Background: Although cervical spinal cord (cSC) area is an established biomarker in MS, there is currently a lack of longitudinal assessments of cSC gray and white matter areas.Objective: We conducted an explorative analysis of longitudinal changes of cSC gray and white matter areas in MS patients.Methods: 65 MS patients (33 relapsing-remitting; 20 secondary progressive and 12 primary progressive) and 20 healthy controls (HC) received clinical and upper cSC MRI assessments over 1.10 +/- 0.28 years. cSC compartments were quantified on MRI using the novel averaged magnetization inversion recovery acquisitions sequence (in-plane resolution=0.67 x 0.67mm2), and in-house developed post-processing methods. Patients were stratified regarding clinical progression.Results: Patients with clinical progression showed faster reduction of cSC areas over time at the level of cSC enlargement (approximate vertebral level C4-C5) compared to stable patients (p<0.05). In addition, when compared to the rostral-cSC (approximate vertebral level C2-C3), a preferential reduction of cSC and white matter areas over time at the level of cSC enlargement (p<0.05 and p<0.01, respectively) was demonstrated only in patients with clinical progression, but not in stable MS patients and HC. Compared to HC, MS patients showed comparable changes over time in all cSC compartments.

BACKGROUND AND PURPOSE:Fully automatic quantification methods of spinal cord compartments are needed to study pathologic changes of the spinal cord GM and WM in MS in vivo. We propose a novel method for automatic spinal cord compartment segmentation (SCORE) in patients with MS.MATERIALS AND METHODS:The cervical spinal cords of 24 patients with MS and 24 sex- and age-matched healthy controls were scanned on a 3T MR imaging system, including an averaged magnetization inversion recovery acquisition sequence. Three experienced raters manually segmented the spinal cord GM and WM, anterior and posterior horns, gray commissure, and MS lesions. Subsequently, manual segmentations were used to train neural segmentation networks of spinal cord compartments with multidimensional gated recurrent units in a 3-fold cross-validation fashion. Total intracranial volumes were quantified using FreeSurfer.RESULTS:The intra- and intersession reproducibility of SCORE was high in all spinal cord compartments (eg, mean relative SD of GM and WM: ≤ 3.50% and ≤1.47%, respectively) and was better than manual segmentations (all P < .001). The accuracy of SCORE compared with manual segmentations was excellent, both in healthy controls and in patients with MS (Dice similarity coefficients of GM and WM: ≥ 0.84 and ≥0.92, respectively). Patients with MS had lower total WM areas (P < .05), and total anterior horn areas (P < .01 respectively), as measured with SCORE.CONCLUSIONS:We demonstrate a novel, reliable quantification method for spinal cord tissue segmentation in healthy controls and patients with MS and other neurologic disorders affecting the spinal cord. Patients with MS have reduced areas in specific spinal cord tissue compartments, which may be used as MS biomarkers.

Background: Spinal cord (SC) gray and white matter pathology plays a central role in multiple sclerosis (MS). Objective: We aimed to investigate the extent, pattern, and clinical relevance of SC gray and white matter atrophy in vivo. Methods: 39 relapsing–remitting patients (RRMS), 40 progressive MS patients (PMS), and 24 healthy controls (HC) were imaged at 3T using the averaged magnetization inversion recovery acquisitions sequence. Total and lesional cervical gray and white matter, and posterior (SCPH) and anterior horn (SCAH) areas were automatically quantified. Clinical assessment included the expanded disability status scale, timed 25-foot walk test, nine-hole peg test, and the 12-item MS walking scale. Results: PMS patients had significantly reduced cervical SCAH — but not SCPH — areas compared with HC and RRMS (both p < 0.001). In RRMS and PMS, the cervical SCAH areas increased significantly less in the region of cervical SC enlargement compared with HC (all p < 0.001). This reduction was more pronounced in PMS compared with RRMS (both p < 0.001). In PMS, a lower cervical SCAH area was the most important magnetic resonance imaging (MRI)-variable for higher disability scores. Conclusion: MS patients show clinically relevant cervical SCAH atrophy, which is more pronounced in PMS and at the level of cervical SC enlargement.

