A comprehensive study on automated muscle segmentation for assessing fat infiltration in neuromuscular diseases

Gadermayr, Michael; Disch, Constantin; Müller, Mira; Merhof, Dorit; Gess, Burkhard

doi:10.1016/j.mri.2017.12.014

Cited by 26 publications

(48 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, the rationale behind this setting is that healthy data can be easily segmented automatically in a fully unsupervised way by means of thresholding. 5 Now we know that healthy data 1) can be effectively augmented by adding elastic deformation and artificial fatty infiltrations, and 2) training with simulated data alone is sufficient to perform a proper segmentation of "severe" samples. A fully unsupervised approach could be based on automated unsupervised segmentation of healthy subjects' images, e.g, using the unsupervised GMM approach.…”

Section: Discussionmentioning

confidence: 99%

“…A fully unsupervised approach could be based on automated unsupervised segmentation of healthy subjects' images, e.g, using the unsupervised GMM approach. 5 The obtained image-label pairs can be augmented (to incorporate pathological data as well) in order to train a segmentation network that is finally applicable to healthy and pathological samples. Such an approach only required labels at the patient level, indicating whether the MR images show fatty infiltrations.…”

Section: Discussionmentioning

confidence: 99%

“…The segmentation of "healthy" data was not considered here, because basic techniques, eg, relying on thresholding only, proved to be sufficient for this purpose. 5,13 For training the data augmentation model, one dataset consisted of "healthy" data and the other one was obtained by merging "moderate" and "severe" MRI data (Table 1). Data showing pathological modifications were merged in order to obtain maximum data for training the GAN.…”

Section: Experimental Settingsmentioning

confidence: 99%

“…This is because pathological muscle tissue cannot be distinguished from subcutaneous fat based on the MR image's signal intensity only and the borders between these two tissue categories are often extremely difficult to detect. 5 Performing manual annotation, on the other hand, is extremely time-consuming, 6,7 especially because recent work 4 also suggested computing the fatfraction of complete 3D data (instead of single 2D slices), providing incentives for the development of automated segmentation approaches. Apart from enhanced efficiency, automated methods also facilitate observer-independent processing.…”

mentioning

confidence: 99%

“…18,19 It was shown that approaches for segmenting healthy muscle tissue in general do not perform well on pathological tissue. 5 To tackle these challenges, approaches for detecting the fascia lata that wraps the muscle tissue and separates it from subcutaneous fat tissue were proposed. [20][21][22] Although showing excellent performance, these approaches were applied to an easier scenario, where all tissue inside the fascia lata was labeled as muscle (apart from the bone) which is not completely correct, but probably serves as a good approximation.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Domain‐specific data augmentation for segmenting MR images of fatty infiltrated human thighs with neural networks

Gadermayr

Müller

et al. 2019

Magnetic Resonance Imaging

Self Cite

View full text Add to dashboard Cite

Background Fat‐fraction has been established as a relevant marker for the assessment and diagnosis of neuromuscular diseases. For computing this metric, segmentation of muscle tissue in MR images is a first crucial step. Purpose To tackle the high degree of variability in combination with the high annotation effort for training supervised segmentation models (such as fully convolutional neural networks). Study Type Prospective. Subjects In all, 41 patients consisting of 20 patients showing fatty infiltration and 21 healthy subjects. Field Strength/Sequence: The T1‐weighted MR‐pulse sequences were acquired on a 1.5T scanner. Assessment To increase performance with limited training data, we propose a domain‐specific technique for simulating fatty infiltrations (i.e., texture augmentation) in nonaffected subjects' MR images in combination with shape augmentation. For simulating the fatty infiltrations, we make use of an architecture comprising several competing networks (generative adversarial networks) that facilitate a realistic artificial conversion between healthy and infiltrated MR images. Finally, we assess the segmentation accuracy (Dice similarity coefficient). Statistical Tests A Wilcoxon signed rank test was performed to assess whether differences in segmentation accuracy are significant. Results The mean Dice similarity coefficients significantly increase from 0.84–0.88 (P < 0.01) using data augmentation if training is performed with mixed data and from 0.59–0.87 (P < 0.001) if training is conducted with healthy subjects only. Data Conclusion Domain‐specific data adaptation is highly suitable for facilitating neural network‐based segmentation of thighs with feasible manual effort for creating training data. The results even suggest an approach completely bypassing manual annotations. Level of Evidence: 4 Technical Efficacy: Stage 3

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Experimental Settingsmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Domain‐specific data augmentation for segmenting MR images of fatty infiltrated human thighs with neural networks

