Acoustic simulation for transcranial focused ultrasound using GAN-based synthetic CT

Numerical Investigation of Layered Homogeneous Skull Model for Simulations of Transcranial Focused Ultrasound

BACKGROUND AND OBJECTIVES

The influence of the intracranial pressure field must be discussed with the development of a single-element transducer for low-intensity transcranial focused ultrasound because the skull plays a significant role in blocking and dispersing ultrasound wave propagation. Ultrasound propagation is mainly affected by the structure and acoustic properties of the skull; thus, we aimed to investigate the impact of simplifying the acoustic properties of the skull on the simulation of the transcranial pressure field to present guidance for efficient skull modeling in full-wave simulations.

MATERIALS AND METHODS

We constructed a three-dimensional computational model for ultrasound transmission with the same structure but varying acoustic properties of the skull. The structural information and heterogeneous acoustic properties of the skull were acquired from computed tomography images, and we segmented the skull into three layers (3 L), including spongy and compact bones. We then assigned homogeneous acoustic properties to a single layer (1 L) or 3 L of the skull. In addition, we investigated the influence of different types of transducers and different ultrasound frequencies (1.1 MHz, 0.5 MHz, and 0.25 MHz) on the intracranial pressure field to provide a comparison of the heterogenous and homogeneous models.

RESULTS

We indicated the importance of numerical simulations in estimating the intracranial pressure field of the skull owing to beam distortions. When we simplified the skull model, both the 1 L and 3 L models showed contours of the acoustic focus comparable to those of the heterogeneous model. When we evaluated the peak pressure and volume of the acoustic focus, the 1 L model produced a better estimation of peak pressure with a difference <10%, and the 3 L model is suitable to obtain smaller errors in the volume of the acoustic focus.

CONCLUSIONS

In conclusion, we examined the possibility of simplification of skull models using 1 L and 3 L homogeneous properties in the numerical simulation for focused ultrasound. The results show that the layered homogeneous model can provide characteristics comparable to those of the acoustic focus in heterogeneous models.

Segmentation Guided Crossing Dual Decoding Generative Adversarial Network for Synthesizing Contrast-Enhanced Computed Tomography Images

Although contrast-enhanced computed tomography (CE-CT) images significantly improve the accuracy of diagnosing focal liver lesions (FLLs), the administration of contrast agents imposes a considerable physical burden on patients. The utilization of generative models to synthesize CE-CT images from non-contrasted CT images offers a promising solution. However, existing image synthesis models tend to overlook the importance of critical regions, inevitably reducing their effectiveness in downstream tasks. To overcome this challenge, we propose an innovative CE-CT image synthesis model called Segmentation Guided Crossing Dual Decoding Generative Adversarial Network (SGCDD-GAN). Specifically, the SGCDD-GAN involves a crossing dual decoding generator including an attention decoder and an improved transformation decoder. The attention decoder is designed to highlight some critical regions within the abdominal cavity, while the improved transformation decoder is responsible for synthesizing CE-CT images. These two decoders are interconnected using a crossing technique to enhance each other's capabilities. Furthermore, we employ a multi-task learning strategy to guide the generator to focus more on the lesion area. To evaluate the performance of proposed SGCDD-GAN, we test it on an in-house CE-CT dataset. In both CE-CT image synthesis tasks-namely, synthesizing ART images and synthesizing PV images-the proposed SGCDD-GAN demonstrates superior performance metrics across the entire image and liver region, including SSIM, PSNR, MSE, and PCC scores. Furthermore, CE-CT images synthetized from our SGCDD-GAN achieve remarkable accuracy rates of 82.68%, 94.11%, and 94.11% in a deep learning-based FLLs classification task, along with a pilot assessment conducted by two radiologists.

Robust deep learning estimation of cortical bone porosity from MR T1-weighted images for individualized transcranial focused ultrasound planning

Objective

Transcranial focused ultrasound (tFUS) is an emerging neuromodulation approach that has been demonstrated in animals but is difficult to translate to humans because of acoustic attenuation and scattering in the skull. Optimal dose delivery requires subject-specific skull porosity estimates which has traditionally been done using CT. We propose a deep learning (DL) estimation of skull porosity from T1-weighted MRI images which removes the need for radiation-inducing CT scans.