There is evidence that multiple sclerosis (MS) pathology leads to distinct patterns of volume loss over time (VLOT) in different central nervous system (CNS) structures. We aimed to use such patterns to identify patient subgroups. MS patients of all classical disease phenotypes underwent annual clinical, blood, and MRI examinations over 6 years. Spinal, striatal, pallidal, thalamic, cortical, white matter, and T2-weighted lesion volumes as well as serum neurofilament light chain (sNfL) were quantified. CNS VLOT patterns were identified using principal component analysis and patients were classified using hierarchical cluster analysis. 225 MS patients were classified into four distinct Groups A, B, C, and D including 14, 59, 141, and 11 patients, respectively). These groups did not differ in baseline demographics, disease duration, disease phenotype distribution, and lesion-load expansion. Interestingly, Group A showed pronounced spinothalamic VLOT, Group B marked pallidal VLOT, Group C small between-structure VLOT differences, and Group D myelocortical volume increase and pronounced white matter VLOT. Neurologic deficits were more severe and progressed faster in Group A that also had higher mean sNfL levels than all other groups. Group B experienced more frequent relapses than Group C. In conclusion, there are distinct patterns of VLOT across the CNS in MS patients, which do not overlap with clinical MS subtypes and are independent of disease duration and lesion-load but are partially associated to sNfL levels, relapse rates, and clinical worsening. Our findings support the need for a more biologic classification of MS subtypes including volumetric and body-fluid markers.

Corneal topography is an essential tool in ophthalmology, in particular for surgical planning and diagnostics. Optical coherence tomography (OCT) enables cross-sectional or volumetric imaging with high resolution. It is, however, not widely used for corneal topography. A major reason for this is that conventional beam-scanning OCT is susceptible to eye motion compared to established modalities, which measure corneal shape in a single shot. To overcome this limitation, we propose a novel pipeline for motion-compensated OCT-based corneal topography. The pipeline includes three main features: (1) continuous, two-dimensional scanning; (2) the three-dimensional continuous motion compensation in postprocessing; and (3) regularised Zernike reconstruction. First, we evaluated our method on an eye phantom that is moved to mimic typical eye motion. The proposed motion compensation was able to determine and correct the movements of the phantom. Second, we performed an in vivo study on 48 eyes, measuring each eye twice with our OCT-based topography, Placido disc topography (Atlas 9000, Carl Zeiss Meditec), and Scheimpflug (Pentacam, Oculus) topography. We then compared the performance of the OCT-based topography to the reference topographies in terms of repeatability and equivalence. The results confirm the necessity and efficiency of the presented motion compensation and validate the proposed methods for scanning and reconstruction.

Background: Brainstem-mediated functions are impaired in neurodegenerative diseases and aging. Atrophy can be visualized by MRI. This study investigates extrinsic sources of brainstem volume variability, intrinsic sources of anatomical variability, and the influence of age and sex on the brainstem volumes in healthy subjects. We aimed to develop efficient normalization strategies to reduce the effects of intrinsic anatomic variability on brainstem volumetry. Methods: Brainstem segmentation was performed from MPRAGE data using our deep-learning-based brainstem segmentation algorithm MD-GRU. The extrinsic variability of brainstem volume assessments across scanners and protocols was investigated in two groups comprising 11 (median age 33.3 years, 7 women) and 22 healthy subjects (median age 27.6 years, 50% women) scanned twice and compared using Dice scores. Intrinsic anatomical inter-individual variability and age and sex effects on brainstem volumes were assessed in segmentations of 110 healthy subjects (median age 30.9 years, range 18–72 years, 53.6% women) acquired on 1.5T (45%) and 3T (55%) scanners. The association between brainstem volumes and predefined anatomical covariates was studied using Pearson correlations. Anatomical variables with associations of |r| > 0.30 as well as the variables age and sex were used to construct normalization models using backward selection. The effect of the resulting normalization models was assessed by % relative standard deviation reduction and by comparing the inter-individual variability of the normalized brainstem volumes to the non-normalized values using paired t- tests with Bonferroni correction. Results: The extrinsic variability of brainstem volumetry across different field strengths and imaging protocols was low (Dice scores > 0.94). Mean inter-individual variability/SD of total brainstem volumes was 9.8%/7.36. A normalization based on either total intracranial volume (TICV), TICV and age, or v-scale significantly reduced the inter-individual variability of total brainstem volumes compared to non-normalized volumes and similarly reduced the relative standard deviation by about 35%. Conclusion: The extrinsic variability of the novel brainstem segmentation method MD-GRU across different scanners and imaging protocols is very low. Anatomic inter-individual variability of brainstem volumes is substantial. This study presents efficient normalization models for variability reduction in brainstem volumetry in healthy subjects.

