Segmentation of fat and muscle from MR images of the thigh by a possibilistic clustering algorithm

Automatic quadriceps and patellae segmentation of MRI with cascaded U2 -Net and SASSNet deep learning model

PURPOSE

Automatic muscle segmentation is critical for advancing our understanding of human physiology, biomechanics, and musculoskeletal pathologies, as it allows for timely exploration of large multi-dimensional image sets. Segmentation models are rarely developed/validated for the pediatric model. As such, autosegmentation is not available to explore how muscle architectural changes during development and how disease/pathology affects the developing musculoskeletal system. Thus, we aimed to develop and validate an end-to-end, fully automated, deep learning model for accurate segmentation of the rectus femoris and vastus lateral, medialis, and intermedialis using a pediatric database.

METHODS

We developed a two-stage cascaded deep learning model in a coarse-to-fine manner. In the first stage, the U² -Net roughly detects the muscle subcompartment region. Then, in the second stage, the shape-aware 3D semantic segmentation method SASSNet refines the cropped target regions to generate the more finer and accurate segmentation masks. We utilized multifeature image maps in both stages to stabilize performance and validated their use with an ablation study. The second-stage SASSNet was independently run and evaluated with three different cropped region resolutions: the original image resolution, and images downsampled 2× and 4× (high, mid, and low). The relationship between image resolution and segmentation accuracy was explored. In addition, the patella was included as a comparator to past work. We evaluated segmentation accuracy using leave-one-out testing on a database of 3D MR images (0.43 × 0.43 × 2 mm) from 40 pediatric participants (age 15.3 ± 1.9 years, 55.8 ± 11.8 kg, 164.2 ± 7.9 cm, 38F/2 M).

RESULTS

The mid-resolution second stage produced the best results for the vastus medialis, rectus femoris, and patella (Dice similarity coefficient = 95.0%, 95.1%, 93.7%), whereas the low-resolution second stage produced the best results for the vastus lateralis and vastus intermedialis (DSC = 94.5% and 93.7%). In comparing the low- to mid-resolution cases, the vasti intermedialis, vastus medialis, rectus femoris, and patella produced significant differences (p = 0.0015, p = 0.0101, p < 0.0001, p = 0.0003) and the vasti lateralis did not (p = 0.2177). The high-resolution stage 2 had significantly lower accuracy (1.0 to 4.4 dice percentage points) compared to both the mid- and low-resolution routines (p value ranged from < 0.001 to 0.04). The one exception was the rectus femoris, where there was no difference between the low- and high-resolution cases. The ablation study demonstrated that the multifeature is more reliable than the single feature.

CONCLUSIONS

Our successful implementation of this two-stage segmentation pipeline provides a critical tool for expanding pediatric muscle physiology and clinical research. With a relatively small and variable dataset, our fully automatic segmentation technique produces accuracies that matched or exceeded the current state of the art. The two-stage segmentation avoids memory issues and excessive run times by using a first stage focused on cropping out unnecessary data. The excellent Dice similarity coefficients improve upon previous template-based automatic and semiautomatic methodologies targeting the leg musculature. More importantly, with a naturally variable dataset (size, shape, etc.), the proposed model demonstrates slightly improved accuracies, compared to previous neural networks methods.

Deep learning for automatic segmentation of thigh and leg muscles

Objective

In this study we address the automatic segmentation of selected muscles of the thigh and leg through a supervised deep learning approach.

Material and methods

The application of quantitative imaging in neuromuscular diseases requires the availability of regions of interest (ROI) drawn on muscles to extract quantitative parameters. Up to now, manual drawing of ROIs has been considered the gold standard in clinical studies, with no clear and universally accepted standardized procedure for segmentation. Several automatic methods, based mainly on machine learning and deep learning algorithms, have recently been proposed to discriminate between skeletal muscle, bone, subcutaneous and intermuscular adipose tissue. We develop a supervised deep learning approach based on a unified framework for ROI segmentation.