Approach

We evaluate the impact of different DL approaches, including network architecture, input size and dimensionality, multichannel inputs, data augmentation, and loss functions. We also propose back-propagation in the mask (BIM), a method whereby only voxels inside the skull mask contribute to training. We evaluate the robustness of the best model to input image noise and MRI acquisition parameters and propagate porosity estimation errors in thousands of beam propagation scenarios.

Main results

Our best performing model is a cGAN with a ResNet-9 generator with 3D 64×64×64 inputs trained with L1 and L2 losses. The model achieved a mean absolute error of 6.9% in the test set, compared to 9.5% with the pseudo-CT of Izquierdo et al. (38% improvement) and 9.4% with the generic pixel-to-pixel image translation cGAN pix2pix (36% improvement). Acoustic dose distributions in the thalamus were more accurate with our approach than with the pseudo-CT approach of both Burgos et al. and Izquierdo et al, resulting in near-optimal treatment planning and dose estimation at all frequencies compared to CT (reference).

Significance

Our DL approach porosity estimates with ~7% error, is robust to input image noise and MRI acquisition parameters (sequence, coils, field strength) and yields near-optimal treatment planning and dose estimates for both central (thalamus) and lateral brain targets (amygdala) in the 200-1000 kHz frequency range.

Model-based navigation of transcranial focused ultrasound neuromodulation in humans: Application to targeting the amygdala and thalamus

BACKGROUND

Transcranial focused ultrasound (tFUS) neuromodulation has shown promise in animals but is challenging to translate to humans because of the thicker skull that heavily scatters ultrasound waves.

OBJECTIVE

We develop and disseminate a model-based navigation (MBN) tool for acoustic dose delivery in the presence of skull aberrations that is easy to use by non-specialists.

METHODS

We pre-compute acoustic beams for thousands of virtual transducer locations on the scalp of the subject under study. We use the hybrid angular spectrum solver mSOUND, which runs in ∼4 s per solve per CPU yielding pre-computation times under 1 h for scalp meshes with up to 4000 faces and a parallelization factor of 5. We combine this pre-computed set of beam solutions with optical tracking, thus allowing real-time display of the tFUS beam as the operator freely navigates the transducer around the subject' scalp. We assess the impact of MBN versus line-of-sight targeting (LOST) positioning in simulations of 13 subjects.

RESULTS

Our navigation tool has a display refresh rate of ∼10 Hz. In our simulations, MBN increased the acoustic dose in the thalamus and amygdala by 8-67 % compared to LOST and avoided complete target misses that affected 10-20 % of LOST cases. MBN also yielded a lower variability of the deposited dose across subjects than LOST.

CONCLUSIONS

MBN may yield greater and more consistent (less variable) ultrasound dose deposition than transducer placement with line-of-sight targeting, and thus could become a helpful tool to improve the efficacy of tFUS neuromodulation.

ENHANCING TRANSCRANIAL FOCUSED ULTRASOUND TREATMENT PLANNING WITH SYNTHETIC CT FROM ULTRA-SHORT ECHO TIME (UTE) MRI: A MULTI-TASK DEEP LEARNING APPROACH

Utilizing a multi-task deep learning framework, this study generated synthetic CT (sCT) images from a limited dataset of Ultrashort echo time (UTE) MRI for transcranial focused ultrasound (tFUS) planning. A 3D Transformer U-Net was employed to produce sCT images that closely replicated actual CT scans, demonstrated by an average Dice coefficient of 0.868 for morphological accuracy. The acoustic simulation with sCT images showed mean focus absolute pressure differences of 8.85±7.29 % for the anterior cingulate cortex, 11.81±8.63 % for the precuneus, and 7.27±3.64 % for the supplemental motor cortex, with focus position discrepancies within 0.9±0.5 mm. These results underscore the efficacy of UTE-MRI as a non-radiative, cost-effective alternative for tFUS planning, with significant potential for clinical application.

Deep learning based synthesis of MRI, CT and PET: Review and analysis

Medical image synthesis represents a critical area of research in clinical decision-making, aiming to overcome the challenges associated with acquiring multiple image modalities for an accurate clinical workflow. This approach proves beneficial in estimating an image of a desired modality from a given source modality among the most common medical imaging contrasts, such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET). However, translating between two image modalities presents difficulties due to the complex and non-linear domain mappings. Deep learning-based generative modelling has exhibited superior performance in synthetic image contrast applications compared to conventional image synthesis methods. This survey comprehensively reviews deep learning-based medical imaging translation from 2018 to 2023 on pseudo-CT, synthetic MR, and synthetic PET. We provide an overview of synthetic contrasts in medical imaging and the most frequently employed deep learning networks for medical image synthesis. Additionally, we conduct a detailed analysis of each synthesis method, focusing on their diverse model designs based on input domains and network architectures. We also analyse novel network architectures, ranging from conventional CNNs to the recent Transformer and Diffusion models. This analysis includes comparing loss functions, available datasets and anatomical regions, and image quality assessments and performance in other downstream tasks. Finally, we discuss the challenges and identify solutions within the literature, suggesting possible future directions. We hope that the insights offered in this survey paper will serve as a valuable roadmap for researchers in the field of medical image synthesis.

