, University Hospital Basel and University of Basel
Translational Imaging in Neurology (ThINk) Basel, Department of Medicine and Biomedical Engineering

Author Of 2 Presentations

Biomarkers and Bioinformatics Oral Presentation

PS09.05 - Value of serum neurofilament light chain levels as a biomarker of suboptimal treatment response in MS clinical practice

Abstract

Background

Serum neurofilament light chain (sNfL) reflects neuro-axonal damage and may qualify as a biomarker of suboptimal response to disease modifying therapy (DMT).

Objectives

To investigate the predictive value of sNfL in clinically isolated syndrome (CIS) and relapsing-remitting (RR) MS patients with established DMT for future MS disease activity in the Swiss MS Cohort Study.

Methods

All patients were on DMT for at least 3 months. sNfL was measured 6 or 12-monthly with the NF-light®assay. The association between sNfL and age was modeled using a generalized additive model for location scale and shape. Z-scores (sNfLz) were derived thereof, reflecting the deviation of a patient sNfL value from the mean value of same age healthy controls (n=8865 samples). We used univariable mixed logistic regression models to investigate the association between sNfLz and the occurrence of clinical events (relapses, EDSS worsening [≥1.5 steps if EDSS 0; ≥1.0 if 1.0-5.5 or ≥0.5 if >5.5] in the following year in all patients, and in those fulfilling NEDA-3 criteria (no relapses, EDSS worsening, contrast enhancing or new/enlarging T2 lesions in brain MRI, based on previous year). We combined sNfLz with clinical and MRI measures of MS disease activity in the previous year (EDA-3) in a multivariable mixed logistic regression model for predicting clinical events in the following year.

Results

sNfL was measured in 1062 patients with 5192 longitudinal samples (median age 39.7 yrs; EDSS 2.0; 4.1% CIS, 95.9% RRMS; median follow-up 5 yrs). sNfLz predicted clinical events in the following year (OR 1.21 [95%CI 1.11-1.36], p<0.001, n=4624). This effect increased in magnitude with increasing sNfLz (sNfLz >1: OR 1.41 [95%CI 1.15-1.73], p=0.001; >1.5: OR 1.80 [95%CI 1.43-2.28], p<0.001; >2: OR 2.33 [95%CI 1.74-3.14], p<0.001). Similar results were found for the prediction of future new/enlarging T2 lesions and brain volume loss. In the multivariable model, new/enlarging T2 lesions (OR 1.88 [95%CI 1.13-3.12], p=0.016) and sNfLz>1.5 (OR 2.18 [95%CI 1.21-3.90], p=0.009) predicted future clinical events (n=853), while previous EDSS worsening, previous relapses and current contrast enhancement did not. In NEDA-3 patients, change of sNfLz (per standard deviation) was associated with a 37% increased risk of clinical events in the subsequent year (OR 1.37 [95%CI 1.04-1.78], p=0.025, n=587).

Conclusions

Our data support the value of sNfL levels, beyond the NEDA3 concept, for treatment monitoring in MS clinical practice.

Collapse
Machine Learning/Network Science Oral Presentation

PS16.04 - RimNet: A deep 3D multimodal MRI architecture for paramagnetic rim lesions assessment in multiple sclerosis

Speakers
Presentation Number
PS16.04
Presentation Topic
Machine Learning/Network Science
Lecture Time
13:27 - 13:39

Abstract

Background

In multiple sclerosis (MS), perilesional chronic inflammation appears on in vivo 3T susceptibility-based magnetic resonance imaging (MRI) as non-gadolinium-enhancing paramagnetic rim lesions (PRL). A higher PRL burden has been recently associated with a more aggressive disease course. The visual detection of PRL by experts is time-consuming and can be subjective.

Objectives

To develop a multimodal convolutional neural network (CNN) capable of automatically detecting PRL on 3D-T2*w-EPI unwrapped phase and 3D-T2w-FLAIR images.

