A novel method for acoustic modeling of cranial bone based on the porosity index Authors: Dechen Kong, Tao Zhang, Guozheng Hou, Gaojie Liu, Xu Qi, Haoyang Xing Published: 2025 ABSTRACT Transcranial ultrasound stimulation (TUS) has emerged as a clinically validated neuromodulation technique. Particularly, phased array ultrasound can be applied in TUS to focus on the cortex or deep brain non-invasively, such as the ventral intermediate thalamic nucleus (VIM) and Precuneus (PCu) region for the treatment of essential tremor (ET) and Alzheimer Disease (AD). Current TUS treatment planning relies on computed tomography (CT)-derived skull porosity measurements, which involve patient radiation exposure and potential registration errors. This study proposes a Porosity Index (PI)-based method, derived from Ultrashort Echo Time (UTE) Magnetic Resonance (MR) images, for establishing skull acoustic models as a viable alternative, aiming to eliminate these limitations. Acoustic simulations using the K-Wave open-source platform were performed to validate the PI method’s accuracy in predicting skull porosity and simulating focal distributions compared to CT. Focal spot characteristics were quantified using five metrics: peak intensity, target intensity, focal positioning error, Dice similarity coefficient, and Pearson correlation coefficient between CT- and PI-based simulation results. Statistical differences between these metrics were assessed using Tukey’s multiple comparisons test. Quantitative comparisons against the gold-standard CT approach demonstrated comparable performance in peak focal intensity (deviation < 5%) and spatial pressure distribution patterns (Dice coefficient > 0.82). No significant differences (p > 0.05) were observed for any of the evaluated metrics. Our findings demonstrate that both the sound pressure distribution and prediction of the porosity are comparable with those from the reference CT. Using the PI to replace the traditional CT porosity has high feasibility and can achieve the purpose of reducing unnecessary radiation exposure and registration error for patients. Introduction Transcranial ultrasound stimulation (TUS) techniques1 have been clinically applied for neuromodulation. This approach leverages the non-invasive capability of transcranial focused ultrasound to safely access deep brain regions pathologically linked to movement disorders, such as essential tremor and Parkinsonian tremor2–4. To overcome phase aberration challenges in clinical settings, numerical simulations based on computed tomography (CT) scans are widely adopted5–8. These simulations utilize personalized three-dimensional (3-D) skull models derived from CT images to calculate tissue-induced phase shifts and time delays, thereby optimizing transducer beam focusing9. Crucially, the accuracy of acoustic properties mapped from CT Hounsfield Units calculations derived from CT images significantly impacts the simulated pressure field10. However, the method’s effectiveness is limited primarily by uncertainty in the skull’s acoustic properties, along with inaccuracies in acoustic simulations and errors from CT-MR image registration11–13. Even minimal changes in skull thickness (as little as 0.1 mm) can lead to errors exceeding 5% in the intracranial pressure14.These limitations can potentially degrade beamforming performance. Furthermore, the use of CT exposes patients to ionizing radiation, which is undesirable for neuromodulation applications. The human skeleton can be conceptualized as a composite material, primarily composed of minerals (approximately 45% by volume), organic matrix (approximately 30% by volume), and water (approximately 20% by volume). The pore network within cortical bone encompasses structures such as Haversian canals and Volkmann’s canals (approximately \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$40 \sim 100$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}m in diameter), osteocyte lacunae (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10 \sim 30$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}m), and canaliculi (approximately 0.1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}m). Free water is commonly present within these pores. Ultrashort echo time (UTE) magnetic resonance (MR) imaging enhances skull signal detection by capturing the proton signal from two distinct water compartments: pore water residing within the bone pores and bound water that forms hydrogen bonds with the organic matrix. Prior research has established a difference in T2* relaxation times between these compartments: bound water exhibits an ultrashort T2* (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T2* \sim 300$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}s), while pore water has a longer T2* (T2* > 1 ms)15. Building on this distinction, Rajapakse et al. (2015)16 introduced the porosity index (PI) concept for assessing cortical bone strength. The PI leverages the signal ratio from UTE sequences acquired at two different echo times (TEs): the signal at a longer TE predominantly reflects pore water, whereas the signal at a shorter TE represents the combined signal from both pore water and bound water. Consequently, the ratio of the long-TE signal to the short-TE signal serves as a measure of bone porosity, thereby providing an indirect assessment of cortical bone strength. This ratio is defined as the PI. UTE-MR imaging has emerged as a valuable technique for visualizing short-T2 tissues, including bone. Consequently, researchers have leveraged UTE-MR images to synthesize computed tomography (CT) scans for calculating acoustic properties and establishing skull acoustic models. Representative methods include the logarithmic transformation of UTE intensity17, MRsimCT15, and deep learning-based synthetic CT approaches18,19. Validation studies support the feasibility of