Musculoskeletal ultrasound: a technical and historical perspective

During the past four decades, Musculoskeletal Ultrasound (MSKUS) has become popular as an imaging modality, particularly outside the United States, due to its low cost, accessibility, and lack of ionizing radiation. The development of MSKUS has paralleled technical advances in diagnostic ultrasound, from the earliest images displaying bistable images of musculoskeletal anatomy to its current state, where exquisite images of muscles, tendons, nerves, ligaments and joints can be obtained^(1–6). With ongoing technical improvements in the quality of grayscale imaging along with some newer applications, such as shear wave elastography, enabling functional assessment of the musculoskeletal soft tissues, MSKUS is becoming ubiquitous^(7–10).

The compactness of some of the currently available scanners makes this modality conducive to bedside imaging as well as imaging in remote settings^(11–13) (Fig. 1). One of the most popular applications of MSKUS has been to provide guidance for a large variety of interventional procedures, since it allows visualization of the relevant anatomy and needle position in real time⁽¹⁴⁾ (Fig. 2, Fig. 3). In addition to radiology, the modality has opened itself up to wider usage by a large variety of clinical subspecialties involved in treating patients with musculoskeletal diseases, as well as anesthesiologists/ pain management physicians who perform nerve blocks, therapeutic or ablative procedures of peripheral nerve lesions^(4–8,10).

Fig. 1.

Fig. 2.

Fig. 3.

Ultrasound technology has undergone extraordinary growth since its inception in the 1940s^(15–18). A brief description of technological developments in ultrasound allows a greater perspective as to the timing when various MSKUS applications first appear. The development of ultrasound technology was possible in large part due to concomitant advances in both solid-state electronics and signal processing. The invention of the transistor and digital computer in the late 1940s was integral in the development of ultrasound technology^(18,19). Gordon Moore in 1964 predicted that the number of microprocessors on a chip would grow exponentially over time. This growth would result in progressive miniaturization of chip design and, therefore, increased computational power.

One of the earliest presentations of ultrasound data employed pulse echo techniques, derived from earlier radar applications. This involved the display of individual A-line data, assuming a fixed speed of sound in soft tissue⁽¹⁵⁾. Each A-line represented the intensity of acoustic backscatter at various depths, adjusted for attenuation. Placing this in a time-varying display allows the display of M-mode data, typically utilized in cardiology⁽¹⁵⁾.

Musculoskeletal soft tissues typically present strong reflecting surfaces (specular reflectors) along with complex internal morphology. A unique feature of many of these tissues is an internal fibrillar architecture, most prominent in tendons⁽²⁰⁾, as well as a linear orientation. One of the most useful features of MSKUS is its real-time nature, allowing the performance of provocative maneuvers and providing real-time guidance during an intervention^(21–23). The ability to visualize features that are useful for musculoskeletal imaging entails two requirements: the display of the back-scattered data as a 2-D anatomic image and rapid acquisition/display (15–20 frames per second) of the image data. The first real-time 2-D scanners were introduced in the late 1960s, involving one or two crystals undergoing translation or rotation to produce a composite bistable image^(15–17,24) (Fig. 4).

Fig. 4.

Ultrasound did not become a widely accepted diagnostic tool, though, until the early 1970s, when grayscale ultrasound, in which non-linear echo amplitudes are mapped to gray levels, became the method of choice to display image data^(15–17). The development of multi-element linear phased array transducers, employing digital image processing, allowed the display of 2-D grayscale images. Duplex ultrasound scanners began to appear during this period, allowing the imaging of anatomy and the measurement of blood flow in a single scanner⁽¹⁸⁾. These systems appeared in the late 1970s/ early 1980s. Some of the earliest images of extremity anatomy, hematomas in muscle, large joint effusions/synovitis appeared as extended-field-of-view (EFOV) B-mode images^(25–32). In 1985, the first color Doppler flow-mapping system that combined Doppler flow imaging in color with B-mode imaging in grayscale was introduced^(15–17).

