Semi-automated Tracing of Hamstring Muscle Architecture for B-mode Ultrasound Images

Fully Automated Analysis of Muscle Architecture from B-Mode Ultrasound Images with DL_Track_US

OBJECTIVE

B-mode ultrasound can be used to image musculoskeletal tissues, but one major bottleneck is analyses of muscle architectural parameters (i.e., muscle thickness, pennation angle and fascicle length), which are most often performed manually.

METHODS

In this study we trained two different neural networks (classic U-Net and U-Net with VGG16 pre-trained encoder) to detect muscle fascicles and aponeuroses using a set of labeled musculoskeletal ultrasound images. We determined the best-performing model based on intersection over union and loss metrics. We then compared neural network predictions on an unseen test set with those obtained via manual analysis and two existing semi/automated analysis approaches (simple muscle architecture analysis [SMA] and UltraTrack). DL_Track_US detects the locations of the superficial and deep aponeuroses, as well as multiple fascicle fragments per image.

RESULTS

For single images, DL_Track_US yielded results similar to those produced by a non-trainable automated method (SMA; mean difference in fascicle length: 5.1 mm) and human manual analysis (mean difference: -2.4 mm). Between-method differences in pennation angle were within 1.5°, and mean differences in muscle thickness were less than 1 mm. Similarly, for videos, there was overlap between the results produced with UltraTrack and DL_Track_US, with intraclass correlations ranging between 0.19 and 0.88.

CONCLUSION

DL_Track_US is fully automated and open source and can estimate fascicle length, pennation angle and muscle thickness from single images or videos, as well as from multiple superficial muscles. We also provide a user interface and all necessary code and training data for custom model development.

Quantitative Muscle Fascicle Tractography Using Brightness-Mode Ultrasound

A muscle's architecture, defined as the geometric arrangement of its fibers with respect to its mechanical line of action, impacts its abilities to produce force and shorten or lengthen under load. Ultrasound and other noninvasive imaging methods have contributed significantly to our understanding of these structure-function relationships. The goal of this work was to develop a MATLAB toolbox for tracking and mathematically representing muscle architecture at the fascicle scale, based on brightness-mode ultrasound imaging data. The MuscleUS_Toolbox allows user-performed segmentation of a region of interest and automated modeling of local fascicle orientation; calculation of streamlines between aponeuroses of origin and insertion; and quantification of fascicle length, pennation angle, and curvature. A method is described for optimizing the fascicle orientation modeling process, and the capabilities of the toolbox for quantifying and visualizing fascicle architecture are illustrated in the human tibialis anterior muscle. The toolbox is freely available.

The architectural characteristics of the hamstring muscles do not differ between male and female elite-level rugby union players

Purpose: To determine whether differences exist in the architectural characteristics of the hamstring muscles of elite-level male and female rugby union players. Methods: Forty elite-level rugby union players (male n = 20, female n = 20) participated in this cross-sectional study. A sonographer acquired static ultrasound images using a 92 mm linear transducer to quantify (via a semi-automated tracing software tool) the architectural characteristics (muscle length, fascicle length, pennation angle, and muscle thickness) of the biceps femoris long head and semimembranosus muscles of participants' left limb. Muscle length and muscle thickness of the biceps femoris short head and semitendinosus muscles of participants' left limb were also quantified. Bonferroni adjusted independent samples t-tests were performed to evaluate whether differences exist in the architectural characteristics of the hamstring muscles of elite-level male and female rugby union players. Results: There were no significant differences in fascicle length or pennation angle of the hamstring muscles of elite-level male and female rugby union players. Some significant differences in muscle thickness (biceps femoris short head, and semimembranosus) and muscle length (biceps femoris short head, biceps femoris long head, semitendinosus, and semimembranosus) were observed; in all cases the male players had thicker and longer muscles. Conclusion: At a group level, hamstring muscle fascicle length and pennation angle are unlikely to be a sex-specific intrinsic risk factor for Hamstring strain injuries.

Reliability and accuracy of ultrasound image analyses completed manually versus an automated tool

Analysis of Brightness-mode ultrasound-captured fascicle angle (FA) and fascicle length (FL) can be completed manually with computer-based programs or by automated programs. Insufficient data exists regarding reliability and accuracy of automated tools. Therefore, the purpose of this study was to determine the test-retest reliability of automatic and manual ultrasound analyses, while determining accuracy of the automatic tool against the manual equivalent. Twenty-three participants (mean ± SD; age = 24 ± 4 years; height = 172.2 ± 10.5 cm; body mass = 73.1 ± 16.1 kg) completed one laboratory visit consisting of two trials where vastus lateralis muscle architecture was assessed with ultrasound. Images were taken at both lower (10 MHz) and higher frequency (12 MHz). Images were analyzed manually in an open-source imaging program and automatically using a separate open-source macro function. Test-retest reliability statistics were calculated for automatic and manual analyses. Accuracy was determined with validity statistics and were calculated for automatic analyses. The results show that manual ultrasound analyses for FA and FL for both lower and higher frequency displayed good reliability (ICC_2,1 = 0.75-0.86). However, automatic ultrasound analyses for FA and FL revealed moderate reliability (ICC_2,1 = 0.61-0.72) for the lower frequency images and poor reliability (ICC_2,1 = 0.16-0.27) for higher frequency images. When assessed against manual techniques, automatic analyses presented greater total error (TE) and standard error of the estimate (SEE) for FA at lower frequency (constant error (CE) = -3.91°, TE = 5.57°, SEE = 3.45°) than higher (CE = -2.78°, TE = -4.54°, SEE = 2.45°). For FL, the higher frequency error (CE = 0.92 cm, TE = 2.12 cm, SEE = 1.15 cm) was similar to lower frequency error (CE = 1.98 cm, TE = 3.66 cm, SEE = 1.57 cm). The findings overall show that manual analyses had good reliability and low absolute error, while demonstrating the automated counterpart had poor to moderate reliability and large errors in analyses. These findings may be impactful as they highlight the good reliability and low error associated with manually analyzed ultrasound images and validate a novel automatic tool for analyzing ultrasound images. Future work should focus on improving reliability and decreasing error in automated image analysis tools. Automated tools are promising for the field as they eliminate biases between analysts and may be more time efficient than manual techniques.