Methods

124 MS cases (87 relapsing remitting MS, 16 primary progressive MS and 21 secondary progressive MS) underwent 3T MRI (MAGNETOM Prisma and MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany). Two neurologists visually inspected FLAIR magnitude and EPI phase images and annotated 462 PRL. 4857 lesions detected by an automatic segmentation (La Rosa et al. 2019) without overlap with PRL were considered non-PRL. The prototype RimNet was built upon two single CNNs, each fed with 3D patches centered on candidate lesions in phase and FLAIR images, respectively. A two-step feature-map fusion, initially after the first convolutional block and then before the fully connected layers, enhances the extraction of low and high-level multimodal features. For comparison, two unimodal CNNs were trained with phase and FLAIR images. The areas under the ROC curve (AUC) were used for evaluation (DeLong et al. 1988). The operating point was set at a lesion-wise specificity of 0.95. The patient-wise assessment was conducted by using a clinically relevant threshold of four rim+ lesions per patient (Absinta et al. 2019).

Results

RimNet (AUC=0.943) outperformed the phase and FLAIR image unimodal networks (AUC=0.913 and 0.855, respectively, P’s <0.0001). At the operating point, RimNet showed higher lesion-wise sensitivity (70.6%) than the unimodal phase network (62.1%), but lower than the experts (77.7%). At the patient level, RimNet performed with sensitivity of 86.8% and specificity of 90.7%. Individual expert ratings yielded averaged sensitivity and specificity values of 76.3% and 99.4%, respectively.

Conclusions

The excellent performance of RimNet supports its further development as an assessment tool to automatically detect PRL in MS. Interestingly, the unimodal FLAIR network performed reasonably well despite the absence of a paramagnetic rim, suggesting that morphometric features such as volume or shape might be a distinguishable feature of PRL.

Collapse

Author Of 7 Presentations

Machine Learning/Network Science Late Breaking Abstracts

LB1213 - Attention-based deep learning identifies a new microstructural diffusion MRI contrast sensitive to focal pathology and related to patient disability (ID 2074)

Speakers
Presentation Number
LB1213
Presentation Topic
Machine Learning/Network Science

Abstract

Background

Microstructural biophysical models reconstructed from advanced diffusion MRI (dMRI) data provide quantitative measures (qMs), which inform about the brain tissue microenvironment, based on different assumptions.

Objectives

To compare the sensitivity of available qMs to focal pathology in multiple sclerosis (MS), and to explore which qMs– or combinations of qMs – are best correlated with patients disability.

Methods

dMRI (1.8 mm isotropic resolution, 149 directions, b-values were 0, 700, 1000, 2000, 3000 s/mm2) was acquired from 67 relapsing-remitting and 33 progressive MS patients (median EDSS: 2.5). The qMs for the isotropic and intra-axonal compartments were derived from the following available models: Ball and Stick, NODDI, SMT-NODDI, MCMDI, NODDIDA, DIAMOND, Microstructure Bayesian approach (MB) and microstructure fingerprinting. In total, 13 qMs were included and subject-wise normalized within brain tissue (nqMs).

To identify the nqMs sensitive to focal pathology, an attention-based convolutional neural network (aCNN) was built to (a) classify randomly sampled WM lesion and perilesional WM patches and (b) generate attention weights (AWs) representing the relative importance of the qMs in the classification. Twenty patients were randomly selected in the test dataset (709 lesion patches and 746 perilesional WM patches), and the rest were in the cross-validation (CV) dataset (2925 lesion patches and 3176 perilesional WM patches). The performance metric was the area under the receiver operating characteristic curve (AUC). Because of the correlation between the nqMs, which may influence the relative AWs, we performed 10-fold CV and selected the nqMS that most contributed to the classification.

To assess which nqMS – or combination of nqMS was best correlated with EDSS, we used Spearman’s correlation coefficient (ρ) with two-sided 20000 permutation tests and followed by Bonferroni correction.

Results

The test AUC was 0.911 indicating the aCNN learned the right AWs to differentiate lesions and perilesional WM. The most discriminating nqMs included isotropic and intra-axonal compartments from MB, the neural density index (NDI) from the NODDI and the intra-axonal compartment from MCMDI.

The sum of isotropic and intra-axonal compartments of the MB (sMB) showed the strongest correlation with EDSS (ρ=-0.40,corr. p<0.0001) followed by the sum of sMB and NDI (ρ=-0.30,corr. p<0.05), and the sum of sMB and intra-axonal compartment from MCMDI (ρ=-0.32,corr. p<0.05). None of the selected nqMs as a single measure and their other combinations correlated with EDSS.