Objective: To investigate medulla oblongata (MO), pons and mesencephalon atrophy on brain 3D T1-weighted images in early stage patients with Multiple Sclerosis (MS) using a fully-automated deep learning-based segmentation approach. Background: Brainstem (BS) involvement is regarded as a bad prognostic sign and can limit life expectancy in MS. In contrast to cortical, subcortical and spinal cord atrophy, BS atrophy has not been systematically investigated in early stage MS. Design/Methods: A novel, fully-automated deep learning-based segmentation approach (Sander et al. 2019) was used to assess BS volumes from high-resolution 3D T1w MPRAGE images obtained at either a 3T Skyra (24.4%) or 1.5T Avanto Scanner (75.6%) (Siemens Healthineers) in 182 patients with early stage MS (mean age 37.7 years, SD 11.2, 61% women, median EDSS 1.75, IQR 1.5, median disease duration 1.4 years, range 0.01–3.98 years) and 80 age- and sex-matched healthy controls (HC; mean age 37.7 years, SD 13.3, 60% women) as part of the ongoing Swiss MS cohort-study. BS function was assessed using the Neurostatus-EDSS BS Functional System Score. Results: Compared to HC, patients showed significant MO volume reductions (p=0.004; % difference 4.96%) in absence of significant atrophy of the mesencephalon and pons. This difference was visible in patients with (p=0.001) and without (p=0.016) clinical BS involvement (adjusted for age and sex). Even the patients with the shortest disease duration below the 25th percentile (range 0.01–0.72 years) showed a significant MO atrophy compared to HC (p=0.024, adjusted for age and sex). Conclusions: Isolated MO atrophy occurs very early in MS in absence of significant pons and mesencephalon atrophy. This observation is independent of clinical BS involvement. Further longitudinal studies are necessary to investigate its role as a marker for disease course monitoring and outcome prediction. Disclosure: Dr. Sander has nothing to disclose. Dr. Horvath has nothing to disclose. Dr. Pezold has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Advanced Osteotomy Tools AG (Basel, CH). Dr. Amann has nothing to disclose. Dr. Andermatt has nothing to disclose. Dr. Yaldizli has nothing to disclose. Dr. Wuerfel has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with CEO of the MIAC AG, also fees from Actelion, Biogen, Sanofi Genzyme, Novartis and Roche. Dr. Kappos has received research support from Bayer, Biogen, Innosuisse, Novartis, the Swiss MS Society, the Swiss National Research Foundation, and the European Union.Dr. Granziera has nothing to disclose. Dr. Gobbi has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Novartis, Biogen, Merck, Almirall, Roche, TEVA.Dr. Kuhle has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Biogen, Genzyme, Novartis, Roche, Teva, Merck. Dr. Kuhle has received research support from ECTRIMS Research Fellowship Programme, University of Basel, Swiss MS Society, Swiss National Research Foundation (320030_160221), Bayer, Biogen, Genzyme, Celgene, Novartis, Roche, Teva, Merck.Dr. Cattin has nothing to disclose. Dr. Schlaeger has received research support from Biogen.