these methods. Guo et al. (2019) demonstrated that CT images synthesized from the logarithmic intensity of UTE data yielded thermal profiles and focal locations consistent with the actual clinical treatment plan17. Furthermore, Leung et al. showed that MRsimCT has the potential to replace conventional CT scans in certain contexts20. However, two key limitations persist in current clinical practice. First, treatment planning that relies directly on CT-derived porosity entails ionizing radiation exposure. Second, when CT data must be registered with MR images for targeting, registration errors may occur. While several UTE-MR based approaches have demonstrated the feasibility of replacing CT through the synthesis of CT-equivalent images, we propose a novel UTE-based method that estimates acoustic porosity directly from the native MR signals, without any intermediate CT synthesis. This approach aims to further advance that MR can fully replace CT for transcranial ultrasound treatment planning, potentially offering a more streamlined workflow. To overcome the limitations mentioned above, we used the bone porosity index (PI) method to measure bone porosity in the current study, using UTE images to establish the skull acoustic model (see Fig. 1). This technique received validation through numerical simulations conducted with K-Wave21, in which the simulated pressure fields and focusing metrics (e.g., focal pressure, positioning error) derived from PI-based models were compared against those from the gold-standard CT-based models. Our primary objective is to determine whether this PI-based approach can achieve acoustic simulation results comparable to the current gold-standard CT-based method. If successful, this method could serve as a viable clinical alternative, eliminating the need for CT and thus avoiding patient radiation exposure and potential registration errors.Fig. 1These images from one representative subject (23 years of age, male). From top to bottom: UTE TE 1 and TE 2 images, reference CT, reference CT of the skull, reference CT of the skull porosity map, and the PI map. These images from one representative subject (23 years of age, male). From top to bottom: UTE TE 1 and TE 2 images, reference CT, reference CT of the skull, reference CT of the skull porosity map, and the PI map. Results Prediction performances The prediction results of the estimated bone porosity using the PI methods are shown in Fig. 2. Following image registration and spatial resolution matching through downsampling, the co-registered CT-MR fusion demonstrates notable consistency in osseous tissue contrast through multiplanar visualization, with corresponding bone porosity profiles across the anatomical cross-section quantitatively validating correspondence. The valley of the curve corresponds to the signal of the skull’s cortical bone position. Notably, we applied Formula (1) to calculate bone porosity from CT images for direct comparison with PI values (a, c). Figure 2b,d display the marginal histograms of the PI method measurement porosity compared with CT. The denser the dots are, the darker the color is. The black line represents the linear regression between voxel-based PI intensity and CT bone porosity based intensity in this case (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CT_{bone porosity} = 0.66 PI + 47.39, R^2=0.40, P<0.05$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CT_{bone porosity} = 0.62 PI + 48.95, R^2=0.37, P<0.05$$\end{document}). The results demonstrate that the PI method provides a voxel-wise estimate of bone porosity that is significantly correlated with the CT-based method (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2 = 0.40$$\end{document}, P<0.05), although the absolute values and distribution shapes differ. The detailed data is shown in Table 4. While the absolute porosity values derived from PI and CT show considerable differences, their spatial patterns exhibit a degree of similarity that may be sufficient for acoustic modeling purposes, as evidenced by the subsequent focusing results.Fig. 2The CT-likeness in the difference images can be quantified by direct comparison of bone porosity. The (a, c) denotes MR(PI) and CT images, skull porosity cross-section plot in the same position. The MR (PI) map intensity represents the Porosity Index (0–100%), while the CT image intensity is in Hounsfield Units (HU). The vertical axis represents porosity and describes the distribution of skull pores, with 100 indicating maximum and 0 indicating no pores. The (b, d) display the marginal histograms of the PI method measurement porosity compared with CT, and the black line represents the linear regression between voxel-based PI intensity and CT bone porosity. The CT-likeness in the difference images can be quantified by direct comparison of bone porosity. The (a, c) denotes MR(PI) and CT images, skull porosity cross-section plot in the same position. The MR (PI) map intensity represents the Porosity Index (0–100%), while the CT image intensity is in Hounsfield Units (HU). The vertical axis represents porosity and describes the distribution of skull pores, with 100 indicating maximum and 0 indicating no pores. The (b, d) display the marginal histograms of the PI method measurement porosity compared with CT, and the black line represents the linear regression between voxel-based PI intensity and CT bone porosity. Focusing performance Although a significant correlation was observed between CT and PI-based skull porosity measures, our primary objective was to evaluate whether PI-based measurements perform comparably to CT-based data in ultrasound treatment simulations.Fig. 3Acoustic simulation results of intensity distribution for two scenarios with focus ultrasound targets placed in the Anterior Precuneus (b-PCu), and ventral intermediate thalamic nucleus (a-VIM). The field of view