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion classify adult-type diffuse gliomas into three tumor subtypes with distinct prognosis. We aimed to evaluate the performance of edited magnetic resonance spectroscopy (MRS) for glioma subtyping in a clinical setting, via the quantification of D-2-hydroxyglutarate (2HG) and cystathionine. The delay between this noninvasive classification and the integrated histomolecular analysis was also quantified. MATERIALS AND METHODS: Subjects with presumed low-grade glioma, eligible for surgery (cohort 1), and subjects with IDH-mutant glioma, previously treated and with progressive disease (cohort 2) were prospectively examined with a singlevoxel Mescher-Garwood point-resolved spectroscopy sequence at 3 T. Spectra were quantified using LCModel. The Cramér-Rao lower bounds (CRLB) threshold was set to 20%. Integrated histomolecular analysis according to the 2021 WHO classification was considered as a ground truth. RESULTS: Thirty-four consecutive subjects were enrolled. Due to poor spectra quality and lack of histological specimen, data from 26 subjects was analyzed. Twenty-one belonged to cohort 1 [11 females; median age: 42 years] and 5 to cohort 2 [3 females; median age: 48 years]. Edited MRS showed 100% specificity for detection of IDH mutation and 91% specificity for prediction of 1p/19q codeletion status. Sensitivities for prediction of IDH and 1p/19q codeletion were 62% and 33%, respectively. The median CRLB values were 14% (13 - 32) for IDH-mutant and 572% (554 - 999) for IDH-wild-type tumors. The time between MRS and surgery was longer for low-grade than high-grade gliomas (p = .03), yet the time between MRS and WHO diagnosis did not differ between grades (p = .07), possibly reflecting molecular analyses induced delays in high-grade gliomas. CONCLUSIONS: Our results, acquired in a clinic setting, confirmed that edited MRS is highly specific for both IDH mutation and 1p/19q codeletion predictions and can provide a faster prognosis stratification. In the upcoming IDH-inhibitor treatment era, incorporation of edited MRS into clinical workflow is desirable. ABBREVIATIONS: 2HG = D-2-hydroxyglutarate; Cth = cystathionine. CRLB: Cramér-Rao lower bound; IDH: isocitrate dehydrogenase.

Hippocampal activations obtained during language fMRI tasks: A complementary tool for predicting postoperative memory prognosis.

In medial temporal lobe epilepsy (MTLE), the benefits of surgery must be balanced against the risk of post-operative memory decline. Prediction of postoperative outcomes based on functional magnetic resonance imaging (fMRI) tasks is increasingly common but remains uncertain. The aim of this retrospective study was to determine whether hippocampal activations elicited by fMRI language tasks could enhance or refine memory fMRI in MTLE patients candidates to surgery. Forty-six patients were included: 30 right and 16 left MTLE, mostly with hippocampal sclerosis. Preoperative assessment included neuropsychological tests and fMRI with language (syntactic verbal fluency) and memory tasks (encoding, delayed, and immediate recognition of images of objects). Thirty patients underwent surgery and had neuropsychological evaluations one year after surgery. Worsening was defined as a degradation of more than 10 % in postoperative forgetting scores compared to preoperative scores in verbal, non-verbal and global memory. Memory fMRI had the best sensitivity with hippocampal activations obtained in 95 % of patients, versus 65 % with language fMRI. Considering the patients who elicited an hippocampal activation, language fMRI led to 80 %, 65 % and 85 % of correct predictions for respectively global, verbal and non verbal memory (versus 71 %, 64 % and 68 % with memory fMRI). Memory and language fMRI predictions outperformed those made by neuropsychological tests. In summary, language fMRI was less sensitive than memory fMRI to elicit hippocampal activations but when it did, the proportion of correct memory predictions was better. Moreover, it proved to be an independent predictive factor regardless of the side of the epileptic focus. Given the ease of setting up a language task in fMRI, we recommend the systematic combination of memory and language tasks to predict the post-operative memory outcome of MTLE patients undergoing epilepsy surgery.

Ultra-short time-echo based ray tracing for transcranial focused ultrasound aberration correction in human calvaria.