Neurodegenerative disorders, such as Alzheimer's disease (AD) and progressive forms of multiple sclerosis (MS), can affect the brainstem and are associated with atrophy that can be visualized by MRI. Anatomically accurate, large-scale assessments of brainstem atrophy are challenging due to lack of automated, accurate segmentation methods. We present a novel method for brainstem volumetry using a fully-automated segmentation approach based on multi-dimensional gated recurrent units (MD-GRU), a deep learning based semantic segmentation approach employing a convolutional adaptation of gated recurrent units. The neural network was trained on 67 3D-high resolution T1-weighted MRI scans from MS patients and healthy controls (HC) and refined using segmentations of 20 independent MS patients' scans. Reproducibility was assessed in MR test-retest experiments in 33 HC. Accuracy and robustness were examined by Dice scores comparing MD-GRU to FreeSurfer and manual brainstem segmentations in independent MS and AD datasets. The mean %-change/SD between test-retest brainstem volumes were 0.45%/0.005 (MD-GRU), 0.95%/0.009 (FreeSurfer), 0.86%/0.007 (manually edited segmentations). Comparing MD-GRU to manually edited segmentations the mean Dice scores/SD were: 0.97/0.005 (brainstem), 0.95/0.013 (mesencephalon), 0.98/0.006 (pons), 0.95/0.015 (medulla oblongata). Compared to the manual gold standard, MD-GRU brainstem segmentations were more accurate than FreeSurfer segmentations (p < .001). In the multi-centric acquired AD data, the mean Dice score/SD for the MD-GRU-manual segmentation comparison was 0.97/0.006. The fully automated brainstem segmentation method MD-GRU provides accurate, highly reproducible, and robust segmentations in HC and patients with MS and AD in 200 s/scan on an Nvidia GeForce GTX 1080 GPU and shows potential for application in large and longitudinal datasets.

The original version of this chapter contained a figure (Fig. 4), which showed the internal electronics as well as the mechanical design of a device. This information should not be made available to the public. This was corrected and Figure 4 was replaced.

Background. When locating the sentinel lymph node (SLN), surgeons use state-of-the-art imaging devices, such as a 1D gamma probe or less widely spread a 2D gamma camera. These devices project the 3D subspace onto a 1D respectively 2D space, hence loosing accuracy and the depth of the SLN which is very important, especially in the head and neck area with many critical structures in close vicinity. Recent methods which use a multi-pinhole collimator and a single gamma detector image try to gain a depth estimation of the SLN. The low intensity of the sources together with the computational cost of the optimization process make the reconstruction in real-time, however, very challenging. Results. In this paper, we use an optimal design approach to improve the classical pinhole design, resulting in a non-symmetric distribution of the pinholes of the collimator. This new design shows a great improvement of the accuracy when reconstructing the position and depth of the radioactive tracer. Then, we introduce our Sentinel lymph node fingerprinting (SLNF) algorithm, inspired by MR-fingerprinting, for fast and accurate reconstruction of the tracer distribution in 3D space using a single gamma detector image. As a further advantage, the method requires no pre-processing, i.e. filtering of the detector image. The method is very stable in its performance even for low exposure times. In our ex vivo experiments, we successfully located multiple Technetium 99m (Tc-99m) sources with an exposure time of only one second and still, with a very small L-2-error. Conclusion. These promising results under short exposure time are very encouraging for SLN biopsy. Although, this device has not been tested on patients yet, we believe: that this approach will give the surgeon accurate 3D positions of the SLN and hence, can potentially reduce the trauma for the patient.

BACKGROUND:Squamous cell carcinoma in the head and neck region is one of the most widespread cancers with high morbidity. Classic treatment comprises the complete removal of the lymphatics together with the cancerous tissue. Recent studies have shown that such interventions are only required in 30% of the patients. Sentinel lymph node biopsy is an alternative method to stage the malignancy in a less invasive manner and to avoid overtreatment. In this paper, we present a novel approach that enables a future augmented reality device which improves the biopsy procedure by visual means.METHODS:We propose a co-calibration scheme for axis-aligned miniature cameras with pinholes of a gamma ray collimating and sensing device and show results gained by experiments, based on a calibration target visible for both modalities.RESULTS:Visual inspection and quantitative evaluation of the augmentation of optical camera images with gamma information are congruent with known gamma source landmarks.CONCLUSIONS:Combining a multi-pinhole collimator with axis-aligned miniature cameras to augment optical images using gamma detector data is promising. As such, our approach might be applicable for breast cancer and melanoma staging as well, which are also based on sentinel lymph node biopsy.