Gadermayr

Müller

et al. 2019

Magnetic Resonance Imaging

Self Cite

View full text Add to dashboard Cite

show abstract

The Impact of Fatty Infiltration on MRI Segmentation of Lower Limb Muscles in Neuromuscular Diseases: A Comparative Study of Deep Learning Approaches

Hostin

Ogier

Michel

et al. 2023

Magnetic Resonance Imaging

View full text Add to dashboard Cite

BackgroundDeep learning methods have been shown to be useful for segmentation of lower limb muscle MRIs of healthy subjects but, have not been sufficiently evaluated on neuromuscular disease (NDM) patients.PurposeEvaluate the influence of fat infiltration on convolutional neural network (CNN) segmentation of MRIs from NMD patients.Study TypeRetrospective study.SubjectsData were collected from a hospital database of 67 patients with NMDs and 14 controls (age: 53 ± 17 years, sex: 48 M, 33 F). Ten individual muscles were segmented from the thigh and six from the calf (20 slices, 200 cm section).Field Strength/SequenceA 1.5 T. Sequences: 2D T1‐weighted fast spin echo. Fat fraction (FF): three‐point Dixon 3D GRE, magnetization transfer ratio (MTR): 3D MT‐prepared GRE, T2: 2D multispin‐echo sequence.AssessmentU‐Net 2D, U‐Net 3D, TransUNet, and HRNet were trained to segment thigh and leg muscles (101/11 and 95/11 training/validation images, 10‐fold cross‐validation). Automatic and manual segmentations were compared based on geometric criteria (Dice coefficient [DSC], outlier rate, absence rate) and reliability of measured MRI quantities (FF, MTR, T2, volume).Statistical TestsBland–Altman plots were chosen to describe agreement between manual vs. automatic estimated FF, MTR, T2 and volume. Comparisons were made between muscle populations with an FF greater than 20% (G20+) and lower than 20% (G20−).ResultsThe CNNs achieved equivalent results, yet only HRNet recognized every muscle in the database, with a DSC of 0.91 ± 0.08, and measurement biases reaching −0.32% ± 0.92% for FF, 0.19 ± 0.77 for MTR, −0.55 ± 1.95 msec for T2, and − 0.38 ± 3.67 cm3 for volume. The performances of HRNet, between G20− and G20+ decreased significantly.Data ConclusionHRNet was the most appropriate network, as it did not omit any muscle. The accuracy obtained shows that CNNs could provide fully automated methods for studying NMDs. However, the accuracy of the methods may be degraded on the most infiltrated muscles (>20%).Evidence Level4.Technical EfficacyStage 1.

show abstract

A novel segmentation framework dedicated to the follow‐up of fat infiltration in individual muscles of patients with neuromuscular disorders

Ogier

Heskamp

Michel

et al. 2019

Magnetic Resonance in Med

View full text Add to dashboard Cite

PurposeTo propose a novel segmentation framework that is dedicated to the follow‐up of fat infiltration in individual muscles of patients with neuromuscular disorders.MethodsWe designed a semi‐automatic segmentation pipeline of individual leg muscles in MR images based on automatic propagation through nonlinear registrations of initial delineation in a minimal number of MR slices. This approach has been validated for the segmentation of individual muscles from MRI data sets, acquired over a 10‐month period, from thighs and legs in 10 patients with muscular dystrophy. The robustness of the framework was evaluated using conventional metrics related to muscle volume and clinical metrics related to fat infiltration.ResultsHigh accuracy of the semi‐automatic segmentation (mean Dice similarity coefficient higher than 0.89) was reported. The provided method has excellent reliability regarding the reproducibility of the fat fraction estimation, with an average intraclass correlation coefficient score of 0.99. Furthermore, the present segmentation framework was determined to be more reliable than the intra‐expert performance, which had an average intraclass correlation coefficient of 0.93.ConclusionThe proposed framework of segmentation can successfully provide an effective and reliable tool for accurate follow‐up of any MRI biomarkers in neuromuscular disorders. This method could assist the quantitative assessment of muscular changes occurring in such diseases.

show abstract

A comprehensive study on automated muscle segmentation for assessing fat infiltration in neuromuscular diseases

Cited by 26 publications

References 28 publications

Domain‐specific data augmentation for segmenting MR images of fatty infiltrated human thighs with neural networks

Domain‐specific data augmentation for segmenting MR images of fatty infiltrated human thighs with neural networks

The Impact of Fatty Infiltration on MRI Segmentation of Lower Limb Muscles in Neuromuscular Diseases: A Comparative Study of Deep Learning Approaches

A novel segmentation framework dedicated to the follow‐up of fat infiltration in individual muscles of patients with neuromuscular disorders

Contact Info

Product

Resources

About