OBJECTIVE: Magnetic resonance guided transcranial focused ultrasound holds great promises for treating neurological disorders. This technique relies on skull aberration correction which requires computed tomography (CT) scans of the skull of the patients. Recently, ultra-short time-echo (UTE) magnetic resonance (MR) sequences have unleashed the MRI potential to reveal internal bone structures. In this study, we measure the efficacy of transcranial aberration correction using UTE images. Approach. We compare the efficacy of transcranial aberration correction using UTE scans to CT based correction on four skulls and two targets using a clinical device (Exablate Neuro, Insightec, Israel). We also evaluate the performance of a custom ray tracing algorithm using both UTE and CT estimates of acoustic properties and compare these against the performance of the manufacturer's proprietary aberration correction software. Main results. UTE estimated skull maps in Hounsfield units (HU) had a mean absolute error of 242 ± 20 HU (n=4). The UTE skull maps were sufficiently accurate to improve pressure at the target (no correction: 0.44 ± 0.10, UTE correction: 0.79 ± 0.05, manufacturer CT: 0.80 ± 0.05), pressure confinement ratios (no correction: 0.45 ± 0.10, UTE correction: 0.80 ± 0.05, manufacturer CT: 0.81 ± 0.05), and targeting error (no correction: 1.06 ± 0.42 mm, UTE correction 0.30 ± 0.23 mm, manufacturer CT: 0.32 ± 0.22) (n=8 for all values). When using CT, our ray tracing algorithm performed slightly better than UTE based correction with pressure at the target (UTE: 0.79 ± 0.05, CT: 0.84 ± 0.04), pressure confinement ratios (UTE: 0.80 ± 0.05, CT: 0.84 ± 0.04), and targeting error (UTE: 0.30 ± 0.23 mm, CT: 0.17 ± 0.15). Significance. These 3D transcranial measurements suggest that UTE sequences could replace CT scans in the case of MR guided focused ultrasound with minimal reduction in performance which will avoid ionizing radiation exposure to the patients and reduce procedure time and cost. .

Combined in vivo MRI assessment of locus coeruleus and nucleus basalis of Meynert integrity in amnestic Alzheimer's disease, suspected-LATE and frontotemporal dementia.

BACKGROUND: The locus coeruleus (LC) and the nucleus basalis of Meynert (NBM) are altered in early stages of Alzheimer's disease (AD). Little is known about LC and NBM alteration in limbic-predominant age-related TDP-43 encephalopathy (LATE) and frontotemporal dementia (FTD). The aim of the present study is to investigate in vivo LC and NBM integrity in patients with suspected-LATE, early-amnestic AD and FTD in comparison with controls. METHODS: Seventy-two participants (23 early amnestic-AD patients, 17 suspected-LATE, 17 FTD patients, defined by a clinical-biological diagnosis reinforced by amyloid and tau PET imaging, and 15 controls) underwent neuropsychological assessment and 3T brain MRI. We analyzed the locus coeruleus signal intensity (LC-I) and the NBM volume as well as their relation with cognition and with medial temporal/cortical atrophy. RESULTS: We found significantly lower LC-I and NBM volume in amnestic-AD and suspected-LATE in comparison with controls. In FTD, we also observed lower NBM volume but a slightly less marked alteration of the LC-I, independently of the temporal or frontal phenotype. NBM volume was correlated with the global cognitive efficiency in AD patients. Strong correlations were found between NBM volume and that of medial temporal structures, particularly the amygdala in both AD and FTD patients. CONCLUSIONS: The alteration of LC and NBM in amnestic-AD, presumed-LATE and FTD suggests a common vulnerability of these structures to different proteinopathies. Targeting the noradrenergic and cholinergic systems could be effective therapeutic strategies in LATE and FTD.

Regional nigral neuromelanin degeneration in asymptomatic leucine-rich repeat kinase 2 gene carrier using MRI.

