Postmortem magnetic resonance imaging (MRI) of formalin-fixed healthy and diseased human brains with ultra-high spatial resolution has the great potential to depict tissue architecture in fine detail, allowing a deeper understanding of pathological processes. Whole-brain imaging is important since it provides neuroanatomic relationships, reference points across distant brain regions, and a comprehensive view of pathologies affecting the brain. However, ultra-high-resolution whole-brain postmortem MRI is challenging and has been so far almost exclusively performed at 7T with specialized hardware.
To develop a 3D isotropic 160µm ultra-high-resolution imaging (URI) approach for human whole-brain ex vivo acquisitions on a standard clinical 3T MRI system. To explore the sensitivity and specificity of the approach to specific pathological features of multiple sclerosis (MS).
A fixed whole human brain from a patient with secondary progressive MS was investigated. Acquisitions were performed on a clinical 3T Siemens Prismafit MRI system with standard hardware components. URI is based on a gradient echo sequence similar to the 7T approach by Edlow et al. 2019. However, it allows to acquire an isotropic 160µm resolution with low hardware demands and to directly reconstruct the image data on the standard 3T MRI system. URI images display a strong, susceptibility-enhanced tissue contrast.
The reconstructed URI images depicted with remarkable quality the diseased human MS brain at 3T field strength. URI allowed to distinguish fine anatomical details such as the subpial molecular layer, the stria of Gennari as well as some intrathalamic nuclei. Additionally, because of the unprecedented spatial resolution and contrast at 3T, URI permitted to easily identify the presence of subpial lesions, detailed features of intracortical lesions such the presence of incomplete/complete iron rims or patterns of iron accumulation in the entire lesion core in both cortical and white matter lesions (CLs/WMLs), lesions affecting the convoluted layers of the cerebellar cortex and nascent submillimetric CLs/WMLs.
URI provides a comprehensive microscopic insight into the whole-human brain at 3T through the micrometric resolution and a tissue-specific, susceptibility-enhanced contrast. We propose URI as an excellent approach to investigate microscopic brain changes of complex pathologies like MS.
Previously, we used frequency domain near-infrared spectroscopy (fdNIRS), a non-invasive imaging modality, to show that brain hypoxia exists in a subset of MS patients. Currently, there is limited knowledge on the effects of hypoxia in MS. However, some studies suggest that hypoxia may exacerbate inflammation of the central nervous system (CNS). It is important to elucidate the time-course of hypoxia in relation to inflammation to further understand its role in MS.
The aim of the present study was to use fdNIRS to determine if hypoxia in MS resolves quickly or if it is a chronic condition.
We used fdNIRS to quantify cortical microvasculature hemoglobin saturation (StO2) in 55 controls and 85 MS patients. StO2 values that were 2 standard deviations (SD) below the control mean were defined as hypoxic (<55.7%). Arterial oxygen saturation (SaO2) was measured using a pulse oximeter to confirm that reduced StO2 was not systemic in origin. To determine whether the temporal pattern of StO2 relates to changes in acute CNS inflammation, we recruited a subset of MS patients (hypoxic: n=12; normoxic: n=7) for a longitudinal study. We measured StO2 once a week for 4 consecutive weeks, and then once a month for 5 subsequent months. Due to COVID-19-related lab closures, we were only able to obtain StO2 data for the first 8 weeks for 13 of these patients (hypoxic: n=8; normoxic: n=5).
StO2 in MS patients was significantly lower compared to controls (57.6±7.6% vs. 62.3±3.6%, respectively, p=0.002), with no differences in SaO2. For the longitudinal study, we found that StO2 values for normoxic and hypoxic MS groups did not change significantly over the course of 8 weeks (F (9, 36.9) =1.44, p=0.255).
To our knowledge, we are the first group to use fdNIRS to identify a subset of MS patients who experience persistent brain hypoxia. As hypoxia in MS patients persists beyond 4 weeks, we argue that it can present as a chronic condition. This indicates that, in these patients, physiological responses such as angiogenesis have not occurred or are not sufficient to result in resolution of hypoxia. With such chronic hypoxia we would predict that in these patients, some symptoms may be a result of this chronic hypoxia. Also, we argue that such chronic hypoxia could exacerbate and further stimulate a pathological immune response (a hypoxia inflammation cycle).
