Experimental setup

All experiments were performed in tanks of de-ionized water filtered to 2µm and degassed to 20% oxygen saturation as measured by a ThermoFisher Orion Star A323 oxygen meter (Waltham, MA). The water temperatures were measured to be between 20 and 21°C. Samples of ex-vivo bovine tissue (liver, kidney, brain; < 2hrs from sacrifice) were procured from a local abattoir. Tissues were cut into 3 cm cubic volumes and degassed under 30inHg vacuum for at least 3 hours to extract any gas that dissolved into the tissues during transit [ ]. Samples were then embedded in 1.5% agarose hydrogel tissue holders [ ]. During treatments (described below), the tissue samples were mounted at the geometric focus of the histotripsy transducer in the water tank and treated using a range of different acoustic doses to generate different levels of induced damage at the treatments’ endpoints ( Figure 1 ). All tissues were treated within four days of acquisition.

Experimental Setup Diagram.

Histotripsy transducers

Two histotripsy transducers, designed and fabricated in-house, were used in this study. The transcranial array is a 360 element, 30 cm aperture diameter hemispherical histotripsy array with a center frequency of 700kHz ( Figure 2 A). The focal zone was 1.2 × 1.2 × 2.3mm ³ in the two transverse and axial axes, respectively. Its geometry was designed to cater to transcranial treatments and was used to treat the bovine brain samples [ ]. The abdominal histotripsy array is a spherically curved 750kHz phased array with a 14.2cm focal radius ( Figure 2 B). The 260 elements of the array are arranged in concentric rings around a 6 cm diameter cutout at the center to allow for the placement of an imaging transducer. The focal zone of the array was measured to be 1.6 × 1.1 × 4.5 mm ³ . The array has an ellipsoidal aperture suitable for coupling to the abdomen, so it was used to treat the bovine liver and kidney samples [ ].

(A) 700kHz, 360-element transcranial histotripsy array used in the bovine brain experiments (15cm diameter; 17 × 17 mm ² elements); (B) 750kHz, 260-element abdominal histotripsy array used in the bovine liver and kidney experiments (14.2cm radius of curvature; 11.5 × 11.5 mm ² elements).

Each transducer element of both arrays was configured to operate in a transmit-receive mode to allow simultaneous generation of the acoustic pulses necessary to generate cavitation and the acquisition of the resulting ACE-signals emitted therefrom. The individual transducer elements were characterized by measuring their output waveform using a fiber-optic hydrophone (HP; Onda HPO-690, Sunnyvale CA) positioned at the focal radius of each array (transcranial – 15cm, abdominal 14.2cm). Acoustic pulses emitted by the elements of each array measured approximately 1.5 acoustic cycles, with corresponding durations of 2.1us (transcranial array) and 2us (abdominal array). The focal zones of the arrays were calculated by firing all of the elements and measuring one-dimensional (1D) transmit beam profiles along three axes using a needle hydrophone (Onda HNR-500, Sunnyvale CA), at low driving pressures (≤1MPa) [ , ]. The sizes of the arrays’ focal zones were defined as the full-width half-maximum (FWHM) of the measured 1D beam profiles. On receive, each transducer element of the arrays was connected to a 12-bit analog-to-digital converter (ADC) with a maximum sampling frequency of up to 50 MHz. Control of the arrays was accomplished using custom programmed system-on-a-chip (SoC) devices, which could store received ACE-signals from every element of the array after every delivered acoustic pulse.