Asymptomatic Leucine-Rich Repeat Kinase 2 Gene (LRRK2) carriers are at risk for developing Parkinson's disease (PD). We studied presymptomatic substantia nigra pars compacta (SNc) regional neurodegeneration in asymptomatic LRRK2 carriers compared to idiopathic PD patients using neuromelanin-sensitive MRI technique (NM-MRI). Fifteen asymptomatic LRRK2 carriers, 22 idiopathic PD patients, and 30 healthy controls (HCs) were scanned using NM-MRI. We computed volume and contrast-to-noise ratio (CNR) derived from the whole SNc and the sensorimotor, associative, and limbic SNc regions. An analysis of covariance was performed to explore the differences of whole and regional NM-MRI values among the groups while controlling the effect of age and sex. In whole SNc, LRRK2 had significantly lower CNR than HCs but non-significantly higher volume and CNR than PD patients, and PD patients significantly lower volume and CNR compared to HCs. Inside SNc regions, there were significant group effects for CNR in all regions and for volumes in the associative region, with a trend in the sensorimotor region but no significant changes in the limbic region. PD had reduced volume and CNR in all regions compared to HCs. Asymptomatic LRRK2 carriers showed globally decreased SNc volume and CNR suggesting early nigral neurodegeneration in these subjects at risk of developing PD.

Sustained reduction of essential tremor with low-power non-thermal transcranial focused ultrasound stimulations in humans.

BACKGROUND: Transcranial ultrasound stimulation (TUS) is a non-invasive brain stimulation technique; when skull aberrations are compensated for, this technique allows, with millimetric accuracy, circumvention of the invasive surgical procedure associated with deep brain stimulation (DBS) and the limited spatial specificity of transcranial magnetic stimulation. OBJECTIVE: /hypothesis: We hypothesize that MR-guided low-power TUS can induce a sustained decrease of tremor power in patients suffering from medically refractive essential tremor. METHODS: The dominant hand only was targeted, and two anatomical sites were sonicated in this exploratory study: the ventral intermediate nucleus of the thalamus (VIM) and the dentato-rubro-thalamic tract (DRT). Patients (N = 9) were equipped with MR-compatible accelerometers attached to their hands to monitor their tremor in real-time during TUS. RESULTS: VIM neurostimulations followed by a low-duty cycle (5 %) DRT stimulation induced a substantial decrease in the tremor power in four patients, with a minimum of 89.9 % reduction when compared with the baseline power a few minutes after the DRT stimulation. The only patient stimulated in the VIM only and with a low duty cycle (5 %) also experienced a sustained reduction of the tremor (up to 93.4 %). Four patients (N = 4) did not respond. The temperature at target was 37.2 ± 1.4 °C compared to 36.8 ± 1.4 °C for a 3 cm away control point. CONCLUSIONS: MR-guided low power TUS can induce a substantial and sustained decrease of tremor power. Follow-up studies need to be conducted to reproduce the effect and better to understand the variability of the response amongst patients. MR thermometry during neurostimulations showed no significant thermal rise, supporting a mechanical effect.

Isolated REM Sleep without Atonia Is Not Equivalent to REM Sleep Behavior Disorder in Early-Stage Parkinson's Disease.

BACKGROUND: In early-stage Parkinson's disease (PD), rapid eye movement (REM) sleep behavior disorder (RBD) predicts poor cognitive and motor outcome. However, the baseline significance and disease evolution associated with isolated REM sleep without atonia (iRWA, ie, enhanced muscle tone during 8.7% of REM sleep, but no violent behavior) are not well understood. OBJECTIVES: The objective is to determine whether iRWA was a mild form of RBD and progressed similarly over time. METHODS: Participants with early PD (<4 years from medical diagnosis) were included from 2014 to 2021 in a longitudinal study. They underwent interviews and examinations in the motor, cognitive, autonomous, psychiatric, sensory, and sleep domains every year for 4 years along with a video polysomnography and magnetic resonance imaging examination of the locus coeruleus/subcoeruleus complex (LC/LsC) at baseline. The clinical characteristics were compared between groups with normal REM sleep, with iRWA and with RBD, at baseline and for 4 years. RESULTS: Among 159 PD participants, 25% had RBD, 25% had iRWA, and 50% had normal REM sleep. At baseline, the non-motor symptoms were less prevalent and the LC/LsC signal intensity was more intense in participants with iRWA than with RBD. Over 4 years, participants with normal REM sleep and with iRWA had a similar cognitive and motor trajectory, whereas participants with RBD had greater cognitive and motor decline. CONCLUSIONS: We demonstrated that iRWA is frequent in early PD, but is not a milder form of RBD. Both groups have distinct baseline characteristics and clinical trajectories. They should be distinguished in clinical routine and research protocols. © 2024 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Contribution of MRI for the Early Diagnosis of Parkinsonism in Patients with Diagnostic Uncertainty.