for all acoustic field plots is 225 mm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 225 mm. The figure a and b also displays 1D acoustic intensity distribution near the focus along lateral (parallel to the transducer plane) and the b longitudinal directions. The colorbar denotes norm sound intensity (0–1). Acoustic simulation results of intensity distribution for two scenarios with focus ultrasound targets placed in the Anterior Precuneus (b-PCu), and ventral intermediate thalamic nucleus (a-VIM). The field of view for all acoustic field plots is 225 mm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 225 mm. The figure a and b also displays 1D acoustic intensity distribution near the focus along lateral (parallel to the transducer plane) and the b longitudinal directions. The colorbar denotes norm sound intensity (0–1). The acoustic field simulation results for all five patient cases are shown in Fig. 3. The focal intensity is an estimated number determined by the K-WAVE simulation. These results encompass simulations of two scenarios that target the PCu and VIM, utilizing models based on actual CT and UTE images. The CT acoustic model simulation results indicate that acoustic energy is focused near the target, and the pressure distribution displays spherical symmetry with sidelobes that are below 40% of the maximum value. All simulated pressure fields were normalized to the spatial peak pressure obtained in the simulation using the gold-standard CT-derived acoustic model. When compared to the CT benchmark, the PI method demonstrated a normalized focal pressure difference of 9.84 ± 8.52% (mean ± standard deviation) for the PCu target and 25.07 ± 6.80% for the VIM target. This indicates that the PI method can accurately recover focal pressure for the PCu target (mean deviation < 10%), while for the deeper VIM target, the larger pressure difference remains within an acceptable range and exhibits lower variability compared to other alternatives like MRsimCT. The focal positioning error was 0.23 ± 0.21 cm for PCu and 0.08 ± 0.1 cm for VIM. This sub-centimeter accuracy, particularly the high precision for the VIM target, demonstrates the method’s effectiveness in steering the acoustic energy to the intended location, which is sufficient for clinical neuromodulation applications. The spatial similarity of the pressure field to the CT result, measured by the Dice coefficient, was 0.814 on average for PCu and 0.924 for VIM. These high similarity coefficients (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$>0.8$$\end{document}) indicate that the focal shape and spatial extent generated by the PI-based models are in excellent agreement with the CT benchmark, which is crucial for ensuring reliable coverage of the target region during stimulation. In summary, while variations in absolute focal pressure exist, the PI method achieves comparable performance to CT in terms of targeting accuracy and focal volume fidelity, supporting its viability as an alternative acoustic modeling approach. The detailed data is shown in Table 1.Table 1Differences in focus pressure (%) and position (cm) between models (PI, MRsimCT, MRBinary) and the reference CT model.Target locationPIMRsimCTMRBinaryPressure difference (%)PCu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$9.84 \pm 8.52$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$15.07 \pm 8.92$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$38.25 \pm 2.16$$\end{document}VIM\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$25.07 \pm 6.80$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$30.21 \pm 19.43$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$68.22 \pm 11.49$$\end{document}Focus difference (cm)PCu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.23 \pm 0.21$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.84 \pm 0.65$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.63 \pm 1.27$$\end{document}VIM\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.08 \pm 0.1$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.12 \pm 0.1$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.18 \pm 0.04$$\end{document} Table 1Differences in focus pressure (%) and position (cm) between models (PI, MRsimCT, MRBinary) and the reference CT model.Target locationPIMRsimCTMRBinaryPressure difference (%)PCu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$9.84 \pm 8.52$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$15.07 \pm 8.92$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$38.25 \pm 2.16$$\end{document}VIM\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$25.07 \pm 6.80$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$30.21 \pm 19.43$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$68.22 \pm 11.49$$\end{document}Focus difference (cm)PCu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.23 \pm 0.21$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.84 \pm 0.65$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.63 \pm 1.27$$\end{document}VIM\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.08 \pm 0.1$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.12 \pm 0.1$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.18 \pm 0.04$$\end{document} Differences in focus pressure (%) and position (cm) between models (PI, MRsimCT, MRBinary) and the reference CT model. Fig. 4Focusing performance metrics for established acoustic modeling methods: (a) Target intensity (striped bars) vs. peak intensity (solid bars) in ventral intermediate nucleus (VIM); (b) Acoustic intensity distribution at precuneus (PCu) target; (c) Dice similarity coefficient of focal regions; (d) Positioning error with standard deviation bars.. Focusing performance metrics for established acoustic modeling methods: (a) Target intensity (striped bars) vs. peak intensity (solid bars) in ventral intermediate nucleus (VIM); (b) Acoustic intensity distribution at precuneus (PCu) target; (c) Dice similarity coefficient of focal regions; (d) Positioning error with standard deviation bars.. The focal spots were evaluated using focus intensity and position error, which are shown in Fig. 4a–d. In (a), the MRsimCT, PI, and CT demonstrated similar performance in phase correction regarding normalized target intensity and normalized peak intensity. Moreover, we tested the Pearson and Dice similarity coefficient (d) of the PI and CT methods and summarized them in Table 2. The results indicate that the PI method can achieve precise focusing comparable to CT results, and the sound field distribution is similar. A one-way repeated measures ANOVA with Tukey’s multiple comparisons test revealed no significant differences for any metric (Table 3).Table 2Similarity performance for PI, MRsimCT and MRBinary set.Target locationIndex#1#2#3#4#5AveragePerasonPcuPI0.980.950.920.970.980.96MRsimCT0.980.920.930.850.980.932MRBinary0.930.750.780.730.880.814VIM (128)PI0.990.970.970.960.980.974MRsimCT0.990.970.940.810.970.936MRBinary0.970.930.90.780.950.906DicePcuPI0.870.780.730.820.870.814MRsimCT0.880.730.740.650.840.768MRBinary0.750.540.560.570.690.622VIM (128)PI0.950.930.920.90.920.924MRsimCT0.950.920.860.820.930.896MRBinary0.920.860.860.770.880.858Table 3Tukey’s multiple comparisons test.IntensityPositionDicePearsonPCuPI-MRsimCT0.640.410.210.3PI-MRBinary0.560.930.510.43MRsimCT-MRBinary0.980.60.780.96VIMPI-MRsimCT0.850.620.160.56PI-MRBinary0.810.940.060.11MRsimCT-MRBinary0.990.80.830.5Table 4Differences in porosity maps between models PI and the reference CT model.RMSEMAEHISampling points#10.760.580.22338399#20.310.220.33455562#30.560.470.18450134#40.310.240.25396343#50.310.220.29472902 Table 2Similarity performance for PI, MRsimCT and MRBinary set.Target locationIndex#1#2#3#4#5AveragePerasonPcuPI0.980.950.920.970.980.96MRsimCT0.980.920.930.850.980.932MRBinary0.930.750.780.730.880.814VIM (128)PI0.990.970.970.960.980.974MRsimCT0.990.970.940.810.970.936MRBinary0.970.930.90.780.950.906DicePcuPI0.870.780.730.820.870.814MRsimCT0.880.730.740.650.840.768MRBinary0.750.540.560.570.690.622VIM (128)PI0.950.930.920.90.920.924MRsimCT0.950.920.860.820.930.896MRBinary0.920.860.860.770.880.858 Similarity performance for PI, MRsimCT and MRBinary set. Table 3Tukey’s multiple comparisons test.IntensityPositionDicePearsonPCuPI-MRsimCT0.640.410.210.3PI-MRBinary0.560.930.510.43MRsimCT-MRBinary0.980.60.780.96VIMPI-MRsimCT0.850.620.160.56PI-MRBinary0.810.940.060.11MRsimCT-MRBinary0.990.80.830.5 Tukey’s multiple comparisons test. Table 4Differences in porosity maps between models PI and the reference CT model.RMSEMAEHISampling points#10.760.580.22338399#20.310.220.33455562#30.560.470.18450134#40.310.240.25396343#50.310.220.29472902 Differences in porosity maps between models PI and the reference CT model. Discussion This study investigated the effectiveness of the porosity index (PI) method and UTE imaging in TUS. We found that this skull porosity measurement method, based on UTE images, can achieve similar acoustic simulation outcomes as with CT images-based treatments. The present study’s findings indicate that a combination of UTE images and the PI method may potentially serve as a replacement for CT images in practical application treatment. On the other hand, these studies on MRsimCT and Cortical bone porosity assessment with MR-UTE imaging22–24 inspired our method. In practice, the registration error between MR and CT always obstructs the treatment accuracy of TUS20,25–27. In addition, to protect patients, we should minimize the radiation dose they receive as much as possible. Thus, We apply the concept of the PI to establish an acoustic model. This method offers a potential alternative to previous synthetic CT approaches18,19 as it is derived from a different physical principle—the ratio of pore water to total water signal. In our simulations, the PI method demonstrated comparable or slightly improved performance relative to the MRsimCT method across several key metrics (Table 1), suggesting its potential utility. The variable performance of MRsimCT compared to the more consistent PI method can be attributed to fundamental differences in their underlying principles and sensitivity to image quality. The MRsimCT value, derived from the normalized difference between short- and long-TE signals (Eq. 3), is more susceptible to noise, B1 field inhomogeneities, and partial volume effects due to the lower spatial resolution of UTE-MR compared to CT. This can lead to instability in estimating Hounsfield Unit equivalents, particularly in regions with thin bone or complex anatomy. In contrast, the PI is a direct ratio of long-TE to short-TE signals (Eq. 2), inherently normalizing for some common noise and bias fields. This ratio more directly captures the pore water fraction, which is the key physical property linked to porosity and thus acoustic properties. Consequently, while MRsimCT can approximate CT-derived porosity reasonably well in ideal conditions, its performance degrades more significantly than the PI method under suboptimal or variable scanning conditions, leading to the observed fluctuations in focal pressure accuracy (e.g., PCu: PI 9.84% vs. MRsimCT 15.07% difference; VIM: PI 25.07% vs. MRsimCT 30.21% difference, Table 1) and spatial similarity (e.g., average PCu Dice: PI 0.814 vs. MRsimCT 0.768, Table 2) across subjects and targets. The comparable performance of the PI method is particularly interesting from a biophysical perspective. As suggested by Webb et al.28, the PI, derived from the ratio of pore water to total water signal, may provide a more direct estimate of bone’s acoustic properties because it is intrinsically tied to the micro-environment of water molecules within the bone pore structure. While our current results show a functional