By 1980, commercial real-time phased array imaging systems were made possible by developments in video microprocessors, digital memory, and the miniaturization offered by programmable integrated circuits. During the 1980s, multi-element array systems allowed rapid growth of transducer technology. Improved materials and piezoelectric composites enabled the production of arrays with several hundred elements, operating at higher frequencies and wider bandwidths. This enhanced flexibility allowed the imaging and operation of other modes within the same transducer at multiple frequencies selectable by the user⁽¹⁸⁾.

With the appearance of these new systems in the 1980s, the applications to the MSK system started to appear at a greater frequency in the imaging literature. Examination of articular cartilage, joint effusions, synovitis, tendon and muscle abnormalities, and nerve imaging using grayscale ultrasound were described, including the first descriptions of tendon morphology and anisotropy^{(20–22,32–46}). Multiple anatomic locations were accessible to this modality, resulting in studies of the Achilles tendon, patellar tendon, and soon also more complex anatomical areas such as the rotator cuff^(22,39–40) (Fig. 5, Fig. 6, Fig. 7). Early assessment of muscle hematomas and soft tissue masses using bistable techniques expanded to more sophisticated descriptions of muscle and other soft tissue pathology^{(26–30,33,36)}. Large joints, such as the knee and hip, were ultimately expanded to include small joint pathology, such as in the hands and feet^(41,45). The real-time capabilities provide a method to assess static anatomy as well as pathology that manifests when accentuated by some maneuver. Examples include neonatal hip examination, assessment of joint effusions as well as examination of rotator cuff impingement^(38,42–43).

Fig. 5.

Fig. 6.

Fig. 7.

By the 1990s, developments enabling more rapid image acquisition and the availability of low cost analog to digital (A/D) chips improved computational capabilities, resulting in faster image processing in smaller devices that could be assembled at a lower cost⁽¹⁸⁾. Imaging systems incorporating these advances evolved into digital architectures and beam formers, permitting beam steering and electronic focusing. The most dramatic changes have been through the continual miniaturization of electronics in accordance with Moore’s law⁽¹⁹⁾. Smaller size components led to the first commercially available phased array imaging systems as well as new portable imaging systems that weigh only a few pounds⁽¹³⁾ (Fig. 1). The current generation of scanners include 1½ and 2-dimensional arrays allowing improved in- and out-of-plane (elevational) focusing, rapid image processing, and 3-D and 4-D imaging (real-time volumetric imaging)^(47–52) (Fig. 7, Fig. 8).

Fig. 8.

The potential utility of ultrasound in evaluating the musculoskeletal system evolved to its current state through a series of advances spanning the 1980s through the first decade of the 21^st century. Many of these features are incorporated in present-day clinical scanners, including assessment of soft tissue vascularity, improvements in anatomic depiction of the relevant musculoskeletal anatomy, and enhanced characterization of soft tissue composition.

Assessment of vascularity plays an important role in musculoskeletal imaging. Inflammation, repair processes, and a variety of musculoskeletal tumors display diverse degrees of vascularity which can serve as a measure of disease activity and/or response to a therapeutic intervention.

Doppler techniques date back to the 1950s, and began to play an important clinical role once incorporated as part of a duplex imaging package within the ultrasound scanner⁽¹⁶⁾. Vascular assessment was further enhanced with the introduction of color Doppler in the 1980s, in which the mean frequency Doppler shift within an interrogated pixel could be color-encoded to reflect mean velocity and direction, producing an anatomic map of vascularity. A series of adjustable parameters and internal discriminators were available to help differentiate true flow from artifact due to low-frequency soft tissue motion. Color Doppler continues to be an important tool in depicting high flow states.

In the musculoskeletal system, vascularity is characteristically low flow and often poly-directional. The limitations associated with color Doppler maps include aliasing and drop out of signal when insonating at 90 degrees, as per the Doppler equation. Increasing color gain to improve flow sensitivity has the effect of introducing low-amplitude noise in the color-encoded image. Many of these issues were resolved with the recognition that flow sensitivity could be improved by encoding the demodulated Doppler signal intensity in color, with the assumption that the noise floor is of low amplitude⁽⁵³⁾ (Fig. 9).

