Pierre-Yves Baudin

Automatic skeletal muscle segmentation through random walks and graph-based seed placement

In this paper we propose a novel skeletal muscle segmentation method driven from discrete optimization. We introduce a graphical model that is able to automatically determine appropriate seed positions with respect to the different muscle classes. This is achieved by taking into account the expected local visual and geometric properties of the seeds through a pair-wise Markov Random Field. The outcome of this optimization process is fed to a powerful graph-based diffusion segmentation method (random walker) that is able to produce very promising results through a fully automated approach. Validation on challenging data sets demonstrates the potentials of our method.

Skeletal Muscle Quantitative Nuclear Magnetic Resonance Imaging and Spectroscopy as an Outcome Measure for Clinical Trials

Recent years have seen tremendous progress towards therapy of many previously incurable neuromuscular diseases. This new context has acted as a driving force for the development of novel non-invasive outcome measures. These can be organized in three main categories: functional tools, fluid biomarkers and imagery. In the latest category, nuclear magnetic resonance imaging (NMRI) offers a considerable range of possibilities for the characterization of skeletal muscle composition, function and metabolism. Nowadays, three NMR outcome measures are frequently integrated in clinical research protocols. They are: 1/ the muscle cross sectional area or volume, 2/ the percentage of intramuscular fat and 3/ the muscle water T2, which quantity muscle trophicity, chronic fatty degenerative changes and oedema (or more broadly, “disease activity”), respectively. A fourth biomarker, the contractile tissue volume is easily derived from the first two ones. The fat fraction maps most often acquired with Dixon sequences have proven their capability to detect small changes in muscle composition and have repeatedly shown superior sensitivity over standard functional evaluation. This outcome measure will more than likely be the first of the series to be validated as an endpoint by regulatory agencies. The versatility of contrast generated by NMR has opened many additional possibilities for characterization of the skeletal muscle and will result in the proposal of more NMR biomarkers. Ultra-short TE (UTE) sequences, late gadolinium enhancement and NMR elastography are being investigated as candidates to evaluate skeletal muscle interstitial fibrosis. Many options exist to measure muscle perfusion and oxygenation by NMR. Diffusion NMR as well as texture analysis algorithms could generate complementary information on muscle organization at microscopic and mesoscopic scales, respectively. 31P NMR spectroscopy is the reference technique to assess muscle energetics non-invasively during and after exercise. In dystrophic muscle, 31P NMR spectrum at rest is profoundly perturbed, and several resonances inform on cell membrane integrity. Considerable efforts are being directed towards acceleration of image acquisitions using a variety of approaches, from the extraction of fat content and water T2 maps from one single acquisition to partial matrices acquisition schemes. Spectacular decreases in examination time are expected in the near future. They will reinforce the attractiveness of NMR outcome measures and will further facilitate their integration in clinical research trials.

Simultaneous muscle water T2 and fat fraction mapping using transverse relaxometry with stimulated echo compensation

Skeletal muscle inflammation/necrosis and fat infiltration are strong indicators of disease activity and progression in many neuromuscular disorders. They can be assessed by muscle T2 relaxometry and water‐fat separation techniques, respectively. In the present work, we exploited differences between water and fat T1 and T2 relaxivities by applying a bi‐component extended phase graph (EPG) fitting approach to simultaneously quantify the muscle water T2 and fat fraction from standard multi‐slice multi‐echo (MSME) acquisitions in the presence of stimulated echoes. Experimental decay curves were adjusted to the theoretical model using either an iterative non‐negative least‐squares (NNLS) procedure or a pattern recognition approach. Twenty‐two patients (age, 49 ± 18 years) were selected to cover a large range of muscle fat infiltration. Four cases of chronic or subchronic juvenile dermatomyositis (age, 8 ± 3 years) were investigated before and 3 months following steroid treatment. For control, five healthy volunteers (age, 25 ± 2 years) were recruited. All subjects underwent the MSME sequence and EPG fitting procedure. The EPG fitting algorithm allowed a precise estimation of water T2 and fat fraction in diseased muscle, even in the presence of large B1+ inhomogeneities. In the whole cohort of patients, there was no overall correlation between water T2 values obtained with the proposed method and the fat fraction estimated inside muscle tissues (R2 = 0.02). In the patients with dermatomyositis, there was a significant decrease in water T2 (‐4.09 ± 3.7 ms) consequent to steroid treatment. The pattern recognition approach resulted in a 20‐fold decrease in processing time relative to the iterative NNLS procedure. The fat fraction derived from the EPG fitting approach correlated well with the fat fraction derived from a standard three‐point Dixon method (≈1.5% bias). The bi‐component EPG fitting analysis is a precise tool to monitor muscle tissue disease activity and is able to handle bias introduced by fat infiltration and B1+ inhomogeneities. Copyright