equivalence to MRsimCT for acoustic modeling, this underlying physical rationale suggests that the PI method could potentially offer more robust estimates across diverse patient populations or imaging conditions. However, validating this hypothesis and determining if PI provides a fundamental advantage in predicting acoustic properties requires further investigation. The MR result chronically underestimates the CT. This is because the lower spatial resolution of UTE-MR compared to CT may lead to partial volume effects, which smooth out high-porosity regions29. We cannot determine the true bone porosity measurements of the subjects. In traditional CT methods, porosity calculations are based on the assumption that water is present only in bone pores, which led to the development of Formula 1 in this manuscript. However, with insights from micro-CT observations, the rationality of the PI method is gradually gaining acceptance. In this study, CT served as the clinical reference standard. The systematic differences observed between PI- and CT-derived values reflect their distinct physical principles. Our primary aim was not to validate PI against a porosity ground truth, but to evaluate whether this radiation-free method could produce skull acoustic models with focusing performance comparable to CT-based models. This would establish PI as a viable alternative that circumvents CT’s limitations of radiation exposure and registration error. However, further validation and refinement are necessary to fully assess its potential to replace CT-based methods. By default, free water only exists in the skull pores, and the porosity is replaced by the ratio of pore water to total water signal. We experimentally obtained a linear relationship between the porosity of PI and CT and compared the acoustic simulation results obtained by the two methods. The results indicate a significant statistical correlation between PI- and CT-derived porosity. Despite these differences, the resulting acoustic simulations show comparable performance. Finally, in the acoustic simulation test, the model established by the PI method can restore 83% of the maximum focus intensity (normalized). Compared with previous statistical methods, such as inverse logarithm, our method is more consistent with physical logic. The PI method starts from basic physical principles and uses different echoes to collect signals of pore water and total water in the skull to calculate the porosity. Moreover, the PI method is simple regarding time and calculation difficulty. The double echo UTE sequence scanning time is generally approximately three minutes, which will not considerably increase the treatment time. Some limitations of the current study are the relatively small sample size and the lack of measurement of acoustic pressure using the hydrophone. In future work, we can also use MRI magnetization transfer technology to distinguish pore water and bound water signals, establish a dual-cell model, and use the macromolecular proton fraction to evaluate porosity to achieve better results than the PI method. Finally, a methodological consideration should be noted regarding the spatial resolution employed in our numerical simulations. The simulations were conducted with a spatial resolution of 3 points per wavelength in water, which represents the minimum requirement to satisfy the stability criteria of the k-Wave toolbox. At this resolution, the primary limitation is the inaccurate representation of wave interactions (e.g., reflection and transmission) at the complex, sharp boundaries of the skull, which can affect focusing accuracy30. While a higher resolution could potentially further enhance accuracy, it would require an approximately eight-fold increase in computational memory and time, rendering it infeasible for our full 3D cranial models with current resources. However, we emphasize that the primary objective of this study was a comparative analysis between the PI-based and CT-based acoustic modelling approaches. Since all simulations were performed under identical numerical conditions, including the same PPW value, any associated numerical errors would systematically affect both models equally. Therefore, the strong agreement observed in all focal metrics between the PI and CT methods is robust and indicates that the conclusions of this comparative study are valid despite this resolution constraint. Future work will prioritize higher-resolution simulations as computational resources allow. In conclusion, PI-based treatment planning achieves accuracy comparable to that of CT for ultrasound simulation results. This method reduces acoustic beam errors from image misalignment and avoids radiation exposure. The approach also benefits emerging neurological treatments like brain stimulation and blood-brain barrier opening, which are moving toward clinical application. Methods Image acquisition Five physically active subjects with no history of nervous system disease participated in the study. The Huaxi MR Research Center Investigational Review Board granted ethical clearance for all experimental protocols, with informed consent secured from each enrolled participant before study procedures. All methods were carried out in accordance with relevant guidelines and regulations. Double echo time 3-D UTE sequence MR images parameters were as follows: TR = 9 ms, flip angle = 12 degrees, TE = 0.05 and 4.6 ms, slice thickness = 1.8 mm, bandwidth = 1250 Hz/px and Field of View (FOV) 23 cm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 23 cm. CT images were acquired on a UIH uCT780 with slice thickness = 3 mm, KVP = 120 kV, X-ray tube current = 174 mA and matrix = 512 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 512. Acoustic model of skull The skull’s 3-D acoustic modeling framework involved CT-UTE multimodal registration and skull segmentation, implemented