BACKGROUND: International clinical criteria are the reference for the diagnosis of degenerative parkinsonism in clinical research, but they may lack sensitivity and specificity in the early stages. OBJECTIVES: To determine whether magnetic resonance imaging (MRI) analysis, through visual reading or machine-learning approaches, improves diagnostic accuracy compared with clinical diagnosis at an early stage in patients referred for suspected degenerative parkinsonism. MATERIALS: Patients with initial diagnostic uncertainty between Parkinson's disease (PD), progressive supranuclear palsy (PSP), and multisystem atrophy (MSA), with brain MRI performed at the initial visit (V1) and available 2-year follow-up (V2), were included. We evaluated the accuracy of the diagnosis established based on: (1) the international clinical diagnostic criteria for PD, PSP, and MSA at V1 ("Clin1"); (2) MRI visual reading blinded to the clinical diagnosis ("MRI"); (3) both MRI visual reading and clinical criteria at V1 ("MRI and Clin1"), and (4) a machine-learning algorithm ("Algorithm"). The gold standard diagnosis was established by expert consensus after a 2-year follow-up. RESULTS: We recruited 113 patients (53 with PD, 31 with PSP, and 29 with MSA). Considering the whole population, compared with clinical criteria at the initial visit ("Clin1": balanced accuracy, 66.2%), MRI visual reading showed a diagnostic gain of 14.3% ("MRI": 80.5%; P = 0.01), increasing to 19.2% when combined with the clinical diagnosis at the initial visit ("MRI and Clin1": 85.4%; P < 0.0001). The algorithm achieved a diagnostic gain of 9.9% ("Algorithm": 76.1%; P = 0.08). CONCLUSION: Our study shows the use of MRI analysis, whether by visual reading or machine-learning methods, for early differentiation of parkinsonism. © 2024 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Sleep disorders in Parkinson's disease, an early and multiple problem.

In Parkinson's disease (PD), it remains unclear whether sleep disorders including insomnia, REM sleep behavior disorder (RBD), excessive daytime sleepiness (EDS), restless legs syndrome (RLS) and sleep-disordered breathing (SDB), are isolated or combined, interact with each other and are associated with clinical factors. We sought to determine the prevalence and combinations of the main sleep disorders, and their clinical and polysomnographic associations in early stage PD. Sleep disorders were systematically diagnosed after medical interview and video-polysomnography in 162 participants with early stage PD and 58 healthy controls from the baseline of the longitudinal ICEBERG cohort. Demographic, clinical (motor, cognitive, autonomic, psychological and sensory tests), therapeutic and polysomnographic associations of sleep disorders were investigated. Sleep disorders were frequent (71%) and combined in half of the patients. The number of sleep disorders increased with disease duration and dysautonomia. Insomnia was the most common (41%), followed by definite RBD (25%), EDS (25%), and RLS (16%). These disorders were more frequent than in controls whereas SDB was rare, moderate and similar in both groups. In patients, insomnia (mainly difficulties maintaining sleep) was associated with female gender, shorter sleep time and RLS, but not with motor or psychological symptoms. RBD was associated with dysautonomia and advanced age, but not with motor and cognitive measures. EDS was associated with psychiatric and motor symptoms as well as the sedative effects of dopamine agonists but not with other sleep disturbances. Sleep disturbances are frequent and combined in early patients with PD. Their determinants and markers are more organic than psychological.