Prospective exploratory muscle biopsy, imaging, and functional assessment in patients with late-onset Pompe disease treated with alglucosidase alfa: The EMBASSY Study

BACKGROUND Late-onset Pompe disease is characterized by progressive skeletal myopathy followed by respiratory muscle weakness, typically leading to loss of ambulation and respiratory failure. In this population, enzyme replacement therapy (ERT) with alglucosidase alfa has been shown to stabilize respiratory function and improve mobility and muscle strength. Muscle pathology and glycogen clearance from skeletal muscle in treatment-naïve adults after ERT have not been extensively examined. METHODS This exploratory, open-label, multicenter study evaluated glycogen clearance in muscle tissue samples collected pre- and post- alglucosidase alfa treatment in treatment-naïve adults with late-onset Pompe disease. The primary endpoint was the quantitative reduction in percent tissue area occupied by glycogen in muscle biopsies from baseline to 6months. Secondary endpoints included qualitative histologic assessment of tissue glycogen distribution, secondary pathology changes, assessment of magnetic resonance images (MRIs) for intact muscle and fatty replacement, and functional assessments. RESULTS Sixteen patients completed the study. After 6months of ERT, the percent tissue area occupied by glycogen in quadriceps and deltoid muscles decreased in 10 and 8 patients, respectively. No changes were detected on MRI from baseline to 6months. A majority of patients showed improvements on functional assessments after 6months of treatment. All treatment-related adverse events were mild or moderate. CONCLUSIONS This exploratory study provides novel insights into the histopathologic effects of ERT in late-onset Pompe disease patients. Ultrastructural examination of muscle biopsies demonstrated reduced lysosomal glycogen after ERT. Findings are consistent with stabilization of disease by ERT in treatment-naïve patients with late-onset Pompe disease.

Discriminative Parameter Estimation for Random Walks Segmentation

The Random Walks (RW) algorithm is one of the most efficient and easy-to-use probabilistic segmentation methods. By combining contrast terms with prior terms, it provides accurate segmentations of medical images in a fully automated manner. However, one of the main drawbacks of using the RW algorithm is that its parameters have to be hand-tuned. we propose a novel discriminative learning framework that estimates the parameters using a training dataset. The main challenge we face is that the training samples are not fully supervised. Specifically, they provide a hard segmentation of the images, instead of a probabilistic segmentation. We overcome this challenge by treating the optimal probabilistic segmentation that is compatible with the given hard segmentation as a latent variable. This allows us to employ the latent support vector machine formulation for parameter estimation. We show that our approach significantly outperforms the baseline methods on a challenging dataset consisting of real clinical 3D MRI volumes of skeletal muscles.

G.P.119

Fatty infiltration of muscles is a marker of disease progression in many neuromuscular disorders. Muscle MRI is capable of revealing patterns of muscles involvement that are disease specific and facilitates the diagnostic workup of patients. Although routine T1-weighted (T1w) imaging can give an indication of the presence or absence of muscular fat infiltration, it is difficult to extract quantitative data from these images. On the contrary, Dixon methods provide quantitative measure of fat fraction. Usually, whole-body (WB) exams consist in the acquisition of T1w images, followed by Dixon acquisitions on targeted regions to quantitatively assess fat infiltration. With the aim of improving and accelerating the qualitative assessment of neuromuscular disorders, we propose to avoid the T1w acquisition altogether, by allowing to perform the visual diagnosis workup on a WB Dixon imaging. 20 patients underwent WB MRI at 3T. WB T1w images were acquired with a 2D TSE sequence (resolution=1.1×1.1mm 2 , slice thickness=6mm, T acq =5min 40s). WB Dixon acquisition consisted in a 3D VIBE sequence with 3 echoes (resolution=1×1×5mm 3 , T acq =14min 5s). Quantitative fat fraction maps were derived using a standard 3-points Dixon reconstruction method. A customed lookup table was embedded in the DICOM file to provide a colored lecture of fat fraction maps corresponding to the Mercuris scale. Our results show that the acquisition of a high resolution WB Dixon imaging is possible in less than 15min using an optimized VIBE sequence. This provides quantitative data that are more suitable than T1w images for longitudinal natural history studies, or therapeutic clinical trials. Moreover, the color representation renders the visual grading of the muscles more convenient and less operator dependent as it is based on actual fat fraction measurements. WB Dixon might then overcome the use of WB T1w images for diagnostic of neuromuscular disorders.