Conclusions

By performing aCNN-aided selection of the openly available WM quantitative measures, we have identified the measures most sensitive to MS focal pathology; furthermore, we have derived a new contrast that – by combining the measures of isotropic and intracellular diffusion – strongly correlated with patients’ disability.

Collapse
Biomarkers and Bioinformatics Poster Presentation

P0097 - Intrathecal immunoglobulin M synthesis is associated with higher serum neurofilament light chain levels and increased MRI disease activity in MS (ID 1089)

Abstract

Background

Intrathecal IgM synthesis was reported to be associated with higher clinical disease activity and severity. We found an association also with earlier use of high efficacy treatments in relapsing MS (RMS).

Objectives

To explore whether patients with intrathecal IgM synthesis show a) higher serum neurofilament light chain levels (sNfL) as a reflection of neuronal damage, or b) signs of increased disease severity in cerebral MRI, in patients with RMS followed in the Swiss MS Cohort Study.

Methods

487 patients were categorized by presence of oligoclonal IgG bands (OCGB) and intrathecally produced IgG/M:

1) OCGB-/IgG-/IgM- (reference [ref]);

2) OCGB+/IgG-/IgM-;

3) OCGB+/IgG+/IgM- and

4) OCGB+/IgG+/IgM+.

sNfL was measured (at baseline and every 6- or 12 months) with the NF-light® assay. Age-dependent sNfL z-scores (sNfLz) were modelled in 8865 healthy control samples to reflect the deviation of a patient sNfL value compared to mean values observed in same age healthy controls. Yearly T2 lesion number and occurrence of new/enlarging T2 lesions were automatically assessed in cerebral MRIs and checked manually. Contrast enhancing lesions (CEL) were manually quantified. Linear or negative binomial mixed models were used to investigate the associations between the four CSF Ig patterns and longitudinal sNfLz and MRI measures, adjusted for DMT and other covariates.

Results

IgM+ patients had higher sNfLz vs reference (estimate 0.50 [CI 0.12, 0.89], p=0.011), whereas those with only OCGB+ (0.11 [-0.28, 0.50], p=0.582) or with OCGB+/IgG+ (0.20 [-0.16, 0.56], p=0.270) did not (n=2970 observations). This was confirmed when analyzing only untreated patients adjusting for T2 and CEL numbers (1.16 [0.47, 1.86], p<0.01 vs 0.58 [-0.11, 1.27], p=0.1022 vs 0.51 [-0.11, 1.13], p=0.108 vs ref, respectively) (n=234).

IgM+ patients had 2.28-fold more T2 lesions ([1.51, 3.44], p<0.01) vs ref; for patients with only OCGB+ (1.61 [1.07, 2.43], p=0.0237) or OCGB+/IgG+ (1.58 [CI 1.08, 2.32], p=0.0179) (n=1580) this association was weaker.

IgM+ was associated with a 2.47-fold risk for new/enlarging T2 lesions on yearly follow-up MRIs vs ref (2.47 [1.28, 4.78], p<0.01) but not the two other patient groups (1.84 [CI 0.93; 3.65], p=0.0799 and 1.61 [CI 0.87; 2.95], p=0.1280) (n=861).

Conclusions

Intrathecal IgM synthesis was consistently associated with quantitative measures of neuro-axonal injury and disease severity in RMS. Our findings strongly support the clinical utiliy of this biomarker.

Collapse
Biomarkers and Bioinformatics Poster Presentation

P0160 - Serum NfL z-scores derived from a large healthy control group reflect different levels of treatment effect in a real-world setting (ID 916)

Abstract

Background

Serum neurofilament light chain (sNfL) levels reflect neuroaxonal damage and relate to disease activity in MS. sNfL may qualify as well as a biomarker of suboptimal treatment response to disease modifying therapies (DMT). Establishment of age-dependent reference ranges in healthy controls is a prerequisite for developing this biomarker for clinical use.

Objectives

To compare on-treatment sNfL levels with values from a healthy control cohort and to investigate the effect of DMTs on sNfL levels in patients from the Swiss MS Cohort Study.