In multiple sclerosis (MS), thalamic integrity is affected both directly by demyelination, neuronal loss and increasing iron concentration, and indirectly by remote gray and white matter lesions affecting neural projections into and out of the thalamus. Thalamic atrophy may therefore reflect a large fraction of MS-related brain damage and thus represent a useful marker of overall damage and therapeutic efficacy.
To assess the efficacy of ocrelizumab (OCR) in patients switching to or maintaining OCR therapy on thalamic atrophy in patients with relapsing MS (RMS) and primary progressive MS (PPMS), participating in the OPERA I/II (NCT01247324/NCT01412333) and ORATORIO (NCT01194570) Phase III trials, respectively.
At the end of the double-blind controlled treatment period in OPERA I/II, patients entered the open‑label extension (OLE), and either continued to receive OCR (OCR-OCR) or switched from interferon β-1a (IFN β-1a) to OCR (IFN β-1a-OCR). In ORATORIO, patients entered the OLE ~3–9 months after the double-blind period cut-off and either continued OCR (OCR-OCR) or switched from placebo (PBO) to OCR (PBO-OCR). Changes in thalamic volume from the core trial baseline were computed using Jacobian integration and analyzed using a mixed-effect repeated measurement model, adjusted for baseline volume, age, baseline gadolinium-enhancing lesions (presence/absence), baseline T2 lesion volume, region (US vs rest of the world), Expanded Disability Status Scale category (<4, ≥4), week, treatment, treatment and time interaction, and treatment and baseline volume interaction.
In the OLE of OPERA I/II, changes (%) in thalamic volume from baseline at OLE Week 46, 94, 142, 190, and 238, were: –2.88/–2.12 (p<0.001), –3.31/–2.36 (p<0.001), –3.61/–2.78 (p<0.001), –3.68/–3.03 (p<0.001), and –4.07/–3.41 (p<0.001), for IFN β-1a-OCR/OCR-OCR patients, respectively. During the OLE of ORATORIO, changes in thalamic volume at OLE Day 1, Week 48, 96, and 144, were: –3.46/–2.44 (p<0.001), –3.93/–2.61 (p<0.001), –4.30/–3.25 (p<0.001), and –4.86/–3.62 (p<0.001), for PBO-OCR/OCR-OCR patients, respectively.
In the OLE, patients with RMS and PPMS who were initially randomized to ocrelizumab experienced less thalamic volume loss compared with those initiating ocrelizumab later.
Multiple-Sclerosis-Partners-Advancing-Technology-Health-Solutions (MSPATHS) is an international multicentre digital database that collects clinical information provided directly by patients together with standardized MRI and biomarkers.
We identify a Benign multiple sclerosis (BMS) population using Patient-Determined-Disease-Steps (PDDS) as a proxy for EDSS. We describe its physical and non-physical characteristics, and explore the features that best discriminate BMS.
Cross-sectional study of MSPATHS patients (Feb 2019). In patients with disease duration ≥10 years, BMS was considered when PDDS score<2. We compared BMS and non-BMS in terms of (1)socio-demographic and clinical characteristics, (2)physical status (lower and upper extremity function by Neuro-QoL (LUEF-NQ) and neurological performance tests: walking speed test (WST), manual dexterity test (MDT), processing speed test (PST), contrast sensitivity test (CST)) and non-physical symptoms (anxiety, depression, fatigue, among other NQ domains), and (3)MRI (gadolinium enhancement and new T2 lesions). We built a random forest model to estimate the importance of each variable. Cohen’s d was used for descriptive statistics to categorize differences in small (d=0.2-0.5), medium (d=0.5-0.8) and large (d>0.8). A sensitivity analysis with a 1:1 matched cohort by disease duration was performed.