Histotripsy treatment parameters

An 8 × 8 × 8 mm ³ cubic ablation region was prescribed to each tissue sample. This volume was comprised of many individual histotripsy steering foci that were arranged in a hexagonal-close-packing (HCP) lattice with 0.5 mm lattice point spacing. During treatment, the focal pressure amplitudes of the array were set to 30 and 40 MPa for the abdominal and transcranial array treatments, respectively. The different absolute pressure amplitudes between arrays were chosen to compensate for the differences in the focal zone shapes and set such that the histotripsy bubble clouds generated by each array would be roughly equivalent in size and have diameters of ∼1.5mm. These pressures were determined from preliminary studies in which brain, liver, and kidney samples were exposed to up to 50 pulses of histotripsy at a single focal location at varying driving pressures, processed for H&E histology, and analyzed to compare the relationship how driving pressure determines focal lesion size for each array. Further, the electric drive power of the arrays was modulated during treatment to compensate for focal steering pressure losses such that these pressures were uniformly maintained throughout the entire steering volume. During treatments, histotripsy pulses were applied to the tissue samples at a pulse repetition frequency (PRF) of 500 Hz. Treatments were delivered such that each subsequent histotripsy pulse would be applied to a different focal location within the steering grid until one pulse had been delivered to all steering locations within the grid. This was then repeated a specified number of times to meet desired acoustic dose, hereafter referred to as pulses-per-location (ppl), of the treatment. In each tissue type, exposures of 3 / 5 / 10 / 20 / 30 ppl were delivered; 4 brain samples and 6 liver and kidney samples were treated per exposure level (80 tissue samples total). To ensure that the cut plane of the tissue during histological sectioning bisected the generated ablation zone, an additional ‘registration’ lesion was generated by applying 500 histotripsy pulses along the line between [0,0,-15] and [0,0,-30] (i.e. the boundary between gel & tissue). These ‘registration’ lesions were placed ≥10mm from the cubic treatment volume, measured approximately 2mm in diameter, were aligned with the center of the cubic treatment volume along its transverse (X, Y) axes, and were visible to the naked eye to assist with histological sectioning.

Acoustic cavitation emission (ACE) signals

ACE-signals were captured using the receive capable elements (Rx) of the histotripsy arrays following every other pulse of treatment [ ] ( Figure 3 ). To further reduce the data volume, signals were acquired at a sampling frequency of 5 MHz. Receive data was sub-sampled in this way in order to manage the volume of recorded data while still maintaining the ability to sample bubble activity across the treatment volume. Phase-delays were applied to compensate for focal steering, then the timings and amplitudes of each shockwave within the acquired signals were calculated as follows and recorded. Given time t in 1:T, elements e in 1:E, and signal S(t,e):

Example ACE Waveform. The sub-figures on the bottom row are zoomed windows of the nucleation and collapse shockwaves as measured by one transducer element of the histotripsy array. Although the analog-to-digital converter (ADC) can sample up to 50MHz, data was captured at 5MHz during ablations to decrease the volume of the resulting data.

First, the coherence of the signal from a single histotripsy event measured across the elements of the acoustic array was calculated as described in Equation 1 following the metric introduced by Mallart and Fink [ ] ( Figure 4 a). Because ACE-signals have high phase coherence compared to reflections from tissue, time windows for the nucleation and collapse signals could be determined via peak detection on the coherence [ , ] ( Figure 4 b). These 20µs windows were applied to the raw data, and singular value decompositions were calculated on the time-windowed data matrices containing nucleation or collapse signals from all elements [ ] ( Figure 4 c,d,e,f). All but the first eigenvalues were discarded to remove background noise and isolate the shockwaves [ , , ]. Peaks in the SVD-filtered signals were detected to identify the shockwave timings and amplitudes for each focal-location.

Shockwave measurement from an ACE signal. (A) Apply “rough” windows around the nucleation and collapse shockwaves in the time x numElements image, and calculate the coherence across both windows. (B) Peak-detect the nucleation and collapse shockwaves, and calculate “tight” (20 µs) windows around each one for an SVD filter. (C) Before and (D) after the SVD filter is applied to the “tight” nucleation window. (E) Before and (F) after the SVD filter is applied to the “tight” collapse window.

Bubble lifespan is calculated as the mean difference between the nucleation and collapse shockwave arrival times measured across all focal locations. Then, bubble lifespan-slope was calculated as a series of linear-fits on the bubble lifespan as a function of ppl using a moving window (size 4 ppl). Finally, percent-reduction (i.e. normalized reduction in lifespan-slope) is calculated with Equation 2 . Given pulse p in 1:P, and LifespanSlope(p):

Histology

Following treatment, each tissue block (embedded in agarose gel) was marked to indicate the positive X/Y/Z axes (relative to the histotripsy array elements’ coordinates) and was stored in 10% formalin for 72 hours to fix before storage in 70% ethanol. To prepare for histology processing, each tissue block was bisected using the ‘registration’ lesion to localize the damaged volume. After bisecting the samples, two notches were cut into the tissue outside of the treatment zone to preserve indicators of the orientation of the sample during treatment following histological processing as needed for co-registration with localized, ACE-signal assessments later. One diagonal notch was placed in the corner of the cube to designate the +X and +Z directions, and another triangular notch was placed midway along the face of the cube to designate the -Z direction ( Figure 4 ). These notches allowed unique identification of the orientation of the sample within the histotripsy array during treatment. Finally, tissue samples were loaded in histology cassettes, embedded in paraffin, and stained using hematoxylin and eosin (H&E) to visualize their structures and the mechanical disruption generated within.