Fig. 9.

Power Doppler (PD), as it has become known, described in the mid-1990s, significantly improved sensitivity to low-flow states, making it ideal for musculoskeletal imaging^(54–58). PD is not subject to aliasing artifact and it is not as sensitive to the insonation angle. PD was established to be a good indicator of hyperemic inflammatory states, capable of depicting blood flow in soft tissues more readily than conventional color Doppler (Fig. 10). PD has, in fact, become an important tool in assessing disease activity in inflammatory arthritis⁽⁵⁸⁾. An added bonus is that PD is more amenable to semi-quantitative estimates of vascularity, noting that it can provide a measure of fractional moving blood volume⁽⁵⁹⁾. A fundamental problem with PD, however, is the presence of so-called blooming artifact, due to high-amplitude low-frequency motion at tissue interfaces (clutter)⁽⁶⁰⁾.

Fig. 10.

In the current generation of high-end ultrasound scanners, blooming artifact is significantly improved by employing newer clutter cancelling techniques, which are computationally intensive, requiring high-frame rates and long ensemble lengths^(60–62). These adaptive filters have a number of different names (slow flow, microvascular flow, etc.), but they generally involve separation of the weakly scattering blood flow from low-frequency motion of strong reflecting tissue boundaries. These microvascular flow techniques are stillbeing investigated, but they appear to improve flow sensitivity and depiction of vascular anatomy, while eliminating blooming artifact^(61–62) (Fig. 11).

Fig. 11.

The most significant improvement in flow detection, however, occurs with the inclusion of ultrasound contrast agents, which are true capillary imaging agents. Contrast agents have long been investigated in ultrasound, dating back to 1968, with the observation that microbubbles produce strong reflections on ultrasound⁽⁶³⁾. This led to intensive research in micro-bubble contrast agents extending into the 1970s–80s, with the first clinically available agent released in 1984. The current (second) generation of contrast agents exhibit a good safety profile and have been used extensively in cardiology and abdominal imaging applications⁽⁶⁴⁾.

Present-day contrast agents are encapsulated microbubbles (1–10 microns in size) that undergo strong backscatter at certain resonant frequencies and are readily detectable, using harmonic imaging (discussed below)⁽⁶⁵⁾ (Fig. 12). These agents are injected intravenously, pass through the pulmonary capillary bed, and then appear in the arterial phase. The bubbles have a half-life of several minutes and the gas is ultimately exhaled in the lungs, while the outer shell is metabolized⁽⁶⁴⁾.

Fig. 12.

Contrast agents have been used in the imaging of the musculoskeletal system in a limited fashion over the last two decades. They have been shown to allow robust quantification of vascularity and blood flow, but fall into the category of off-label usage in the United States. Contrast has allowed remarkably improved flow sensitivity in the detection and quantification of hyperemic states, such as inflammatory myopathies, inflammatory arthritis, tendinosis and tendon repairs^(66–72) (Fig. 13, Video 1). In addition to improved flow sensitivity and flow quantification, contrast agents may serve as therapeutic delivery agents in the future⁽⁶⁴⁾.

Fig. 13.

The advent of grayscale imaging has allowed the assessment of internal tissue morphology⁽¹⁶⁾. A feature of grayscale images is the appearance of an inherent granular noise within the image, known as speckle⁽⁷³⁾ (Fig. 14). The latter arises due to small inhomogeneities (e.g. variation in acoustic impedance) within the soft tissues that scatter ultrasound but are in and of themselves too small to resolve. Within a resolution element, there can be many such scatterers, which can only be characterized in a statistical sense^(74,75). Since these scatterers are fixed in space, small translations, rotations or compressions by the ultrasound transducers may contain some degree of correlation to the previous transducer position⁽⁷⁶⁾ (Fig. 14).

Fig. 14.

