D. Reich
NINDS, NIH Translational Neuroradiology SectionAuthor Of 2 Presentations
PS11.05 - Inclusion of small ovoid lesions in central vein sign assessment improves sensitivity for multiple sclerosis
Abstract
Background
The ‘central vein sign’ (CVS) is increasingly recognized as a valuable biomarker with high specificity and sensitivity for multiple sclerosis (MS) MRI lesions. Current consensus North American Imaging in Multiple Sclerosis (NAIMS) guidelines recommend excluding lesions <3mm in diameter in any plane for CVS assessment. However, different lesion-size exclusion cut-offs for CVS have not been systematically evaluated.
Objectives
To evaluate the impact of different lesion size cut-offs and exclusion methodologies on CVS analysis and select3* criteria for MS diagnosis.
Methods
MS patients and non-MS controls were recruited as part of the National Institute of Neurological Disorders and Stroke MS natural history study and underwent 3T MRI on either Siemens Skyra or Philips Achieva scanners. MS lesions were segmented using a deep learning-based method and manually corrected by a single rater. Individual lesions were extracted as clusters of connected voxels, and their principal axes lengths (calculated as the lengths of the major axes of an ellipsoid with the same normalized second central moments) were used to measure lesion size in 3 dimensions. Ground truth CVS assessment was conducted by two raters on all lesions regardless of size. Two paradigms of lesion exclusion were compared: (1) excluding lesions if any dimension was less than threshold (ExcAny), or (2) if all dimensions were less than threshold (ExcAll).
Results
A total of 3920 lesions from 71 subjects (8 healthy controls, 36 RRMS, 12 SPMS, 14 PPMS, 1 CIS) were included in the analysis. CVS+ lesions were more likely to be ovoid and less spherical compared to their CVS- counterparts, as measured by the fractional anisotropy of lesion dimensions (mean difference 0.02, p=0.001). Of the 1679 CVS+ lesions in the cohort, 82% met the ExcAny criteria to be excluded at a 3mm cut-off, which was reduced to 29% when ExcAll criteria were used (McNemar test, p < 0.001). At the subject-level, an increase in the sensitivity of select3* CVS criteria for MS diagnosis was noted at 3mm using the less strict ExcAll (95%) compared to the more conservative ExcAny criteria (61%), without impacting specificity (100% for both methods). There was a reduction in specificity for both ExcAny and ExcAll criteria when size cut-offs less than or equal to 2mm were used (88% for both).
Conclusions
Compared to the current NAIMS guidelines, ExcAll criteria for CVS lesion analysis allow the inclusion of a larger proportion of CVS+ lesions and improve the sensitivity of select3* criteria for MS diagnosis. These findings improve the applicability of the CVS as a diagnostic marker for MS in clinical practice and provide evidence for future modifications of CVS lesion exclusion guidelines.
PS16.04 - RimNet: A deep 3D multimodal MRI architecture for paramagnetic rim lesions assessment in multiple sclerosis
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.