Methods

sNfL was measured (at baseline and every 6- or 12 months) with the NF-light® assay. Age-dependent sNfL z-scores (sNfLz) were modeled in healthy controls using a generalized additive model for location scale and shape to reflect the deviation of a patient sNfL value from the mean value of same age healthy controls. Linear mixed models were used to investigate the associations between clinical characteristics, DMT and longitudinal sNfLz. Interaction terms and splines were used to model sNfLz and for comparison log(NfL), and their dynamics under treatment.

Results

sNfL was measured in 1368 patients with 7550 longitudinal samples (baseline: median age: 41.9 yrs; 5.4% CIS, 83.2% RRMS, 5.6% SPMS, 5.8% PPMS; median EDSS: 2.0; median follow-up: 4.6 yrs) and 4133 healthy controls with 8865 samples (median age: 44.8 yrs). In the multivariable model, sNfLz increased with EDSS (0.131/step, [95% CI 0.101;0.161]), recent (<120 days) relapse (0.739 [0.643;0.835]) decreased with age (-0.014/year [-0.02;-0.009]), and time on DMT (-0.040/year [-0.054;-0.027]); sNfLz were lower when sampled while on more effective DMT (oral versus platform injectables: -0.229 [-0.344;-0.144]; monoclonal antibodies (mAB) versus platform injectables: -0.349 [-0.475;-0.224]), (p<0.001 for all associations). sNfLz were inversely associated with the hierarchy in efficacy of mAB over orals and orals over platform therapies with regard to slope and extent of decrease (interaction between time under DMT and DMT class: p<0.001). sNfLz, but not log(NfL) showed normalization of sNfL levels by mAB to healthy control levels.

Conclusions

The dynamic change of sNfLz on DMT reflects closely their relative clinical efficacy and is more meaningful than log(sNfL) by excluding age as a confounding factor. Use of sNfLz based on a large normative database as an age-independent sNfL measure improves the accuracy of the sNfL signal and hence their clinical utility.

Collapse
Imaging Poster Presentation

P0538 - Applying advanced diffusion MRI in MS: a comparison of 20 diffusion MRI models to identify microstructural features of focal damage (ID 1338)

Speakers
Presentation Number
P0538
Presentation Topic
Imaging

Abstract

Background

Advanced diffusion-weighted MRI (DW-MRI) sequences, in combination with biophysical models, provide unprecedented information on the microstructural properties of both healthy and pathological brain tissue.

Nevertheless, it is nowadays challenging to identify the most accurate biophysical model to describe focal microstructural pathology in multiple sclerosis (MS) patients, due to the lack of appropriate comparative studies.

Objectives

To investigate the specificity and sensitivity of 124 independent features derived from 20 diffusion microstructural models to differentiate specific features of tissue alterations in white matter (WM) lesions compared to the surrounding normal-appearing WM (NAWM).

Methods

The study included 102 MS patients: RRMS: 66%, SPMS: 18%, PPMS: 16%, mean age 46±14; female 64%, disease duration 12.16±18.18 years, median Expanded Disability Status Scale (EDSS): 2.5.

DW-MRI data were acquired with 1.8mm isotropic resolution isotropic and with the b-values [0, 700, 1000, 2000, 3000] s/mm2.

Lesion masks were generated with a deep learning network algorithm and manually corrected if required. Voxels of NAWM tissue were randomly chosen outside the lesion masks.

The following microstructural models were applied: DTI, Non-parametric DTI, DKI, Ball and Stick, Ball and Sticks, Ball and Rockets, NODDI-Watson, AMICO-NODDI, NODDI-Bingham, SMT-NODDI, NODDIDA, SMT, MCMDI, CHARMED, IVIM, sIVIM, Microstructure Fingerprinting, Microstructure Bayesian, DIAMOND, and DIAMOND isotropic-restricted.

The classification was performed using logistic regression on 300’000 voxels, equally divided in lesion and NAWM voxels. Features were scored according to the Area Under the Curve (AUC), sensitivity, and specificity.