Fully supervised segmentation methods require a large training cohort of already segmented images, providing information at the pixel level of each image. We present a method to automatically segment and model pathologies in medical images, trained solely on data labelled on the image level as either healthy or containing a visual defect. We base our method on CycleGAN, an image-to-image translation technique, to translate images between the domains of healthy and pathological images. We extend the core idea with two key contributions. Implementing the generators as residual generators allows us to explicitly model the segmentation of the pathology. Realizing the translation from the healthy to the pathological domain using a variational autoencoder allows us to specify one representation of the pathology, as this transformation is otherwise not unique. Our model hence not only allows us to create pixelwise semantic segmentations, it is also able to create inpaintings for the segmentations to render the pathological image healthy. Furthermore, we can draw new unseen pathology samples from this model based on the distribution in the data. We show quantitatively, that our method is able to segment pathologies with a surprising accuracy being only slightly inferior to a state-of-the-art fully supervised method, although the latter has per-pixel rather than per-image training information. Moreover, we show qualitative results of both the segmentations and inpaintings. Our findings motivate further research into weakly-supervised segmentation using image level annotations, allowing for faster and cheaper acquisition of training data without a large sacrifice in segmentation accuracy.

BACKGROUND AND PURPOSE:Currently, accurate and reproducible spinal cord GM segmentation remains challenging and a noninvasive broadly accepted reference standard for spinal cord GM measurements is still a matter of ongoing discussion. Our aim was to assess the reproducibility and accuracy of cervical spinal cord GM and WM cross-sectional area measurements using averaged magnetization inversion recovery acquisitions images and a fully-automatic postprocessing segmentation algorithm.MATERIALS AND METHODS:The cervical spinal cord of 24 healthy subjects (14 women; mean age, 40 ± 11 years) was scanned in a test-retest fashion on a 3T MR imaging system. Twelve axial averaged magnetization inversion recovery acquisitions slices were acquired over a 48-mm cord segment. GM and WM were both manually segmented by 2 experienced readers and compared with an automatic variational segmentation algorithm with a shape prior modified for 3D data with a slice similarity prior. Precision and accuracy of the automatic method were evaluated using coefficients of variation and Dice similarity coefficients.RESULTS:The mean GM area was 17.20 ± 2.28 mm2 and the mean WM area was 72.71 ± 7.55 mm2 using the automatic method. Reproducibility was high for both methods, while being better for the automatic approach (all mean automatic coefficients of variation, ≤4.77%; all differences, P < .001). The accuracy of the automatic method compared with the manual reference standard was excellent (mean Dice similarity coefficients: 0.86 ± 0.04 for GM and 0.90 ± 0.03 for WM). The automatic approach demonstrated similar coefficients of variation between intra- and intersession reproducibility as well as among all acquired spinal cord slices.CONCLUSIONS:Our novel approach including the averaged magnetization inversion recovery acquisitions sequence and a fully-automated postprocessing segmentation algorithm demonstrated an accurate and reproducible spinal cord GM and WM segmentation. This pipeline is promising for both the exploration of longitudinal structural GM changes and application in clinical settings in disorders affecting the spinal cord.

Quantification of cerebral white matter hyperintensities (WMH) of presumed vascular origin is of key importance in many neurological research studies. Currently, measurements are often still obtained from manual segmentations on brain MR images, which is a laborious procedure. The automatic WMH segmentation methods exist, but a standardized comparison of the performance of such methods is lacking. We organized a scientific challenge, in which developers could evaluate their methods on a standardized multi-center/-scanner image dataset, giving an objective comparison: the WMH Segmentation Challenge. Sixty T1 FLAIR images from three MR scanners were released with the manual WMH segmentations for training. A test set of 110 images from five MR scanners was used for evaluation. The segmentation methods had to be containerized and submitted to the challenge organizers. Five evaluation metrics were used to rank the methods: 1) Dice similarity coefficient; 2) modified Hausdorff distance (95th percentile); 3) absolute log-transformed volume difference; 4) sensitivity for detecting individual lesions; and 5) F1-score for individual lesions. In addition, the methods were ranked on their inter-scanner robustness; 20 participants submitted their methods for evaluation. This paper provides a detailed analysis of the results. In brief, there is a cluster of four methods that rank significantly better than the other methods, with one clear winner. The inter-scanner robustness ranking shows that not all the methods generalize to unseen scanners. The challenge remains open for future submissions and provides a public platform for method evaluation.