Real-Time Acoustic Simulation Framework for tFUS: A Feasibility Study Using Navigation System

Transcranial focused ultrasound (tFUS), in which acoustic energy is focused on a small region in the brain through the skull, is a non-invasive therapeutic method with high spatial resolution and depth penetration. Image-guided navigation has been widely utilized to visualize the location of acoustic focus in the cranial cavity. However, this system is often inaccurate because of the significant aberrations caused by the skull. Therefore, acoustic simulations using a numerical solver have been widely adopted to compensate for this inaccuracy. Although the simulation can predict the intracranial acoustic pressure field, real-time application during tFUS treatment is almost impossible due to the high computational cost. In this study, we propose a neural network-based real-time acoustic simulation framework and test its feasibility by implementing a simulation-guided navigation (SGN) system. Real-time acoustic simulation is performed using a 3D conditional generative adversarial network (3D-cGAN) model featuring residual blocks and multiple loss functions. This network was trained by the conventional numerical acoustic simulation program (i.e., k-Wave). The SGN system is then implemented by integrating real-time acoustic simulation with a conventional image-guided navigation system. The proposed system can provide simulation results with a frame rate of 5 Hz (i.e., about 0.2 s), including all processing times. In numerical validation (3D-cGAN vs. k-Wave), the average peak intracranial pressure error was 6.8 ± 5.5%, and the average acoustic focus position error was 5.3 ± 7.7 mm. In experimental validation using a skull phantom (3D-cGAN vs. actual measurement), the average peak intracranial pressure error was 4.5%, and the average acoustic focus position error was 6.6 mm. These results demonstrate that the SGN system can predict the intracranial acoustic field according to transducer placement in real-time.

A transducer positioning method for transcranial focused ultrasound treatment of brain tumors

Purpose

As a non-invasive method for brain diseases, transcranial focused ultrasound (tFUS) offers higher spatial precision and regulation depth. Due to the altered path and intensity of sonication penetrating the skull, the focus and intensity in the skull are difficult to determine, making the use of ultrasound therapy for cancer treatment experimental and not widely available. The deficiency can be effectively addressed by numerical simulation methods, which enable the optimization of sonication modulation parameters and the determination of precise transducer positioning.

Methods

A 3D skull model was established using binarized brain CT images. The selection of the transducer matrix was performed using the radius positioning (RP) method after identifying the intracranial target region. Simulations were performed, encompassing acoustic pressure (AP), acoustic field, and temperature field, in order to provide compelling evidence of the safety of tFUS in sonication-induced thermal effects.

Results

It was found that the angle of sonication path to the coronal plane obtained at all precision and frequency models did not exceed 10° and 15° to the transverse plane. The results of thermal effects illustrated that the peak temperatures of tFUS were 43.73°C, which did not reach the point of tissue degeneration. Once positioned, tFUS effectively delivers a Full Width at Half Maximum (FWHM) stimulation that targets tumors with diameters of up to 3.72 mm in a one-off. The original precision model showed an attenuation of 24.47 ± 6.13 mm in length and 2.40 ± 1.42 mm in width for the FWHM of sonication after penetrating the skull.

Conclusion

The vector angles of the sonication path in each direction were determined based on the transducer positioning results. It has been suggested that when time is limited for precise transducer positioning, fixing the transducer on the horizontal surface of the target region can also yield positive results for stimulation. This framework used a new transducer localization method to offer a reliable basis for further research and offered new methods for the use of tFUS in brain tumor-related research.