Results

The proposed network generates segmentation maps with high accuracy, consisting in Dice Scores ranging from 0.89 to 0.95, with respect to “ground truth” manually segmented labelled images, also showing high average performance in both mild and severe cases of disease involvement (i.e. entity of fatty replacement).

Discussion

The presented results are promising and potentially translatable to different skeletal muscle groups and other MRI sequences with different contrast and resolution.

Overview of MR Image Segmentation Strategies in Neuromuscular Disorders

Neuromuscular disorders are rare diseases for which few therapeutic strategies currently exist. Assessment of therapeutic strategies efficiency is limited by the lack of biomarkers sensitive to the slow progression of neuromuscular diseases (NMD). Magnetic resonance imaging (MRI) has emerged as a tool of choice for the development of qualitative scores for the study of NMD. The recent emergence of quantitative MRI has enabled to provide quantitative biomarkers more sensitive to the evaluation of pathological changes in muscle tissue. However, in order to extract these biomarkers from specific regions of interest, muscle segmentation is mandatory. The time-consuming aspect of manual segmentation has limited the evaluation of these biomarkers on large cohorts. In recent years, several methods have been proposed to make the segmentation step automatic or semi-automatic. The purpose of this study was to review these methods and discuss their reliability, reproducibility, and limitations in the context of NMD. A particular attention has been paid to recent deep learning methods, as they have emerged as an effective method of image segmentation in many other clinical contexts.

A Valid and Precise Semiautomated Method for Quantifying Intermuscular Fat Intramuscular Fat in Lower Leg Magnetic Resonance Images

The accumulation of INTERmuscular fat and INTRAmuscular fat (IMF) has been a hallmark of individuals with diabetes, those with mobility impairments such as spinal cord injuries and is known to increase with aging. An elevated amount of IMF has been associated with fractures and frailty, but the imprecision of IMF measurement has so far limited the ability to observe more consistent clinical associations. Magnetic resonance imaging has been recognized as the gold standard for portraying these features, yet reliable methods for quantifying IMF on magnetic resonance imaging is far from standardized. Previous investigators used manual segmentation guided by histogram-based region-growing, but these techniques are subjective and have not demonstrated reliability. Others applied fuzzy classification, machine learning, and atlas-based segmentation methods, but each is limited by the complexity of implementation or by the need for a learning set, which must be established each time a new disease cohort is examined. In this paper, a simple convergent iterative threshold-optimizing algorithm was explored. The goal of the algorithm is to enable IMF quantification from plain fast spin echo (FSE) T1-weighted MR images or from water-saturated images. The algorithm can be programmed into Matlab easily, and is semiautomated, thus minimizing the subjectivity of threshold-selection. In 110 participants from 3 cohort studies, IMF area measurement demonstrated a high degree of reproducibility with errors well within the 5% benchmark for intraobserver, interobserver, and test-retest analyses; in contrast to manual segmentation which already yielded over 20% error for intraobserver analysis. This algorithm showed validity against manual segmentations (r > 0.85). The simplicity of this technique lends itself to be applied to fast spin echo images commonly ordered as part of standard of care and does not require more advanced fat-water separated images.

Healthy versus pathological learning transferability in shoulder muscle MRI segmentation using deep convolutional encoder-decoders

Fully-automated segmentation of pathological shoulder muscles in patients with musculo-skeletal diseases is a challenging task due to the huge variability in muscle shape, size, location, texture and injury. A reliable automatic segmentation method from magnetic resonance images could greatly help clinicians to diagnose pathologies, plan therapeutic interventions and predict interventional outcomes while eliminating time consuming manual segmentation. The purpose of this work is three-fold. First, we investigate the feasibility of automatic pathological shoulder muscle segmentation using deep learning techniques, given a very limited amount of available annotated pediatric data. Second, we address the learning transferability from healthy to pathological data by comparing different learning schemes in terms of model generalizability. Third, extended versions of deep convolutional encoder-decoder architectures using encoders pre-trained on non-medical data are proposed to improve the segmentation accuracy. Methodological aspects are evaluated in a leave-one-out fashion on a dataset of 24 shoulder examinations from patients with unilateral obstetrical brachial plexus palsy and focus on 4 rotator cuff muscles (deltoid, infraspinatus, supraspinatus and subscapularis). The most accurate segmentation model is partially pre-trained on the large-scale ImageNet dataset and jointly exploits inter-patient healthy and pathological annotated data. Its performance reaches Dice scores of 82.4%, 82.0%, 71.0% and 82.8% for deltoid, infraspinatus, supraspinatus and subscapularis muscles. Absolute surface estimation errors are all below 83 mm² except for supraspinatus with 134.6 mm². The contributions of our work offer new avenues for inferring force from muscle volume in the context of musculo-skeletal disorder management.