Following staining, ground-truth damage results were assessed by a board-certified surgical pathologist (Jiaqi Shi). Damaged regions, similar to the region outlined in black in Figure 5 , were manually segmented on each slide for analysis. Then the pathologist, blinded to the treatment applied to the tissue, graded the damage in each region based on the overall percentage of tissue necrosis and the remaining areas of residual structures (e.g., epithelial cells, neurons, blood vessels, stroma, glia, etc.) using the following grading scales [ ]:

• 
  

  Necrosis score: 0 (0%), 1 (1-25%), 2 (26-50%), 3 (51-75%), 4 (76-95%), 5 (>95%)

Sample H&E of treated bovine brain. Histotripsy was applied in the -Z -to- +Z direction. Blue dotted lines indicate cuts to register H&E and histotripsy array coordinate systems. The black dashed line indicates the prescribed treatment area. The red dotted line indicates ablated tissue.

Residual Structure: 0 (none present), 1 (patchy/small areas), 2 (patchy/large areas), 3 (diffuse areas), 4 (intact). Several samples were omitted from the final analysis due to the histology slices not intersecting the center of the ablation lesion. This resulted when larger tissue samples surpassed the ‘registration’ lesion, and clean slices of the center of the ablation zone could not be procured following histotripsy exposure. The following samples were removed from their respective tissue-type/acoustic-dose groups: 2 brain-20ppl, 1 liver-3ppl, 1 liver-5ppl, 1 kidney-5ppl, 1 kidney-10ppl.

ACE Correlation with Pathology

As the pathologist evaluations are discrete measurements, the Spearman correlation was used to evaluate the efficacy of using ACE-signals to predict tissue disruption. The lifespan, lifespan-slope, and percent-reduction from the last pulse of each ablation were correlated against the corresponding necrosis scores and residual structure scores.

In addition, ideal thresholds for binary evaluations of no/partial damage vs. near-complete/complete damage were determined for each of the three ACE-features (lifespan, lifespan-slope, percent reduction). For the ground-truth data, a necrosis grade 4 and 5 was used as the threshold for near-complete/complete damage. Although grade 4 retained some viable tissue, most of the residual tissues are detached, minute islands of cells suspended in necrotic debris with no blood supplies and would likely not survive. For the experimental data, final lifespan/lifespan-slope/percent-reduction was used to predict the presence of near-complete/complete damage. ROC curves were calculated by sweeping threshold values for each of the ACE-features, and calculating the corresponding sensitivities ( Eq. 3 ) and specificities ( Eq. 4 ) to the pathological results. Area-under-the-curve (AUC) was calculated using integration with the trapezoidal method. The ideal threshold for each the ACE-feature/tissue-type pair was identified as the value with the greatest F1-score ( Eq. 5 ). For this value, the “peak” sensitivity, specificity, accuracy ( Eq. 6 ), positive-predictive value (PPV) ( Eq. 7 ), and negative-predictive value (NPV) ( Eq. 8 ) are calculated. Without a defined clinical standard for damage monitoring of mechanical ablation, validation of this technique was performed by comparing its efficacy against those reported for other methods [ , , , ].

For true-positive counts TP , true-negative counts TN , false-positive counts FP , and false-negative counts FP :

Related posts:

Quantitative Analysis of Contrast-Enhanced Ultrasound Images of Brain-Dead Donor Livers: Prediction of Early Allograft Dysfunction Ultrasound Shear Wave Elastography and Shear Wave Dispersion: Correlation with Histopathological Changes in Autoimmune Hepatitis Patients Diagnosis of Salivary Gland Tumors Using Ultrasound Radiomics Empirical Model of Focused Ultrasound-Mediated Treatment for Chemotherapy Delivery to Brain Tumors High-Frequency Ultrasound Characterization of Periodontal Soft Tissues Pre- and Post-Bacterial Inoculation SDRD-Net: A Symmetric Dual-branch Residual Dense Network for OCT and US Image Fusion

Stay updated, free articles. Join our Telegram channel

Join