Generating Patient-Specific Acoustic Simulations for Transcranial Focused Ultrasound Procedures Based on Optical Tracking Information

Optical tracking is a real-time transducer positioning method for transcranial focused ultrasound (tFUS) procedures, but the predicted focus from optical tracking typically does not incorporate subject-specific skull information. Acoustic simulations can estimate the pressure field when propagating through the cranium but rely on accurately replicating the positioning of the transducer and skull in a simulated space. Here, we develop and characterize the accuracy of a workflow that creates simulation grids based on optical tracking information in a neuronavigated phantom with and without transmission through an ex vivo skull cap. The software pipeline could replicate the geometry of the tFUS procedure within the limits of the optical tracking system (transcranial target registration error (TRE): 3.9 ± 0.7 mm). The simulated focus and the free-field focus predicted by optical tracking had low Euclidean distance errors of 0.5±0.1 and 1.2±0.4 mm for phantom and skull cap, respectively, and some skull-specific effects were captured by the simulation. However, the TRE of simulation informed by optical tracking was 4.6±0.2, which is as large or greater than the focal spot size used by many tFUS systems. By updating the position of the transducer using the original TRE offset, we reduced the simulated TRE to 1.1 ± 0.4 mm. Our study describes a software pipeline for treatment planning, evaluates its accuracy, and demonstrates an approach using MR-acoustic radiation force imaging as a method to improve dosimetry. Overall, our software pipeline helps estimate acoustic exposure, and our study highlights the need for image feedback to increase the accuracy of tFUS dosimetry.

Evaluation of synthetically generated computed tomography for use in transcranial focused ultrasound procedures

Purpose

Transcranial focused ultrasound (tFUS) is a therapeutic ultrasound method that focuses sound through the skull to a small region noninvasively and often under magnetic resonance imaging (MRI) guidance. CT imaging is used to estimate the acoustic properties that vary between individual skulls to enable effective focusing during tFUS procedures, exposing patients to potentially harmful radiation. A method to estimate acoustic parameters in the skull without the need for CT is desirable.

Approach

We synthesized CT images from routinely acquired T1-weighted MRI using a 3D patch-based conditional generative adversarial network and evaluated the performance of synthesized CT (sCT) images for treatment planning with tFUS. We compared the performance of sCT with real CT (rCT) images for tFUS planning using Kranion and simulations using the acoustic toolbox, k-Wave. Simulations were performed for 3 tFUS scenarios: (1) no aberration correction, (2) correction with phases calculated from Kranion, and (3) phase shifts calculated from time reversal.

Results

From Kranion, the skull density ratio, skull thickness, and number of active elements between rCT and sCT had Pearson's correlation coefficients of 0.94, 0.92, and 0.98, respectively. Among 20 targets, differences in simulated peak pressure between rCT and sCT were largest without phase correction (12.4 % ± 8.1 % ) and smallest with Kranion phases (7.3 % ± 6.0 % ). The distance between peak focal locations between rCT and sCT was < 1.3    mm for all simulation cases.

Conclusions

Real and synthetically generated skulls had comparable image similarity, skull measurements, and acoustic simulation metrics. Our work demonstrated similar results for 10 testing cases comparing MR-sCTs and rCTs for tFUS planning. Source code and a docker image with the trained model are available at https://github.com/han-liu/SynCT_TcMRgFUS.

BabelBrain: An Open-Source Application for Prospective Modeling of Transcranial Focused Ultrasound for Neuromodulation Applications

BabelBrain is an open-source standalone graphic-user-interface application designed for studies of neuromodulation using transcranial-focused ultrasound (FUS). It calculates the transmitted acoustic field in the brain tissue, taking into account the distortion effects caused by the skull barrier. The simulation is prepared using scans from magnetic resonance imaging (MRI) and, if available, computed tomography (CT) and zero-echo time MRI scans. It also calculates the thermal effects based on a given ultrasound regime, such as the total duration of exposure, the duty cycle, and acoustic intensity. The tool is designed to work in tandem with neuronavigation and visualization software, such as 3-DSlicer. It uses image processing to prepare domains for ultrasound simulation and uses the BabelViscoFDTD library for transcranial modeling calculations. BabelBrain supports multiple GPU backends, including Metal, OpenCL, and CUDA, and works on all major operating systems including Linux, macOS, and Windows. This tool is particularly optimized for Apple ARM64 systems, which are common in brain imaging research. The article presents the modeling pipeline used in BabelBrain and a numerical study where different methods of acoustic properties mapping were tested to select the best method that can reproduce the transcranial pressure transmission efficiency reported in the literature.