Novel stochastic framework for automatic segmentation of human thigh MRI volumes and its applications in spinal cord injured individuals

Severe spinal cord injury (SCI) leads to skeletal muscle atrophy and adipose tissue infiltration in the skeletal muscle, which can result in compromised muscle mechanical output and lead to health-related complications. In this study, we developed a novel automatic 3-D approach for volumetric segmentation and quantitative assessment of thigh Magnetic Resonance Imaging (MRI) volumes in individuals with chronic SCI as well as non-disabled individuals. In this framework, subcutaneous adipose tissue, inter-muscular adipose tissue and total muscle tissue are segmented using linear combination of discrete Gaussians algorithm. Also, three thigh muscle groups were segmented utilizing the proposed 3-D Joint Markov Gibbs Random Field model that integrates first order appearance model, spatial information, and shape model to localize the muscle groups. The accuracy of the automatic segmentation method was tested both on SCI (N = 16) and on non-disabled (N = 14) individuals, showing an overall 0.93±0.06 accuracy for adipose tissue and muscle compartments segmentation based on Dice Similarity Coefficient. The proposed framework for muscle compartment segmentation showed an overall higher accuracy compared to ANTs and STAPLE, two previously validated atlas-based segmentation methods. Also, the framework proposed in this study showed similar Dice accuracy and better Hausdorff distance measure to that obtained using DeepMedic Convolutional Neural Network structure, a well-known deep learning network for 3-D medical image segmentation. The automatic segmentation method proposed in this study can provide fast and accurate quantification of adipose and muscle tissues, which have important health and functional implications in the SCI population.

Repeatability of Dixon magnetic resonance imaging and magnetic resonance spectroscopy for quantitative muscle fat assessments in the thigh

BACKGROUND

Changes in muscle fat composition as for example observed in sarcopenia or muscular dystrophy affect physical performance and muscular function, like strength and power. The purpose of the present study is to measure the repeatability of Dixon magnetic resonance imaging (MRI) for assessing muscle volume and fat in the thigh. Furthermore, repeatability of magnetic resonance spectroscopy (MRS) for assessing muscle fat is determined.

METHODS

A prototype 6-point Dixon MRI method was used to measure muscle volume and muscle proton density fat fraction (PDFF) in the left thigh. PDFF was measured in musculus semitendinosus of the left thigh with a T2-corrected multi-echo MRS method. For the determination of short-term repeatability (consecutive examinations), the root mean square coefficients of variation of Dixon MRI and MRS data of 23 young and healthy (29 ± 5 years) and 24 elderly men with sarcopenia (78 ± 5 years) were calculated. For the estimation of the long-term repeatability (13 weeks between examinations), the root mean square coefficients of variation of MRI data of seven young and healthy (31 ± 7 years) and 23 elderly sarcopenic men (76 ± 5 years) were calculated. Long-term repeatability of MRS was not determined.

RESULTS

Short-term errors of Dixon MRI volume measurement were between 1.2% and 1.5%, between 2.1% and 1.6% for Dixon MRI PDFF measurement, and between 9.0% and 15.3% for MRS. Because of the high short-term repeatability errors of MRS, long-term errors were not determined. Long-term errors of MRI volume measurement were between 1.9% and 4.0% and of Dixon MRI PDFF measurement between 2.1% and 4.2%.

