Peter J. Harding

Ultrasound-Based Detection of Fasciculations in Healthy and Diseased Muscles

Involuntary muscle activations are diagnostic indicators of neurodegenerative pathologies. Currently detected by invasive intramuscular electromyography, these muscle twitches are found to be visible in ultrasound images. We present an automated computational approach for the detection of muscle twitches, and apply this to two muscles in healthy and motor neuron disease-affected populations. The technique relies on motion tracking within ultrasound sequences, extracting local movement information from muscle. A statistical analysis is applied to classify the movement, either as noise or as more coherent movement indicative of a muscle twitch. The technique is compared to operator identified twitches, which are also assessed to ensure operator agreement. We find that, when two independent operators manually identified twitches, higher interoperator agreement (Cohens k) occurs when more twitches are present (k = 0.94), compared to a lower number (k = 0.49). Finally, we demonstrate, via analysis of receiver operating characteristics, that our computational technique detects muscle twitches across the entire dataset with a high degree of accuracy (0.83 <; accuracy <; 0.96).

Real-Time Ultrasound Segmentation, Analysis and Visualisation of Deep Cervical Muscle Structure.

Despite widespread availability of ultrasound and a need for personalised muscle diagnosis (neck/back pain-injury, work related disorder, myopathies, neuropathies), robust, online segmentation of muscles within complex groups remains unsolved by existing methods. For example, Cervical Dystonia (CD) is a prevalent neurological condition causing painful spasticity in one or multiple muscles in the cervical muscle system. Clinicians currently have no method for targeting/monitoring treatment of deep muscles. Automated methods of muscle segmentation would enable clinicians to study, target, and monitor the deep cervical muscles via ultrasound. We have developed a method for segmenting five bilateral cervical muscles and the spine via ultrasound alone, in real-time. Magnetic Resonance Imaging (MRI) and ultrasound data were collected from 22 participants (age: 29.0±6.6, male: 12). To acquire ultrasound muscle segment labels, a novel multimodal registration method was developed, involving MRI image annotation, and shape registration to MRI-matched ultrasound images, via approximation of the tissue deformation. We then applied polynomial regression to transform our annotations and textures into a mean space, before using shape statistics to generate a texture-to-shape dictionary. For segmentation, test images were compared to dictionary textures giving an initial segmentation, and then we used a customized Active Shape Model to refine the fit. Using ultrasound alone, on unseen participants, our technique currently segments a single image in

Mutual Information for the Detection of Crush

{\approx } 0.45\text {s}

Deep Residual Networks for Quantification of Muscle Fiber Orientation and Curvature from Ultrasound Images

to over 86% accuracy (Jaccard index). We propose this approach is applicable generally to segment, extrapolate and visualise deep muscle structure, and analyse statistical features online.

Validation of a New Semi-Automated Technique to Evaluate Muscle Capillarization.

Fatal crush conditions occur in crowds with tragic frequency. Event organizers and architects are often criticised for failing to consider the causes and implications of crush, but the reality is that both the prediction and prevention of such conditions offer a significant technical challenge. Full treatment of physical force within crowd simulations is precise but often computationally expensive; the more common method of human interpretation of results is computationally “cheap” but subjective and time-consuming. This paper describes an alternative method for the analysis of crowd behaviour, which uses information theory to measure crowd disorder. We show how this technique may be easily incorporated into an existing simulation framework, and validate it against an historical event. Our results show that this method offers an effective and efficient route towards automatic detection of the onset of crush.

The application of deep convolutional neural networks to ultrasound for modelling of dynamic states within human skeletal muscle.