One method to achieve improved image quality, therefore, is to reduce the degree of unwanted speckle within the image. There are a variety of ways to accomplish this, one being temporal averaging (i.e. changing persistence), which has the effect of smoothing the image, while reducing spatial resolution and decreasing the frame rate^(77–81). One of the most successful methods to reduce speckle, which has significantly impacted musculoskeletal imaging, is spatial compounding using speckle decorrelation^(81–82) (Fig. 15). In this latter technique, image data from multiple look directions (as many as eight) are averaged, producing a speckle-free image with improved signal-to-noise ratio and an anatomically improved image. The one caveat in using spatial compounding is that certain artifacts which are sometimes diagnostically useful, such as posterior acoustic enhancement or refraction artifact, are diminished.

Fig. 15.

Speckle has played an important role in a number of applications that have impacted MSKUS. These applications utilize the fact that images derived from small changes in transducer positioning may remain correlated. Using this correlated motion gave rise to speckle tracking algorithms (late 1980s–early 1990s), which in turn provided an accurate method to produce in-plane image registration^(76,83). Examples include improved depiction of muscle and tendon pathology, selected soft tissue neoplasms and, in some cases, neural pathology.

Extended field-of-view (EFOV) imaging entails in-plane speckle tracking while manually translating the transducer to allow consecutive images to be combined into a single image depicting a larger anatomic region^(84–86) (Fig. 16). This has led to improved depiction of pathology relative to adjacent anatomy as part of a large field-of-view image (up to approximately 60 cm). The advantage of EFOV imaging resides more in visualizing and describing an abnormal finding to clinicians or other imagers not present during the initial real-time assessment. Examples include depicting muscle and tendon tears, which are shown in relation to normal adjacent anatomy as well as allowing the full extent of an abnormality to be displayed (Fig. 16). Obtaining these images remains somewhat of an art, though, since the tissue of interest is tracked using a free-hand technique. Poor speckle tracking and complex anatomy are potential complicating factors.

Fig. 16.

Ultrasound elastography was initially described as an extension of soft tissue palpation, albeit in situations where an abnormality might not be clinically evident. The wide dynamic range of soft tissue mechanical properties is conducive to this type of imaging⁽⁸⁷⁾. Thus, while research regarding soft tissue mechanical properties has been around since the 1970s, it was not until the 1990s that “sonopalpation” or compression-based elastography became clinically available^(88–90). This is a free-hand technique, whereby small compressions of soft tissue by the transducer result in small axial displacements that can be tracked using speckle tracking techniques. A strain map is obtained that is depth-dependent. The contrast derived in such a map is largely determined by the relative hardness of different tissues at the same depth⁽⁹⁰⁾. While the technique is well-suited to small low-contrast soft tissue masses, it is less suitable for global tissue properties. Nevertheless, it has been used successfully in a variety of musculoskeletal applications, particularly ten-dinosis^(91–92) (Fig. 17). There is an art to achieving optimal imaging by knowing what part of the compression/relaxation cycle to obtain measurements from and how light a touch is required. Most scanners have built-in tools to optimize the level of compression used, optimally producing 1–2% strain. A color map is generally employed, in which red denotes softer objects (greater strain or more deformable), while blue hues denote low strain (stiffer objects). The addition of a reference of known Young’s modulus has been shown to be helpful in calculating relative strain in a manner that is more robust and less operator-dependent⁽⁹³⁾.

Fig. 17.

Another variant of elasticity imaging refers to the production of a shear wave either through direct mechanical means (transient elastography) or using the transducer itself to produce a push pulse (acoustic radiation force impulse imaging or ARFI)^(94–98). Historically, Doppler techniques have also been employed to detect the induced shear waves in transient elastography⁽⁹⁶⁾. In either case, the transducer also serves to track the resultant shear wave propagation using a series of tracker pulses and speckle tracking to determine axial displacement at each element location (Fig. 18). The speed of the shear wave (measured in meters per second) reflects the local shear elastic (or equivalently Young’s) modulus with the assumption that soft tissues are incompressible. While transient elastography may more accurately reflect the elastic moduli, ARFI techniques have been implemented in most high-end clinical scanners. The shear waves produced are generally bandwidth limited between approximately 100–1000 Hz, and the group velocity of the induced wave packet is estimated, which is then used to determine the Young’s modulus⁽⁹⁰⁾.