CONCLUSIONS

The high degree of repeatability of volume and PDFF Dixon MRI supports its use to predict future mobility impairment and measures the success of therapeutic interventions, for example, in sarcopenia in aging populations and muscular dystrophy. Because of possible inhomogeneity of fat infiltration in muscle tissue, the application of MRS for PDFF measurements in muscle is more problematic because this may result in high repeatability errors. In addition, the tissue composition within the MRS voxel may not be representative for the whole muscle.

Muscle Tissue Labeling of Human Lower Limb in Multi-Channel mDixon MR Imaging: Concepts and Applications

With increasing resolutions and number of acquisitions, medical imaging more and more requires computer support for interpretation as currently not all imaging data is fully used. In our work, we show how multi-channel images can be used for robust air masking and reliable muscle tissue detection in the human lower limb. We exploit additional channels that are usually discarded in clinical routine. We use the common mDixon acquisition protocol for MR imaging. A series of thresholding, morphological, and connectivity operations is used for processing. We demonstrate our fully automated approach on four subjects and present a comparison with manual labeling. We discuss how this work is used for advanced and intuitive visualization, the quantification of tissue types, pose estimation, initialization of further segmentation methods, and how it could be used in clinical environments.

Computer-based assessment for facioscapulohumeral dystrophy diagnosis

The paper presents a computer-based assessment for facioscapulohumeral dystrophy (FSHD) diagnosis through characterisation of the fat and oedema percentages in the muscle region. A novel multi-slice method for the muscle-region segmentation in the T1-weighted magnetic resonance images is proposed using principles of the live-wire technique to find the path representing the muscle-region border. For this purpose, an exponential cost function is used that incorporates the edge information obtained after applying the edge-enhancement algorithm formerly designed for the fingerprint enhancement. The difference between the automatic segmentation and manual segmentation performed by a medical specialists is characterised using the Zijdenbos similarity index, indicating a high accuracy of the proposed method. Finally, the fat and oedema are quantified from the muscle region in the T1-weighted and T2-STIR magnetic resonance images, respectively, using the fuzzy c-mean clustering approach for 10 FSHD patients.

Automated segmentation and shape characterization of volumetric data

Characterization of complex shapes embedded within volumetric data is an important step in a wide range of applications. Standard approaches to this problem employ surface-based methods that require inefficient, time consuming, and error prone steps of surface segmentation and inflation to satisfy the uniqueness or stability of subsequent surface fitting algorithms. Here we present a novel method based on a spherical wave decomposition (SWD) of the data that overcomes several of these limitations by directly analyzing the entire data volume, obviating the segmentation, inflation, and surface fitting steps, significantly reducing the computational time and eliminating topological errors while providing a more detailed quantitative description based upon a more complete theoretical framework of volumetric data. The method is demonstrated and compared to the current state-of-the-art neuroimaging methods for segmentation and characterization of volumetric magnetic resonance imaging data of the human brain.

Manual segmentation of individual muscles of the quadriceps femoris using MRI: a reappraisal

PURPOSE

To propose a manual segmentation method for individual quadriceps femoris (QF) muscles and to test its reliability for muscle volume estimation.

MATERIALS AND METHODS

Images were acquired every 5 mm along the thigh using a 3T MRI scanner on 10 young (mean age: 25 years) and 10 older (mean age: 75 years) adults using a three-point 3D Dixon sequence. In each slice, anatomical cross-sectional areas of the individual quadriceps muscles of the dominant leg were outlined by two operators working independently. Differences between operators were assessed by means of Bland-Altman plots and intraclass correlation coefficients (ICC). This study was approved by the local Ethics Committee.

RESULTS

Precise delimitation of individual muscles along the femur often remains challenging, particularly near their insertion areas where some muscles may be partially or totally fused. There was, however, an excellent interoperator segmentation reliability despite a systematic significant difference between operators (ICC > 0.99), mainly due to delineation divergences. Considering all subjects and muscles, differences between operators were all lower than 4.4%.

CONCLUSION

This work has demonstrated the excellent reliability of manual segmentation to assess cross-sectional areas and therefore the volume of individual QF muscles using MRI. It may serve as a basis for a future segmentation consensus of the QF muscles.

