Muscle Cross-Sectional Area Segmentation in Transverse Ultrasound Images Using Vision Transformers

Abstract

Automatically measuring a muscle’s cross-sectional area is an important application in clinical practice that has been studied extensively in recent years for its ability to assess muscle architecture. Additionally, an adequately segmented cross-sectional area can be used to estimate the echogenicity of the muscle, another valuable parameter correlated with muscle quality. This study assesses state-of-the-art convolutional neural networks and vision transformers for automating this task in a new, large, and diverse database. This database consists of 2005 transverse ultrasound images from four informative muscles for neuromuscular disorders, recorded from 210 subjects of different ages, pathological conditions, and sexes. Regarding the reported results, all of the evaluated deep learning models have achieved near-to-human-level performance. In particular, the manual vs. the automatic measurements of the cross-sectional area exhibit an average discrepancy of less than 38.15 mm2, a significant result demonstrating the feasibility of automating this task. Moreover, the difference in muscle echogenicity estimated from these two readings is only 0.88, another indicator of the proposed method’s success. Furthermore, Bland–Altman analysis of the measurements exhibits no systematic errors since most differences fall between the 95% limits of agreements and the two readings have a 0.97 Pearson’s correlation coefficient (p < 0.001, validation set) with ICC (2, 1) surpassing 0.97, showing the reliability of this approach. Finally, as a supplementary analysis, the texture of the muscle’s visible cross-sectional area was examined using deep learning to investigate whether a classification between healthy subjects and patients with pathological conditions solely from the muscle texture is possible. Our preliminary results indicate that such a task is feasible, but further and more extensive studies are required for more conclusive results.

MyoVision-US: an Artificial Intelligence-Powered Software for Automated Analysis of Skeletal Muscle Ultrasonography

“…Using a deep learning image segmentation model known for performance, this study presents a relevant opportunity to improve the analysis of muscle ultrasound images with clinical and research applications. While previous studies have shown excellent reliability for muscle thickness, muscle CSA, muscle EI, fascicle length, and pennation angle measurements in ultrasound images of healthy muscles 23,24,46,47 , the present study adds findings with excellent inter-rater consistency for CSA, thickness, and EI of QC and TA muscles in patients with acute and chronic illness. This allowed the spectrum of muscle parameters to be increased, covering high and low values of thickness, CSA, and muscle EI.…”

The role of artificial intelligence in electrodiagnostic and neuromuscular medicine:Currentstate and future directions

“…One notable application is in the challenging task of muscle segmentation, by which DL models demonstrated near human-level accuracy in identifying muscle contours and automating the calculation of muscle CSA and echogenicity. 83 This could potentially enable further characterization of muscle disorders by detecting changes consistent with denervation, myositis, or dystrophic changes using different NMUS features. 84 DL-based segmentation has also been explored in other musculoskeletal tissues including tendons, bones, vertebral bodies, and discs, yielding substantial accuracy and facilitating the diagnosis and monitoring of various musculoskeletal injuries.…”

“…DL techniques have also demonstrated utility with NMUS muscle imaging. One notable application is in the challenging task of muscle segmentation, by which DL models demonstrated near human‐level accuracy in identifying muscle contours and automating the calculation of muscle CSA and echogenicity 83 . This could potentially enable further characterization of muscle disorders by detecting changes consistent with denervation, myositis, or dystrophic changes using different NMUS features 84 .…”

Neuromuscular Ultrasound in Intensive Care Unit-Acquired Weakness: Current State and Future Directions

“…In regard to NMUS, the application of DL has already been investigated with promising results. The parameters of muscle quantity (MLT and muscle CSA) and quality (ME and PA) in healthy and pathological conditions have been tested, with good to excellent accuracy compared to manual assessment by trained operators [ 86 , 87 , 88 , 89 , 90 ]. As increased ME can hamper a clear identification of muscle within surrounding tissue, the accuracy of DL in myopathic conditions with pathological elevated ME seems questionable.…”