Background

In the constantly evolving field of MS, personalized medicine is still one of the most important unmet need that requires further attention

Objectives

We aimed to develop a dynamic risk calculator to predict the long-term prognosis of MS in the context of a large MS Centre in Catalonia

Methods

This is an observational study based on data prospectively acquired from a deeply phenotyped CIS cohort from Barcelona. We first built a natural history baseline risk score (BRS) for predicting moderate disability, integrating baseline prognostic factors: Sex, age at CIS, CIS topography, number of T2 lesions, contrast-enhancing lesions (CEL) and oligoclonal bands. This BRS was designed as follows: For untreated patients, we built a Weibull model to estimate the median time to confirmed EDSS 3.0 and with these estimates we identified risk groups based on the median of the cut-offs of 2000 survival trees. Then we obtained the BRS of the full cohort. In patients with more than ten years of follow-up, we performed an inverse probability weighting to balance patients during their follow up for the propensity of being treated or lost to follow-up. The weights were estimated via a proportional hazards (PH) Cox model considering both baseline information (CIS year, BRS) and time-dependent (diagnosis status, new T2 lesions, CEL and cumulative number of relapses). Finally, a weighted PH Cox model was built to estimate the time to confirmed EDSS 3.0 considering the BRS and time-dependent events (new T2 lesions, cumulative number of relapses and first or second-line treatment use). Sensitivity analyses using other disability outcomes and different follow-ups were conducted.

Results

Of 956 patients, 577 (60.4%) were untreated before confirmed EDSS 3.0. Two BRS were obtained: low and high-BRS. Of 400 patients followed for more than ten years, 226 (56.5%) were low-BRS and 174 (43.5%) were high-BRS. High-BRS showed a HR=2.16 95%CI (1.16,4.02). Each new T2 lesion presented HR=1.04 95%CI (1.00,1.08) and each new relapse HR=1.46 95%CI (1.23,1.74). Being on second-line treatment showed a protective effect (HR=0.23 95%CI (0.06,0.94)) but no association was found for first-line treatments (HR=1.32 95%CI (0.67,2.60). Sensitivity analyses confirmed the association between BRS, new T2 lesions and the accumulation of relapses with the prognosis. However, treatment results were inconclusive.

Conclusions

Presenting a high-BRS doubles the risk of reaching moderate disability. Each new lesion and new relapse increses the risk by 4% and 46%, respectively; and second-line treatments seem to be protective. If validated, this risk calculator could be a crucial step to personalized medicine.

Background

Synthetic MRI reduces acquisition time and could be an alternative to conventional sequences for the assessment of multiple sclerosis (MS).

Objectives

To perform a qualitative and quantitative comparison of conventional and synthetic MRI sequences to evaluate their value in the assessment of brain demyelinating lesions.

Methods

Twenty-seven RRMS patients (18 women), mean age of 44.0 years, median EDSS of 3.5 were examined in a 1.5T MRI scanner. A 2D QRAPMASTER sequence was added to the brain MRI protocol (proton density [PD] and T2w fast spin-echo, fast T2w FLAIR, and T1w spin-echo sequences). SyMRI software version 8.0.4 used QRAPMASTER images to generate synthetic images with the same TR, TE and TI used for conventional MRI.

Four raters performed a blinded qualitative analysis of the images in a random order to evaluate global image quality (GIQ), global image contrast, presence of flow artifacts in posterior fossa, contrast of lesions to white matter, and level of confidence for supratentorial and infratentorial lesion assessment. Moreover, the number of periventricular, juxtacortical, brainstem, and cerebellum lesions, and the contrast-to-noise ratio (CNR) between regions were evaluated.

Statistical analysis was performed in SPSS v. 25. Crosstabs were used to evaluate the degree of agreement between sequences for qualitative data. Wilcoxon signed rank test was used to evaluate differences for quantitative data.

Results

GIQ showed a predominance of better scores for conventional MRI. All other image quality parameters showed a degree of agreement similar or greater to the predominance of better scores for conventional MRI. There were no significant differences in the degree of agreement between pairs of raters in the assessment of conventional and synthetic MRI except between raters 1 and 2. However, we found a clear predominance of disagreement for all pairs of raters. Synthetic PD, T2w and T2w-FLAIR showed higher CNR than conventional sequences for most of the regions. Two raters found a greater number of brainstem lesions in conventional PD and one in synthetic T2w-FLAIR images. Three raters found a greater number of cerebellum lesions in conventional PD, and two of them in conventional T2w-FLAIR images.

Conclusions

Synthetic MRI obtained lower scores for some qualitative rater-related parameters while quantitative CNR data showed higher values. Synthetic MRI shows potential to be used as an alternative to conventional brain MRI sequences in the assessment of MS.

Background

New T2 lesions count is routinely used for assessing disease activity in multiple sclerosis (MS), although their visual detection is challenging (low sensitivity, high variability).

Objectives

We assessed two different automatic decision-support systems to detect new T2 lesions in longitudinal brain MRIs of patients with MS.

Methods

The study included 100 MS patients with two MRI exams (median interval 12 months [range 3-27 months]; relapse free 85%). MRI scans were acquired on a 3T magnet following a standardized protocol (3D-FLAIR, 3D- MPRAGE, and 2D dual-echo T2-weighted sequences).

Two different automated methods were used: M1, based on an unsupervised approach that used intensity-derived features from the subtraction images together with deformation fields information obtained from the non-rigid registration between the two scans; and M2, a supervised approach based on the application of convolutional neural networks (CNN) trained to detect the presence of new T2 lesions in the follow-up scan. The outcomes of these automated tools were compared to the results of two operator-related methods based on visual analysis: the standard radiological report (O1); and revision of the MRIs by an expert observer non-blinded to the radiological report (O2). A “Gold Standard Outcome” (GOS) was created by consensus of two expert observers based on combined visual assessment of all the MRI images, the radiological reports, and the outcomes of the automated methods.

Results

GOS identified 104 new T2 lesions in 38 patients. Automated tools doubled the number of new T2 lesions (125 for M1; and 119 for M2) compared to operator-related methods (59 for O1 and 73 for O2). Specificity for detecting patients with at least one new T2 lesion was 100% for operator related methods while for automatic tools was 83% for M1 and 87% for M2. Sensitivity was higher with both automated tools (92.1% for M1; 97.4% for M2) compared to operator related methods (76.3% for O1, and 89.5% for O2).

Conclusions

The CNN model was more sensitive for detecting new T2 lesions and active patients, compared to standard and expert visual analysis, and to an unsupervised automated tool. However, visual supervision of the CNN model outcomes is still required due to its suboptimal specificity. Automatic tools, based on the application of CNN models are promising for detecting MRI disease activity, and shows potential to be used as an aid to the neuroradiological visual assessment in clinical practice.