Automated quantification of muscle and fat in the thigh from water-, fat-, and nonsuppressed MR images

PURPOSE

To introduce and validate an unsupervised muscle and fat quantification algorithm based on joint analysis of water-suppressed (WS), fat-suppressed (FS), and water and fat (nonsuppressed) volumetric magnetic resonance imaging (MRI) of the mid-thigh region.

MATERIALS AND METHODS

We first segmented the subcutaneous fat by use of a parametric deformable model, then applied centroid clustering in the feature domain defined by the voxel intensities in WS and FS images to identify the intermuscular fat and muscle. In the final step we computed volumetric and area measures of fat and muscle. We applied this algorithm on datasets of water-, fat-, and nonsuppressed volumetric MR images acquired from 28 participants.

RESULTS

We validated our tissue composition analysis against fat and muscle area measurements obtained from semimanual analysis of single-slice mid-thigh computed tomography (CT) images of the same participants and found very good agreement between the two methods. Furthermore, we compared the proposed approach with a variant that uses nonsuppressed images only and observed that joint analysis of WS and FS images is more accurate than the nonsuppressed only variant.

CONCLUSION

Our MRI algorithm produces accurate tissue quantification, is less labor-intensive, and more reproducible than the original CT-based workflow and can address interparticipant anatomic variability and intensity inhomogeneity effects.

Automatic quantification of muscle volumes in magnetic resonance imaging scans of the lower extremities

Muscle volume measurements are essential for an array of diseases ranging from peripheral arterial disease, muscular dystrophies, neurological conditions to sport injuries and aging. In the clinical setting, muscle volume is not routinely measured due to the lack of standardized ways for its repeatable quantification. In this paper, we present magnetic resonance muscle quantification (MRMQ), a method for the automatic quantification of thigh muscle volume in magnetic resonance imaging (MRI) scans. MRMQ integrates a thigh segmentation and nonuniform image gradient correction step, followed by feature extraction and classification. The classification step leverages prior probabilities, introducing prior knowledge to a maximum a posteriori classifier. MRMQ was validated on 344 slices taken from 60 MRI scans. Experiments for the fully automatic detection of muscle volume in MRI scans demonstrated an averaged accuracy, sensitivity and specificity for leave-one-out cross-validation of 88.3%, 93.6% and 87.2%, respectively.

Effect of exercise intervention on thigh muscle volume and anatomical cross-sectional areas--quantitative assessment using MRI

The objective of this study was to evaluate the location-specific magnitudes of an exercise intervention on thigh muscle volume and anatomical cross-sectional area, using MRI. Forty one untrained women participated in strength, endurance, or autogenic training for 12 weeks. Axial MR images of the thigh were acquired before and after the intervention, using a T1-weighted turbo-spin-echo sequence (10 mm sections, 0.78 mm in-plane resolution). The extensor, flexor, adductor, and sartorius muscles were segmented between the femoral neck and the rectus femoris tendon. Muscle volumes were determined, and anatomical cross-sectional areas were derived from 3D reconstructions at 10% (proximal-to-distal) intervals. With strength training, the volume of the extensors (+3.1%), flexors (+3.5%), and adductors (+3.9%) increased significantly (P < 0.05) between baseline and follow-up, and with endurance training, the volume of the extensor (+3.7%) and sartorius (+5.1%) increased significantly (P < 0.05). No relevant or statistically significant change was observed with autogenic training. The greatest standardized response means were observed for the anatomical cross-sectional area in the proximal aspect (10-30%) of the thigh and generally exceeded those for muscle volumes. The study shows that MRI can be used to monitor location-specific effects of exercise intervention on muscle cross-sectional areas, with the proximal aspect of the thigh muscles being most responsive.

Magnetic resonance imaging of the erector spinae muscles in Duchenne muscular dystrophy: implication for scoliotic deformities

Background

In Duchenne muscular dystrophy (DMD), the muscular degeneration often leads to the development of scoliosis. Our objective was to investigate how anatomical changes in back muscles can lead to scoliosis. Muscular volume and the level of fat infiltration in those muscles were thus evaluated, in non-scoliotic, pre-scoliotic and scoliotic patients. The overlying skin thickness over the apex level of scoliotic deformations was also measured to facilitate the interpretation of electromyographic signals when recorded on the skin surface.