through FSL’s FLIRT module31 and MIPAV computational tools32. Critical acoustic parameters governing skull wave propagation—specifically wave velocity33, mass density, and the attenuation coefficients34,35 were incorporated to construct the cranial acoustic analog. Previous studies have shown that these parameters can be estimated from bone porosity. To date, this remains the only clinically validated method in practical applications36. Reference CT imaging The approximation of bone porosity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varphi$$\end{document} was determined from the CT Hounsfield unit:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \varphi =1-\frac{HU}{HU_{max}} \end{aligned}$$\end{document}where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$HU_{max}$$\end{document} represents the maximum HU value of 2000 observed in the clinical CT images of all five patients, indicating peak skull density. MR-UTE imaging The UTE images were used to generate the PI images using the following expressions:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} Porosity\ Index(\%)=\frac{TE_{long} intensity}{TE_{short} intensity}\times 100 \end{aligned}$$\end{document}where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE_{short}$$\end{document} is the signal intensity image from the short echo time, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE_{long}$$\end{document} is the signal intensity image from the long echo time. The skull cortex contains bound water, and the skull pores are full of free water. The free and bound water have different T2* values, allowing us to obtain the skull porosity through the UTE sequence. The first echo obtained under the shortest TE contains proton signals from all water, while the second echo obtained under the longer TE mainly comes from pore water. Therefore, the selected dual-echo UTE sequence can get images of two different TEs in one scan. A short TE value was chosen to obtain the bound water (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T2^*\sim 300 \mu s$$\end{document}) and pore water (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T2^*$$\end{document}>0.1ms). However, in the case of a long TE, the detected proton signal is essentially from the water in the pore space because the amplitude of the shorter T2* component can be ignored (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<5\%$$\end{document})16. Finally, the MR-simulated-CT (MRsimCT) images15 are as follows:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} MRsimCT=\frac{TE_{short} intensity-TE_{long} intensity}{TE_{short} intensity+TE_{long} intensity} \times HU_{bone} \end{aligned}$$\end{document}where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$HU_{bone}$$\end{document} is the maximum HU of bone (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$HU_{bone}=2000$$\end{document}). The PI images directly correlate with skull porosity, enabling the inference of acoustic characteristics and facilitating the establishment of skull acoustic models. Subsequently, the MRsimCT images rely on Eq. (3), following the same computational framework as reference CT approaches. For the MR-binary acoustic model, a simple thresholding approach was applied directly to the MRsimCT images to create a two-tissue model. Specifically, the MRsimCT images–after removal of air voxels–were processed using a thresholding algorithm. Voxels with values greater than 1000 HU were classified as bone and assigned a fixed value of 1000 HU, while those with values less than or equal to 1000 HU were classified as non-bone (approximated as water) and assigned a value of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-1000$$\end{document} HU. Using the method of Miller et al.29 to create MR binary image masks enabled a direct benchmarking of the results from this study against theirs. Acoustic property assignment Bone porosity was derived from CT and MRI signals by applying Eqs. 1 and 2, as described in the previous subsection. The speed of sound, density, and attenuation coefficient of the skull were estimated as follows:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} C_{skull}={\varphi } C_{water}+(1-\varphi ) C_{bone} \end{aligned}$$\end{document}5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \rho _{skull}={\varphi } \rho _{water}+(1-\varphi ) \rho _{bone}\end{aligned}$$\end{document}6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \alpha _{skull}=\alpha _{skull-min}+(\alpha _{skull-max}-\alpha _{skull-min})\varphi \end{aligned}$$\end{document}where C, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\rho$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document} are the compressional sound speed, density and the attenuation coefficient, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varphi$$\end{document} was bone porosity. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha _{skull-max}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha _{skull-min}$$\end{document} are the maximum and minimum attenuation coefficients. These equations were solved using the compressional acoustic properties of various tissues derived from Aubry et al.36. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha _{water}$$\end{document}was assumed to be zero. Simulation Numerical simulations of transcranial ultrasound propagation were performed using the k-Wave MATLAB toolbox (version 1.3)21, which solves the first-order acoustic wave equations using a pseudo-spectral, time-domain method. This approach allows for the modeling of wave propagation through heterogeneous media with defined acoustic properties. Parallel simulations were performed using a validated acoustic model8 with modified transducer and skull acoustic parameters. The simulation domain was defined by the segmented skull mask, embedded within a homogeneous water-like medium (speed of sound \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c_0 = 1500$$\end{document} m/s, density \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\rho _0 = 1000$$\end{document} kg/m3) representing the brain tissue. The spatial resolution was fixed at 2000 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}m (points per wavelength, PPW = 3), corresponding to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda /3$$\end{document} in water at 250 kHz, satisfying K-Wave’s stability criteria30. The temporal step \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta t = 2.38~\mu \text {s}$$\end{document} (points per period, PPP = 16.8) was automatically determined by the K-Wave toolbox to maintain a Courant–Friedrich–Lewy (CFL) condition of 0.3 within cortical bone, with total time steps also set to auto. The pressure field was recorded over the entire computational domain by defining a full-domain sensor mask. This allowed for the complete 3D pressure time-series to be captured, enabling post-hoc analysis at any point within the field. The complex pressure field \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p(\textbf{x}, t)$$\end{document} was extracted from the simulated waveforms. The peak positive pressure \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {peak}}$$\end{document} at each voxel was calculated as the maximum pressure value over the recorded time series. Phase aberration correction was performed for all simulations to compensate for distortions induced by the heterogeneous skull. This was achieved using a time-reversal approach, where a virtual source was first defined at the desired focal location. The acoustic waves propagating backward from this virtual source through the skull model were simulated to calculate the required phase corrections for each transducer element. These phase corrections were then applied to the transmitted signals to achieve constructive interference at the intended focus. Two sets of simulations were conducted using different transducer array configurations: a hemispherical array and a bowl-shaped array. The hemispherical array was defined by 128 point sources distributed over a hemisphere with a radius of 0.1 m. The bowl-shaped array featured an aperture diameter of 6 cm and a radius of curvature of 11 cm, created using the makeBowl function in k-Wave. The bowl source represents the entire concave surface vibrating in phase. Where, a 128-element hemispherical ultrasonic array (radius = 0.1 m) delivered 40 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}s tonebursts centered at 250 kHz. The use of a 40 µs pulse duration was justified based on the time required for the acoustic field to reach a steady state. A 40 µs tone burst contains 10 cycles at 250 kHz. Considering the propagation time and the number of cycles, the total time for the field to establish and stabilize exceeds the one-way propagation time. Therefore, the 40 µs duration, when combined with the recorded time window which encompasses the entire propagation, is sufficient to capture the fully developed pressure field at the focal region. Primary simulations targeted the ventral intermediate thalamic nucleus (VIM)—a validated focus for transcranial ultrasound management of essential tremor37—with stereotaxic coordinates defining a \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$70 \pm 5$$\end{document} mm axial span from the cranial vertex. Additional targets (e.g., precuneus, PCu) were evaluated using a single-element transducer (60 mm radius of curvature, 60 mm aperture diameter) with identical methodology. The workflow is illustrated in Fig. 5.Fig. 5Flow chart of the simulation, where reference CT simulation result denotes the gold standard. The blue dashed box denotes four different methods for building models. Flow chart of the simulation, where reference CT simulation result denotes the gold standard. The blue dashed box denotes four different methods for building models. Data analysis To quantitatively evaluate the agreement between the porosity maps derived from CT and UTE-PI methods, as well as the performance of the resulting acoustic simulations, the following metrics were calculated. Voxel-wise comparison of porosity maps For each subject, the co-registered CT-based porosity map (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi _{\text {CT}}$$\end{document}) and the UTE-PI-based porosity map (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi _{\text {PI}}$$\end{document}) were compared on a voxel-by-voxel basis within the segmented skull mask. The following standard error metrics were computed:Root Mean Square Error (RMSE): Measures the average magnitude of the differences between the two maps. 7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {RMSE} = \sqrt{ \frac{1}{N} \sum _{i=1}^{N} \left( \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right) ^2 } \end{aligned}$$\end{document} where N is the total number of voxels within the skull mask.Mean Absolute Error (MAE): Provides a more intuitive average of the absolute differences. 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {MAE} = \frac{1}{N} \sum _{i=1}^{N} \left| \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right| \end{aligned}$$\end{document} Root Mean Square Error (RMSE): Measures the average magnitude of the differences between the two maps. 7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {RMSE} = \sqrt{ \frac{1}{N} \sum _{i=1}^{N} \left( \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right) ^2 } \end{aligned}$$\end{document} where N is the total number of voxels within the skull mask. Root Mean Square Error (RMSE): Measures the average magnitude of the differences between the two maps. 7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {RMSE} = \sqrt{ \frac{1}{N} \sum _{i=1}^{N} \left( \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right) ^2 } \end{aligned}$$\end{document} where N is the total number of voxels within the skull mask. Mean Absolute Error (MAE): Provides a more intuitive average of the absolute differences. 