Results

The intra-axonal signal fraction of the Microstructure Bayesian approach scored maximum with AUC=0.87, for threshold=0.5 sensitivity=0.79, sensitivity=0.83. AUC = 0.86 were attributed to the intra-axonal signal fraction of Ball and rockets, NODDI-Watson, AMICO-NODDI, NODDI-Bingham, SMT-NODDI and the extra-axonal perpendicular signal fraction of the Microstructure Bayesian approach. Low AUC scores (<0.75) were achieved by DTI and parameters not related to signal fractions, e.g. orientation dispersion.

Conclusions

Among available microstructural models, the Microstructure Bayesian appeared to best differentiate voxels with microstructural damage in WM lesions compared to NAWM. Very similar, albeit slightly lower accuracy, was achieved by NODDI-based models. In general, models with estimates intra-axonal signal fraction tend to perform better in this type of classification, showing that intra-axonal component may be the dominant factor in distinguishing the two types of tissue. Further analysis will explore the advantage of including combinations of independent features.

Collapse
Imaging Poster Presentation

P0545 - Automatic MS lesions segmentation using LeMan-PV as a clinical decision-support tool: a longitudinal analysis (ID 1590)

Abstract

Background

LeMan-PV is a prototype that performs cross-sectional and longitudinal detection of Multiple Sclerosis (MS) lesions, which has been validated on conventional (cMRI) and advanced magnetic resonance imaging at 3T (Fartaria et al. 2019). Since this software provides a report that is available shortly after image acquisition, it may be ideal as clinical decision-support tool.

Objectives

To assess LeMan-PV as clinical decision-support tool in a monocentric real-world cMRI dataset from the Swiss Multiple Sclerosis Cohort.

Methods

262 MS patients underwent cMRI at Basel University Hospital in a mean of 3.5 follow-up sessions, with an average of 399 days between two consecutive sessions. cMRI sequences were acquired at 1.5T and 3T in 725 and 195 sessions, respectively. Cross-sectional and longitudinal MS lesions segmentation (i.e. identification of new and enlarging lesions - NLs, ELs) was performed using the LeMAN-PV prototype software. An expert neuroradiologist performed a radiological reading of the number of NLs and ELs in the most recent acquisition by comparing it to the previous one (ground truth, GT), considering only lesions with a diameter larger than 3 mm. The minimum volume thresholds to identify an NL and an EL were chosen by minimizing the patient-wise error between the automated count and the expert ground truth. Two scenarios were evaluated by first assuming disease activity if one or more EL were present, and second by considering activity if NL were present in the new acquisition.

Results

The volume thresholds chosen were 11 and 12 mm3 for ELs and NLs, respectively. For those, LeMan-PV detected 11% more of both ELs and NLs than the neuroradiologist. In the patient-wise evaluation of cases with both sessions acquired at 1.5T (70%), LeMan-PV showed sensitivities of 93% and 78% and specificities of 62% and 43% when evaluating ELs and NLs. For the 3T pairs of sessions (8%), values were 68% and 72% for ELs and 73% and 68% for NLs. Finally, for cases with a first acquisition at 1.5T and a second at 3T (22%), values were 76% and 73% for ELs and 71% and 65% for NLs.

Conclusions

The count of new and enlarging MS lesions using LeMan-PV were close to the one performed by an expert neuroradiologist; the software performed better when assessing disease activity via detection of enlarging lesions rather than by identifying new lesions. More 3T data is being currently collected at 3T to provide a size-matched inter-scanner comparison.

Collapse
Imaging Poster Presentation

P0580 - Focal inflammatory activity and lesion repair are associated with brain atrophy rates in MS patients (ID 1092)

Abstract

Background

The pathogenesis of neurodegeneration in multiple sclerosis (MS) is multifactorial and the determinants of brain atrophy rates are not completely understood.

Objectives

To investigate the association between annualized atrophy rate (AAR) of multiple brain measures (regional cortical thickness (CTh), volumes of basal ganglia, thalamus, white matter, gray matter, brain and brain parenchymal fraction (BPF)) and: (1) annualized rate of new and enlarging white matter lesions (WMLs); (2) annualized rate of resolved WMLs; (3) occurrence of progression independent of relapse activity (PIRA) during follow-up.

