Page 23 - Spring2019

range of possible effects and the typical clinical requirement for only a specific subset of these at any one instance under- scores the need for careful monitoring during treatment. De- spite the myriad bubble behaviors and resulting bioeffects, very few are noninvasively detectible in vivo. For example, light generation during inertial cavitation (sonolumines- cence) produced under controlled laboratory conditions may be detectible one meter away, but in soft tissue, both the relatively weak light production and its rapid absorption ren- ders in vivo measurement effectively impossible. Magnetic resonance (MR) techniques, known for generating three- dimensional anatomic images, may also be used to measure temperature elevation, with clinically achievable resolution on the order of 1°C, 1 s, and 1 mm (Rieke and Pauly, 2008). However, because these techniques are generally agnostic to the cause(s) of heating, they cannot mechanistically identify bubble contributions to temperature elevation nor can they generally indicate nonthermal actions of bubble activity. On the basis of high-resolution availability of bubble-specific re- sponse cues, active and passive ultrasonic methods appear best suited for noninvasive clinical monitoring.
Active Ultrasound
The enhanced acoustic-scattering strength of a bubble excit- ed near resonance (Ainslie and Leighton, 2011) is exploitable with diagnostic systems that emit and receive ultrasound pulses; bubbles can be identified as regions with an elevated backscatter intensity relative the low-contrast scattering that is characteristic of soft tissues and biological liquids. Intra- venous introduction of microbubbles may therefore sub- stantially improve the diagnostic visibility of blood vessels (where microbubbles are typically confined) by increasing both echo amplitude and bandwidth (Stride and Coussios, 2010). Spatial resolution approaching 10 μm may be clini- cally feasible with “super-resolution” methods employing microbubbles exposed to a high-frame rate sequence of low- amplitude sound exposures (Couture et al., 2018), yielding in vivo microvascular images with a level of detail that was previously unthinkable with diagnostic ultrasound.
Active ultrasound transmissions temporally interleaved with therapeutic ultrasound pulses have been used for bubble detec- tion and tracking (Li et al., 2014). This timing constraint means that cavitation activity occurring during therapeutic ultrasound exposures (often thousands of cycles) would be missed, and with it, information about the therapy process would remain unknown. This limitation is especially important if using solid cavitation nuclei, which are essentially anechoic unless driven with a low-frequency excitation (e.g., therapy beam).
Passive Ultrasound
Passive acoustic mapping (PAM) methods are intended to de- tect, localize, and quantify cavitation activity based on analysis of bubble-scattered sound (for a review of the current art, see Haworth et al., 2017; Gray and Coussios, 2018). Unlike active- imaging modalities, PAM images may be computed at any time during a therapeutic ultrasound exposure and are decou- pled from restrictions on timing, pulse length, or bandwidth imposed by the means of cavitation generation. Because of this timing independence, PAM methods yield fundamentally dif- ferent (and arguably more relevant) information about micro- bubble activity. For example, features in the passively received spectrum such as half-integer harmonics (Figure 1) may be used to distinguish cavitation types from each other and from nonlinearities in the system being monitored and therefore may indicate the local therapeutic effects being applied.
The processes for PAM image formation are illustrated in Figure 2. Suppose a single bubble located at position xs radi- ates sound within a region of interest (ROI), such as a tumor undergoing therapeutic ultrasound exposure. Under ideal conditions, the bubble-emitted field propagates spherically, where it may be detected by an array of receivers, commonly a conventional handheld diagnostic array placed on the skin. PAM methods estimate the location of the bubble based on relative delays between array elements due to the curvature of the received wave fronts. This stands in contrast to con- ventional active methods that localize targets using the time delay between active transmission and echo reception.
After filtering raw array data to remove unwanted signals (such as the fundamental frequency of the active ultrasound that created the cavitation), PAM algorithms essentially run a series of spatial and temporal signal similarity tests consisting of two basic steps. First, the array data are steered to a point (x) in the ROI by adding time shifts to compensate for the path length between x and each array element. If the source actually was located at the steering location (x = xs), then all the array signals would be temporally aligned. Second, the steered data are combined, which in its simplest form involves summation. In more sophisticated forms, the array signal covariance ma- trix is scaled by weight factors optimized to reduce interfer- ence from neighboring bubbles (Coviello et al., 2015).
Regardless of the procedural details, the calculation typically yields an array-average signal power for each steering location in the ROI. This quantity is maximized when the array has been steered to the source location, and the map has its best spatial resolution when interference from other sources has been sup- pressed. The processing illustration in Figure 2 is in the time
Spring 2019 | Acoustics Today | 21