Classical and Learned MR to Pseudo-CT Mappings for Accurate Transcranial Ultrasound Simulation

Model-based treatment planning for transcranial ultrasound therapy typically involves mapping the acoustic properties of the skull from an X-ray computed tomography (CT) image of the head. Here, three methods for generating pseudo-CT (pCT) images from magnetic resonance (MR) images were compared as an alternative to CT. A convolutional neural network (U-Net) was trained on paired MR-CT images to generate pCT T images from either T1-weighted or zero-echo time (ZTE) MR images (denoted tCT and zCT, respectively). A direct mapping from ZTE to pCT was also implemented (denoted cCT). When comparing the pCT and ground-truth CT images for the test set, the mean absolute error was 133, 83, and 145 Hounsfield units (HU) across the whole head, and 398, 222, and 336 HU within the skull for the tCT, zCT, and cCT images, respectively. Ultrasound simulations were also performed using the generated pCT images and compared to simulations based on CT. An annular array transducer was used targeting the visual or motor cortex. The mean differences in the simulated focal pressure, focal position, and focal volume were 9.9%, 1.5 mm, and 15.1% for simulations based on the tCT images; 5.7%, 0.6 mm, and 5.7% for the zCT; and 6.7%, 0.9 mm, and 12.1% for the cCT. The improved results for images mapped from ZTE highlight the advantage of using imaging sequences, which improves the contrast of the skull bone. Overall, these results demonstrate that acoustic simulations based on MR images can give comparable accuracy to those based on CT.

Comparison between MR and CT imaging used to correct for skull-induced phase aberrations during transcranial focused ultrasound

Transcranial focused ultrasound with the InSightec Exablate system uses thermal ablation for the treatment of movement and mood disorders and blood brain barrier disruption for tumor therapy. The system uses computed tomography (CT) images to calculate phase corrections that account for aberrations caused by the human skull. This work investigates whether magnetic resonance (MR) images can be used as an alternative to CT images to calculate phase corrections. Phase corrections were calculated using the gold standard hydrophone method and the standard of care InSightec ray tracing method. MR binary image mask, MR-simulated-CT (MRsimCT), and CT images of three ex vivo human skulls were supplied as inputs to the InSightec ray tracing method. The degassed ex vivo human skulls were sonicated with a 670 kHz hemispherical phased array transducer (InSightec Exablate 4000). 3D raster scans of the beam profiles were acquired using a hydrophone mounted on a 3-axis positioner system. Focal spots were evaluated using six metrics: pressure at the target, peak pressure, intensity at the target, peak intensity, positioning error, and focal spot volume. Targets at the geometric focus and 5 mm lateral to the geometric focus were investigated. There was no statistical difference between any of the metrics at either target using either MRsimCT or CT for phase aberration correction. As opposed to the MRsimCT, the use of CT images for aberration correction requires registration to the treatment day MR images; CT misregistration within a range of ± 2 degrees of rotation error along three dimensions was shown to reduce focal spot intensity by up to 9.4%. MRsimCT images used for phase aberration correction for the skull produce similar results as CT-based correction, while avoiding both CT to MR registration errors and unnecessary patient exposure to ionizing radiation.

Differential evolution method to find optimal location of a single-element transducer for transcranial focused ultrasound therapy

BACKGROUND AND OBJECTIVE

Focused ultrasound (FUS) has been receiving growing attention as a noninvasive brain stimulation tool because of its superior spatial specificity and depth penetrability. However, the large mismatch of acoustic properties between the skull and water can disrupt and shift the acoustic focus in the brain. In this paper, we present a numerical method to find the optimal location of a single-element FUS transducer, which creates focus on the target region.

METHODS

The score function, representing the superposition of acoustic waves according to the relative phase difference and transmissibility, was defined based on time-reversal invariance of acoustic waves and depending on the spatial location of the transducer. The optimal location of the transducer was then determined using a differential evolution algorithm. To assess the proposed method, we conducted a forward simulation and compared the resulting focal location to the desired target point. We also performed experimental validation by measuring the acoustic pressure field through an ex vivo human skull in a water tank.

RESULTS

The numerical results indicated that the score function had a positive proportional relationship with the acoustic pressure at the target. Moreover, for the optimized transducer location, both the numerical and experimental results showed that the normalized acoustic pressure at the target was higher than 0.9.

CONCLUSIONS

In this study, we developed an optimization method to place a single-element transducer that effectively transmits acoustic energy to the targeted region in the brain. Our numerical and experimental results demonstrate that the proposed method can provide an optimal transducer location for safe and efficient FUS treatment.