Methods

We included 1573 1.5T or 3T brain MRI scans from 378 patients of the Swiss MS Cohort Study (331 relapsing-remitting MS (RRMS), 27 clinically isolated syndrome (CIS), 11 secondary-progressive MS (SPMS), 9 primary-progressive MS (PPMS); 70% female; median age: 41.9 yrs; disease duration: 8.3 yrs; EDSS: 2.0; follow-up time: 4.0 yrs). Longitudinal changes in WMLs were obtained using an automated prototype (LeMan-PV). Brain volumes and CTh AARs were obtained using FreeSurfer longitudinal pipeline (v6.0) after WMLs filling. In patients fulfilling PIRA an EDSS progression had to be confirmed ≥6 months after the index event. Multivariable generalized linear models were used to model the association between AAR (dependent variable) and independent variables (1-3), correcting for age, sex, disease duration and baseline EDSS. p-values were adjusted for Bonferroni multiple comparison correction; for vertex-wise CTh analysis, Monte Carlo Z simulation was performed (cluster threshold p<0.05).

Results

We found positive associations between annualized rate of new and enlarging WMLs and (i) CTh AAR of 8 extensive clusters (bilateral frontal, temporal and occipital regions and right insula, all p<0.01) and (ii) AAR of: caudate bilaterally (p=0.02), white matter volume, brain volume and BPF (p<0.001 for all).

We also found a negative association between annualized rate of resolved WMLs and CTh AAR in 3 cortical clusters (right insula, precentral area and anterior cingulate region, all p<0.05); no associations with AAR of volumes emerged.

57 patients fulfilled PIRA whereas 295 experienced no EDSS progression events: no significant differences in AAR measures were found between these two groups.

Conclusions

In a large cohort of MS patients, with a median follow-up of 4 years, local radiological inflammatory and reparative activity were associated with AAR in multiple brain regions. PIRA did not seem to be related to increased AAR in any of the regions studied.

Collapse
Imaging Poster Presentation

P0638 - Role of Gadolinium-based contrast agents to detect subclinical disease activity in clinically stable patients in the Swiss MS Cohort Study (ID 821)

Abstract

Background

Gadolinium (Gd)-based contrast agents are widely used to assess disease activity and treatment response by MRI in multiple sclerosis (MS). There is, however, increasing concern about their safety as their repeated administration may lead to brain parenchymal accumulation, while preclinical models suggest that they induce mitochondrial toxicity and neuronal cell death. Moreover, recent reports have demonstrated that three-dimensional (3D) T2-weighted Fluid-Attenuated-Inversion-Recovery (FLAIR) is highly sensitive in detecting new or enlarging MS lesions.

Objectives

To explore whether the presence of contrast enhancing lesions (CEL) based on Gd injection is more sensitive in detecting lesional activity in clinically stable MS patients in comparison to the analysis of new or enlarging MS lesions by 3D FLAIR.

Methods

MS patients being part of the observational, multicenter Swiss Multiple Sclerosis Cohort Study (SMSC) with contrast enhanced T1-weighted (T1w) images were included. Clinical stability was defined as no relapse and no Expanded Disability Status Scale (EDSS) increase during at least twelve months prior to MRI. Presence of CEL was assessed on contrast enhanced T1w images. Presence of new or enlarging T2w lesions was assessed manually on 3D FLAIR in an independent analysis by a different investigator in clinically stable MS patients presenting with CEL.

Results

3930 MRI scans (3.0 Tesla n=1497 (38%)) in 1057 participants (685 women, median age 42.0 years, 941 with relapsing MS, 116 with progressive MS, median EDSS 2.0 (range 1.5-3.5), median disease duration 7.4 years) were included.

Of 2620 MRI scans (66.7%) acquired in clinically stable conditions 46 (1.8%) demonstrated CEL. In all of these, new or enlarging T2w lesions were detectable by 3D FLAIR when a previous MRI was available for comparison (previous MRI available in 29/46; median number of new or enlarging T2w lesions: 3 (range 1-41, total number 176); median number of CEL: 1 (range 1-4, total number 47)).

Conclusions

In our large cohort from clinical practice, the assessment of new or enlarging lesions by 3D FLAIR was equally sensitive as the quantification of CEL to detect disease activity in clinically stable MS patients, challenging current practice of the use of Gd-enhanced MRI for monitoring of MS in clinical routine.

Collapse