Sentinel lymph node biopsy (SNB) is a surgical method to stage certain cancer types in a minimally invasive manner. However, the current sensing methods for SNB are limited in accuracy, as they are based on acoustic feedback radiation probes to detect tracer enriched sentinel lymph nodes. We present a deep neural network approach to learn the latent spatial activity distributions from a simulated gamma source on 2D activity images. Data processing can then be applied for multi-pinhole collimator optimization, lymph node visualization or surgical navigation to further support SNB. Using simulations of photon multi-pinhole collimator interaction, we generate labeled synthetic 2D activity images to train convolutional neural networks (CNN). These CNNs are then evaluated on synthetic as well as on real experimental data from a radioactive point-like source, collected by our own stationary small form factor multi-pinhole collimator. We achieve good results on synthetic data for the xy-component ensemble learners with a localization class accuracy of 0.97, while depth estimation achieves a localization class accuracy of 0.55. Accuracy on real experimental data is limited due to the small sample set and its variability, compared to the simulation.

Background: Functional lung imaging with MRI allows evaluation of ventilation and perfusion deficits in lung disease. To quantify deficits, segmentation of lung tissue is required, which is subjective and time-consuming. Deep learning (DL) algorithms could accelerate this process with high precision. Aim: We assessed the variation in relative impaired ventilation and perfusion resulting from segmentations from different readers. Methods: This study included 29 MRI scans from 17 children with cystic fibrosis and 9 healthy. Pulmonary tissue was segmented on base images. A matrix pencil algorithm computed perfusion- and ventilation-weighted maps of the lung and calculated the relative fractional ventilation (RFV) and perfusion (RQ) impairment. The RFV and RQ resulting from the segmentations of two experienced human readers (A & B) and a recurrent neural network (DL) were compared. Reader A repeated segmentation after 24 hours to investigate intra-reader variability. Agreement between readers was assessed with paired t-test and intra-class correlation coefficient (ICC). Results: There was very good agreement between all readers (ICC: 0.98; 0.92 - 0.99). There was a significant difference in RFV from reader A (mean=26.8; SD=6.4) to B (28.2; 6.1; p<0.001) and DL (28.0; 6.3; p<0.001). No significant differences in repeated segmentation from reader A (difference: -0.03; p=0.62) were found. The differences for RQ were comparable to RFV. Conclusion: We found small but statistically significant differences in outcomes between observers. The inter-reader variability of RFV and RQ between human and machine was similar to the variability between human readers. DL may be a promising alternative to human segmentation.

The small butterfly shaped structure of spinal cord (SC) gray matter (GM) is challenging to image and to delinate from its surrounding white matter (WM). Segmenting GM is up to a point a trade-off between accuracy and precision. We propose a new pipeline for GM-WM magnetic resonance (MR) image acquisition and segmentation. We report superior results as compared to the ones recently reported in the SC GM segmentation challenge and show even better results using the averaged magnetization inversion recovery acquisitions (AMIRA) sequence. Scan-rescan experiments with the AMIRA sequence show high reproducibility in terms of Dice coefficient, Hausdorff distance and relative standard deviation. We use a recurrent neural network (RNN) with multi-dimensional gated recurrent units (MD-GRU) to train segmentation models on the AMIRA dataset of 855 slices. We added a generalized dice loss to the cross entropy loss that MD-GRU uses and were able to improve the results.

Registration of thoracic images is central when studying for example physiological changes of the lung. Due to sliding organ motion and intensity changes based on respiration the registration of thoracic images is challenging. We present a novel regularisation method based on adaptive anisotropic graph diffusion. Without the need of a mask it preserves discontinuities of the transformation at sliding organ boundaries and enforces smoothness in areas with similar motion. The graph diffusion regularisation provides a direct way to achieve anisotropic diffusion at sliding organ boundaries by reducing the weight of corresponding edges in the graph which cross the sliding interfaces. Since the graph diffusion is defined by the edge weights of the graph, we develop an adaptive edge weight function to detect sliding boundaries. We implement the adaptive graph diffusion regularisation method in the Demons registration framework. The presented method is tested on synthetic 2D images and on the public 4D-CT DIR-Lab data set, where we are able to correctly detect the sliding organ boundaries.