Methods

In 8 DMD patients and two healthy controls with no known muscular deficiencies, magnetic resonance imaging (MRI) was used to measure continuously at 3 mm intervals the distribution of the erector spinae (ES) muscle in the T8-L4 region as well as fat infiltration in the muscle and overlying skin thickness: four patients were non-scoliotic (NS), two were pre-scoliotic (PS, Cobb angle < 15°) and two were scoliotic (S, Cobb angle ≥ 15°). For each subject, 63 images 3 mm thick of the ES muscle were obtained in the T8-L4 region on both sides of the spine. The pixel dimension was 0.39 × 0.39 mm. With a commercial software, on each 12 bits image, the ES contour on the left and on the right sides of the spine were manually determined as well as those of its constituents i.e., the iliocostalis (IL), the longissimus (LO) and the spinalis (SP) muscles. Following this segmentation, the surfaces within the contours were determined, the muscles volume were obtained, the amount of fat infiltration inside each muscle was evaluated and the overlying skin thickness measured.

Findings

The volume of the ES muscle of our S and PS patients was found smaller on the convex side relative to the concave one by 5.3 ± 0.7% and 2.8 ± 0.2% respectively. For the 4 NS patients, the volume difference of this muscle between right and left sides was 2.1 ± 1.5% and for the 2 controls, it was 1.4 ± 1.2%. Fat infiltration for the S and the PS patients was larger on the convex side than on the concave one (4.4 ± 1.6% and 4.5 ± 0.7% respectively) and the difference was more important near the apex. Infiltration was more important in the lateral IL muscle than in the medial SP and it was always larger near L2 than at any other spinal level. Fat infiltration was much more important in the ES for the DMD patients (49.9% ± 1.6%) than for the two controls (2.6 ± 0.8%). As for the overlying skin thickness measured near the deformity of the patients, it was larger on the concave than on the convex side: 14.8 ± 6.1 vs 13.5 ± 5.7 mm for the S and 10.3 ± 6.3 vs 9.8 ± 5.6 mm for the PS.

Interpretation

In DMD patients, our results indicate that a larger replacement of muscles fibers by fat infiltration on one side of the spine is a factor that can lead to the development of scoliosis. Efforts to slow such an infiltration on the most affected side of the spine could thus be beneficial to those patients by delaying the apparition of the scoliotic deformation. In addition to anatomical considerations, results obtained from the same patients but in experiments dealing with electromyography recordings, point to differences in the muscular contraction mechanisms and/or of the neural input to back muscles. This is similar to the adolescent idiopathic scoliosis (AIS) where a role of the nervous system in the development of the deformation has also been suggested.

Physical function, muscle strength and muscle mass in children on peritoneal dialysis

The aim of this study was to examine the physical function and muscle strength of children on peritoneal dialysis (PD) and to assess whether the muscle structure alterations influence physical function and muscle strength in these children. Twenty-two children on PD and 16 healthy children were enrolled into the study. A 6-min walk distance and gait speed tests were used to evaluate physical performance. Quadriceps muscle strength (QMS) was measured with a hand-held dynamometer. Magnetic resonance imaging was used to determine the cross-sectional area (CSA) and T2 signal intensity of the quadriceps muscle. Significant differences in the performance of these functional tests were found between PD patients and controls. Quadriceps muscle strength was significantly lower in PD patients than in controls. The CSA corrected for the body mass index (CSA/BMI) was not different between groups, whereas T2 signal intensity was significantly higher in PD patients than in the controls. Physical functioning tests and QMS had a close relationship with muscle CSA/BMI and with T2 signal intensity. In conclusion, along with the other previously documented mechanisms, increased fat in muscles may contribute to the decreased physical functioning and muscle strength in PD patients.