From 15,257 patients included, 8,349 had a disease duration ≥10 years and 3,852 (46.1%) were classified as BMS. (1)BMS and non-BMS patients were similar for gender, age at disease onset and diagnosis, ethnicity, years of education and smoking status. Compared to non-BMS, BMS had small differences in disease duration (median, 17.2 (12,9-23,4) vs. 20.9 (15,1-28,8 years); d=0.39) but medium/large differences in (2)physical status (LUEF-NQ d=2.06 and 1.53, WST d=0.81, MDT d=0.97, PST d=0.82 and CST d=0.56), as well as, in all non-physical symptoms evaluated by NQ (anxiety d=0.53, depression d=0.69, fatigue d=0.84, stigma d=1.32, cognition d=0.69, social role satisfaction (SRS) d=1.11 and participation (SRP) d=1.19). (3)No differences were found on MRI activity. With 0.88 sensitivity and 0.86 specificity, LUEF-NQ was the most contributing variable for the random forest followed by stigma, SRP, WST, and SRS. The sensitivity analysis showed similar results.
PDDS seems to be a useful disability proxy to identify BMS when using digital technology. LUEF-NQ, stigma, SRP and SRS seem to better discriminate BMS.
A deep-learning prototype method, called CVSNet, was recently introduced for the automated detection of the central vein sign (CVS) in brain lesions and demonstrated effective and accurate discrimination of multiple sclerosis (MS) from its mimics. However, this method solely considered focal lesions displaying the central vein sign (CVS+) or not (CVS−), therefore requiring a manual pre-selection of the lesions to be evaluated by eliminating the so-called excluded lesions (CVSe) as defined by the NAIMS criteria. CVSe lesions may however play an important role in differential diagnosis. Moreover, extending the automated CVS classification to these lesions would facilitate the integration of CVSNet with existing MS lesion segmentation algorithms in a fully automated pipeline.
To develop an improved version of the CVSNet prototype method able to classify all types of lesions (CVS+, CVS− and CVSe).
Patients with an established MS or CIS diagnosis (RRMS 29; SPMS 10; PPMS 10; CIS 1; mean ± SD age: 50 ± 11 years; male/female: 23/27), and healthy controls (n=8; mean ± SD age: 41 ± 9 years; male/female: 5/3), underwent 3T brain MRI (MAGNETOM Skyra and MAGNETOM Prisma, Siemens Healthcare, Erlangen, Germany, or Achieva, Philips Healthcare, Best, Netherlands). Brain lesions were automatically segmented and manually corrected by a single rater. CVS assessment was conducted on FLAIR* images by two raters, according to the NAIMS guidelines, yielding 1542 CVS+, 1004 CVS−, and 1131 CVSe lesions. A convolutional neural network (CNN) based on the CVSnet architecture was trained with different configurations using 3021 samples (1261 CVS+, 847 CVS−, and 913 CVSe) and evaluated in 656 unseen samples (281 CVS+, 157 CVS−, and 218 CVSe, from 13 patients) for final testing. The configurations relied on different combinations of the following channels as input: (i) FLAIR*, (ii) T2*, (iii) lesion mask, and (iv) CSF and brain tissue concentration maps obtained from a partial-volume estimation algorithm. Lesion-wise classification performance was evaluated for the different configurations by estimating the sensitivity, specificity, and accuracy for each lesion class.
The results were similar across the different configurations. The best performance in the unseen testing set was obtained when all channels were used as input (sensitivity: 0.71, 0.73; specificity: 0.71, 0.81; and accuracy: 0.71, 0.79 for CVS+, CVS−, respectively). For CVSe, this approach achieved 0.52 sensitivity, 0.94 specificity, and 0.80 accuracy.
We introduced a modified CVSNet prototype method that can analyze the presence of the central vein for all types of brain lesions, enabling its integration with current MS lesion segmentation algorithms. This new feature will allow a fully automated assessment of the CVS in patients’ brains, speeding up the evaluation of CVS as a diagnostic biomarker for differentiating MS from mimicking diseases.