Fig. 18.

Shear wave elastography (SWE) has a range of advantages: it is easy to implement, less operator-dependent, and produces a quantitative measure or parametric map of soft tissue stiffness. In this respect, SWE has been useful to produce more global and quantitative soft tissue assessment. SWE has been used to study tendon, ligament, joint and muscle abnormalities with encouraging preliminary results, although large controlled series are still lacking^(99–105) (Fig. 19).

Fig. 19.

A limitation of SWE is the lack of uniformity across different scanners, or when using different transducers within the same scanner. The parametric images are of low spatial resolution due to the method used to track wave propagation. Two additional problems also arise in shear wave imaging, which have not as yet been addressed by vendors. Musculoskeletal soft tissues are inherently anisotropic, so that measurements of shear elastic or Young’s modulus are directionally dependent^(97,100,106). Two-dimensional tracking techniques, using matrix arrays, may ultimately deal with this issue by displaying the angle-dependent two-dimensional propagation patterns following a cylindrically symmetric push pulse. Secondly, the shear wave speed is frequency-dependent at the higher end of the frequency spectrum due to viscoelastic effects^(100,106). This is not generally taken into consideration in any clinically available system, although it is an area of ongoing research.

Much of ultrasound is based on the assumption of a fixed speed of sound in soft tissue. An approximate plane wave of a given frequency reflects off a surface and is reflected back to the transducer, a coherent process, or the wave can undergo scattering by point-like scatterers, which is an incoherent process. This is complicated by attenuation, which exponentially diminishes the pressure amplitude with a decay coefficient proportional to frequency. To a first approximation, many of these processes can be understood as being linear: an example being scattering of a plane wave from a point scatterer. The propagation of sound in biological soft tissues, however, is inherently non-linear, particularly at high enough pressure amplitudes. During the compressional phase, the speed of sound increases slightly, whereas during relaxation the speed of sound will decrease. The extent to which this non-linearity should be taken into account is determined by the non-linearity factor, B/A^(107,108).

When B/A is sufficiently large, higher order harmonics will be generated; the wave is no longer represented as a simple plane wave but rather as a summation of waves consisting of higher order terms, each propagating at some multiple of the original (or fundamental) frequency. These higher order terms can be scattered back to the transducer (Fig. 20). Detection of these lower amplitude harmonics can have a number of distinct advantages, such as improved signal-to-noise ratio, contrast, and axial and lateral resolution⁽¹⁰⁹⁾. Harmonic imaging generally refers to second harmonic detection (propagates at twice the fundamental frequency) and imaging. A broadband transducer operating at a lower center frequency (i.e., 6 MHz) can transmit into the soft tissue, being less susceptible to attenuation. The second harmonic frequency (12 MHz) can be detected to produce a higher axial resolution image (Fig. 21), which is less attenuated at the higher frequency due to the shortened propagation distance. Tissue harmonic imaging (THI) is particularly well-suited to deeper structures, such as the supraspinatus tendon, muscle or interdigital web spaces, where subcutaneous fat can be problematic. In scanning very superficial structures, THI is less of an advantage⁽¹¹⁰⁾.

Fig. 20.

Fig. 21.

Second-harmonic imaging has become the method of choice when used in conjunction with contrast agents⁽¹¹¹⁾. Bubbles oscillate at a resonant frequency corresponding to the appropriate insonating frequency (fundamental), to generate higher order harmonics in the backscattered wave. The corresponding waves are of significantly higher amplitude (orders of magnitude) in comparison to back-scattered waves that derive from scatterers within the soft tissue. In this manner, areas of enhancement are selectively imaged. There are a variety of methods used to detect these harmonic waves, including frequency selection, pulse inversion or amplitude modulation⁽¹¹¹⁾. The resulting image displays the distribution of microbubbles over time on a capillary level, allowing quantitative estimates of soft tissue perfusion and blood volume (Fig. 14, Video 1).