Reliability of tarsal bone segmentation and its contribution to MR kinematic analysis methods

The purpose of this study was to determine the reliability of tarsal bone segmentation based on magnetic resonance (MR) imaging using commercially available software. All tarsal bones of five subjects were segmented five times each by two operators. Volumes and second moments of volume were calculated and used to determine the intra- as well as interoperator reproducibility. The results show that these morphological parameters had excellent interclass correlation coefficients (>0.997) indicating that the presented tarsal bone segmentation is a reliable procedure and that operators are in fact interchangeable. The consequences on differences in MR kinematic analysis methods of segmentation due to repetition were also determined. It became evident that one analysis method--fitting surface point clouds--was considerable less affected by repeated segmentation (cuboid: up to 0.2 degrees, other tarsal bones up to 0.1 degrees) compared to a method using principal axes (cuboid up to 6.7 degrees, other tarsal bones up to 0.8 degrees). Thus, the former method is recommended for investigations of tarsal bone kinematics by MR imaging.

Segmentation of kidney from ultrasound B-mode images with texture-based classification

The segmentation of anatomical structures from sonograms can help physicians evaluate organ morphology and realize quantitative measurement. It is an important but difficult issue in medical image analysis. In this paper, we propose a new method based on Laws' microtexture energies and maximum a posteriori (MAP) estimation to construct a probabilistic deformable model for kidney segmentation. First, using texture image features and MAP estimation, we classify each image pixel as inside or outside the boundary. Then, we design a deformable model to locate the actual boundary and maintain the smooth nature of the organ. Using gradient information subject to a smoothness constraint, the optimal contour is obtained by the dynamic programming technique. Experiments on different datasets are described. We find this method to be an effective approach.

Segmentation of fascias, fat and muscle from magnetic resonance images in humans: the DISPIMAG software

Segmentation of human limb MR images into muscle, fat and fascias remains a cumbersome task. We have developed a new software (DISPIMAG) that allows automatic and highly reproducible segmentation of lower-limb MR images. Based on a pixel intensity analysis, this software does not need any previous mathematical or statistical assumptions. It displays a histogram with two main signals corresponding to fat and muscle, and permits an accurate quantification of their relative spatial distribution. To allow a systematic discrimination between muscle and fat in any subject, fixed boundaries were first determined manually in a group of 24 patients. Secondly, an entirely automatic process using these boundaries was tested by three operators on four patients and compared to the manual approach, showing a high concordance.

An investigation into the use of MR imaging to determine the functional cross sectional area of lumbar paraspinal muscles

The purpose of this study was to investigate the use of magnetic resonance (MR) imaging and image processing software to determine the functional cross-sectional area (FCSA) (the area of muscle isolated from fat) of the lumbar paraspinal muscles. The measurement of the morphology of the lumbar paraspinal muscles has become the focus of several recent investigations into the aetiology of low back pain. However, the reliability and validity of determining the FCSA of the lumbar paraspinal muscles using MR imaging are yet to be reported. T2 axial MR scans at the L1-S1 spinal levels of six subjects were obtained using identical MR systems and scanning parameters. Lean paraspinal muscle, vertebral body bone and intermuscular fat were manually segmented using image analysis software to assign a grey scale range to the MR signal intensity emitted by each tissue type. The resultant grey scale range for muscle was used to determine FCSA measurements for each of the paraspinal muscles, psoas, quadratus lumborum, erector spinae and lumbar multifidus on each scan slice. As various biological, instrument and measurement factors can affect MR signal intensity, a sensitivity analysis was conducted to determine the error associated in calculating FCSA for paraspinal muscle using a discrete grey scale range. Cross-sectional area and FCSA measurements were repeated three times and reliability indices for the FCSA measurements were obtained, showing excellent reliability, intra class correlation coefficient (mean=0.97, range 0.90-0.99) and %SEM (mean=2.6%, range 0.7-4.8%). In addition, the error associated with miscalculation of the grey scale range for the MR signal intensity of muscle was calculated and found to be low with an error of 20 grey scale units at the upper end of the muscle's grey scale range resulting in a very small error in the measured muscle FCSA. The method presented in this paper has a variety of practical applications in areas such as evidence-based rehabilitation, biomechanical modelling and the determination of segmental inertial parameters.