This paper concerns fully automatic and objective measurement of human skeletal muscle fiber orientation directly from standard b-mode ultrasound images using deep residual (ResNet) and convolutional neural networks (CNN). Fiber orientation and length is related with active and passive states of force production within muscle. There is currently no non-invasive way to measure force directly from muscle. Measurement of forces and other contractile parameters like muscle length change, thickness, and tendon length is not only important for understanding healthy muscle, but such information has contributed to understanding, diagnosis, monitoring, targeting and treatment of diseases ranging from myositis to stroke and motor neurone disease (MND). We applied well established deep learning methods to ultrasound data recorded from 19 healthy participants (5 female, ages: 30 ± 7.7) and achieved state of the art accuracy in predicting fiber orientation directly from ultrasound images of the calf muscles. First we used a previously developed segmentation technique to extract a region of interest within the gastrocnemius muscle. Then we asked an expert to annotate the main line of fiber orientation in \( 4 \times 4 \) partitions of 400 normalized images. A linear model was then applied to the annotations to regulate and recover the orientation field for each image. Then we applied a CNN and a ResNet to predict the fiber orientation in each image. With leave one participant out cross-validation and dropout as a regulariser, we were able to demonstrate state of the art performance, recovering the fiber orientation with an average error of just \( 2^\circ \).

A technical note on variable inter-frame interval as a cause of non-physiological experimental artefacts in ultrasound

The method of capillary domains has often been used to study capillarization of skeletal and heart muscle. However, the conventional data processing method using a digitizing tablet is an arduous and time-consuming task. Here we compare a new semi-automated capillary domain data collection and analysis in muscle tissue with the standard capillary domain method. The capillary density (1481±59 vs. 1447±54 caps mm(-2); R2:0.99; P<0.01) and heterogeneity of capillary spacing (0.085±0.002 vs. 0.085±0.002; R2:0.95; P<0.01) were similar in both methods. The fiber cross-sectional area correlated well between the methods (R2:0.84; P<0.01) and did not differ significantly (~8% larger in the old than new method at P=0.08). The latter was likely due to differences in outlining the contours between the two methods. In conclusion, the semi-automated method gives quantitatively and qualitatively similar data as the conventional method and saves a considerable amount of time.

Mutual information based gesture recognition

This paper concerns the fully automatic direct in vivo measurement of active and passive dynamic skeletal muscle states using ultrasound imaging. Despite the long standing medical need (myopathies, neuropathies, pain, injury, ageing), currently technology (electromyography, dynamometry, shear wave imaging) provides no general, non-invasive method for online estimation of skeletal intramuscular states. Ultrasound provides a technology in which static and dynamic muscle states can be observed non-invasively, yet current computational image understanding approaches are inadequate. We propose a new approach in which deep learning methods are used for understanding the content of ultrasound images of muscle in terms of its measured state. Ultrasound data synchronized with electromyography of the calf muscles, with measures of joint torque/angle were recorded from 19 healthy participants (6 female, ages: 30 ± 7.7). A segmentation algorithm previously developed by our group was applied to extract a region of interest of the medial gastrocnemius. Then a deep convolutional neural network was trained to predict the measured states (joint angle/torque, electromyography) directly from the segmented images. Results revealed for the first time that active and passive muscle states can be measured directly from standard b-mode ultrasound images, accurately predicting for a held out test participant changes in the joint angle, electromyography, and torque with as little error as 0.022°, 0.0001V, 0.256Nm (root mean square error) respectively.

Blunted angiogenesis and hypertrophy are associated with increased fatigue resistance and unchanged aerobic capacity in old overloaded mouse muscle.

Ultrasound (US) imaging is a well-recognized technique for the study of static tissues but its suitability for studying tissue dynamics depends upon accurate frame time information, which may not always be available to users. Here we present methods to quantify the inter-frame interval (IFI) variability, and evaluate different procedures for collecting temporal information from two US-imaging devices. The devices tested exhibited variable IFIs that could only be confirmed by direct measures of timing signals, available by means of electrical signals (triggers) and/or temporal information contained in the software used for the US data collection. Interpolating frame-by-frame measures of dynamic changes within image sequences using individual IFI values provided improved synchronization between measures of skeletal muscle movement and activation; validating US as a valuable technique for the study of musculoskeletal tissue dynamics, when correctly implemented.

An early warning method for crush

Proliferation of gestural interfaces necessitates the creation of robust gesture recognition systems. A novel technique using Mutual Information to classify gestures in a recognition system is presented. As this technique is based on well-known information theory metrics the underlying operation is not as complex as many other techniques which allows for this technique to be easily implemented. A high recognition rate of 98.55% was achieved, with recognition occurring in under 10ms.