With the appearance of real-time grayscale imaging in the late 1970s/early 1980s, and noting the conspicuity of metallic needles relative to adjacent soft tissue anatomy, a large variety of procedures have been shown to be readily performed under ultrasound guidance^(17,112). Since fluid is of high contrast on ultrasound, the earliest musculoskeletal applications (1980s/1990s) included joint and cyst aspirations^(113–115). Subsequently, ultrasound guidance was shown to be useful in performing a large variety of musculoskeletal interventions, including bursal, tendon sheath and joint injections, and in the treatment of calcific tendinosis^(14,116). In recent years, ultrasound has been also used extensively in perineural blocks and therapeutic injections as well as for localization for ablative therapy, such as alcohol, RF and cryoablations^(117,118) (Fig. 2, Fig. 3).

The majority of technical developments during this period relate to the compactness of the ultrasound scanners and transducers, which have become more portable along with the availability of smaller footprint transducers (e.g., hockey stick) and higher operating frequencies^(15–17). Efforts to increase needle conspicuity, such as beam steering and echo tip needles, have been developed to improve confidence during needle positioning.

Since many procedures are performed in patients who already have pre-existing imaging in the form of computed tomography (CT) or magnetic resonance imaging (MRI), image co-registration capabilities have been added to more advanced ultrasound systems over the last 10–15 years⁽¹¹⁹⁾ (Fig. 22). These largely entail the use of an electromagnetic field generator to localize the transducer location and orientation in space, identifying common fiduciary markers between both sets of imaging data (ultrasound/CT or ultrasound/ MRI) and using the least squares techniques to calculate the appropriate mathematical transformation to superimpose both imaging data sets. Likewise, a transmitter can be placed on a needle to localize it in the image space during positioning. One can then effectively scan using either ultrasound or CT/MRI in a real-time setting to identify a target area for injection, aspiration or biopsy^(120–121). While these systems have proven to be useful in a variety of musculoskeletal procedures, where the target is not readily visualized on a baseline ultrasound, they continue to be somewhat cumbersome and potentially expensive for routine applications.

Fig. 22.

Applications of AI to the musculoskeletal system have received a great deal of attention in the last decade⁽¹²²⁾. These techniques have been shown to be particularly adept at classifying and identifying soft tissue pathology, as well as performing image segmentation (Fig. 23). Convolutional neural networks (CNN) and deep learning (DL) techniques in particular are proficient at image analysis and will likely be implemented as aids to radiologic interpretation and improvements to workflows in the near future^(123,124).

Fig. 23.

With respect to image analysis, most applications involve a supervised learning approach, whereby a series of input images are used to train the CNN, which entails using a series of filters applied to the data, determining whether there are any dominant features present in the convolved data set, and pooling those features to be compared with a set of truth data. Examples would include detection of tissue boundaries or lesion classification. The operations may be computationally intensive, but are well-suited to the capabilities of the current generation of high-end ultrasound scanners. Based on the author’s experience, these have only been used for relatively simple applications to date, although the current literature has demonstrated promising applications for MSKUS in inflammatory arthritis, segmentation of muscular anatomy, and lesion characterization ^(125–127).

MSKUS has developed rapidly over the last 40 years, paralleling advances in ultrasound technology, software, and hardware in accordance with Moore’s law. The appearance of a large variety of transducer capabilities and rapid image processing, along with the ability to assess vascularity and tissue properties, has already expanded and continues to expand the role of MSKUS. It should also be recognized that these developments have in large part been possible due to a number of individuals who had the insight to see the potential applications of this developing technology to a host of relevant clinical musculoskeletal problems.

Exquisite high-resolution images of both deep and small superficial musculoskeletal anatomy, assessment of vascularity on a capillary level and tissue mechanical properties can be obtained. Ultrasound has also been recognized as the method of choice to perform a large variety of interventional procedures. Time-dependent 3-D imaging along with automatic image analysis depicting abnormal tissue morphology and function, as well as greater sophistication in performing interventions, may be natural outgrowths of these developments.