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {MAE} = \frac{1}{N} \sum _{i=1}^{N} \left| \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right| \end{aligned}$$\end{document} Mean Absolute Error (MAE): Provides a more intuitive average of the absolute differences. 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {MAE} = \frac{1}{N} \sum _{i=1}^{N} \left| \phi _{\text {CT}, i} - \phi _{\text {PI}, i} \right| \end{aligned}$$\end{document} Distribution similarity The similarity between the histograms of the two porosity maps was quantified using the Histogram Intersection (HI) metric. Let \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H_{\text {CT}}(k)$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H_{\text {PI}}(k)$$\end{document} be the normalized counts of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k^{\text {th}}$$\end{document} bin of the histograms for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi _{\text {CT}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi _{\text {PI}}$$\end{document}, respectively, with K bins.9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {HI} = \sum _{k=1}^{K} \min \left( H_{\text {CT}}(k), H_{\text {PI}}(k) \right) \end{aligned}$$\end{document} The HI value ranges from 0 (no overlap) to 1 (identical distributions). Acoustic simulation performance metrics The quality of the focused ultrasound field generated by each acoustic model was assessed using the following metrics, calculated on the simulated complex pressure field p.Normalized Peak Intensity (NPI): The intensity at the focal peak (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}} = |p_{\text {peak}}|^2$$\end{document}) normalized to the peak intensity achieved by the gold-standard CT-based simulation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}). 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NPI} = \frac{I_{\text {peak}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document}Normalized Target Intensity (NTI): The intensity at the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {target}}$$\end{document}) normalized to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}. 11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NTI} = \frac{I_{\text {target}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document}Focal Positioning Error (FPE): The Euclidean distance between the location of the peak intensity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {peak}}$$\end{document}) and the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {target}}$$\end{document}). 12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {FPE} = \Vert \textbf{x}_{\text {peak}} - \textbf{x}_{\text {target}}\Vert _2 \end{aligned}$$\end{document}Dice Similarity Coefficient (DSC): Measures the volumetric overlap of the focal volume. The focal volume was defined as the region where the acoustic pressure exceeds 50% of the peak pressure (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|p_{\text {peak}}|$$\end{document}) in each simulation. 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {DSC} = \frac{2 \times \left| V_{\text {sim}} \cap V_{\text {CT}} \right| }{ \left| V_{\text {sim}} \right| + \left| V_{\text {CT}} \right| } \end{aligned}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {CT}}$$\end{document} are the binary masks of the focal volumes for the test simulation and the CT-reference simulation, respectively.Pearson Correlation Coefficient (PCC): Calculated on the pressure values within a volumetric region of interest (ROI) encompassing the main focal lobe. 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {PCC} = \frac{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})(p_{\text {CT}, i} - \bar{p}_{\text {CT}}) }{ \sqrt{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})^2 } \sqrt{ \sum _{i=1}^{M} (p_{\text {CT}, i} - \bar{p}_{\text {CT}})^2 } } \end{aligned}$$\end{document} where M is the number of voxels in the ROI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {CT}}$$\end{document} are the pressure values from the test and reference simulations, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{p}$$\end{document} denotes the mean pressure within the ROI.Statistical comparisons of these acoustic metrics across the different modeling methods were performed using a one-way repeated measures ANOVA followed by Tukey’s honestly significant difference test for post-hoc multiple comparisons. A p-value of less than 0.05 was considered statistically significant. Normalized Peak Intensity (NPI): The intensity at the focal peak (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}} = |p_{\text {peak}}|^2$$\end{document}) normalized to the peak intensity achieved by the gold-standard CT-based simulation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}). 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NPI} = \frac{I_{\text {peak}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document} Normalized Peak Intensity (NPI): The intensity at the focal peak (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}} = |p_{\text {peak}}|^2$$\end{document}) normalized to the peak intensity achieved by the gold-standard CT-based simulation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}). 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NPI} = \frac{I_{\text {peak}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document} Normalized Target Intensity (NTI): The intensity at the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {target}}$$\end{document}) normalized to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}. 11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NTI} = \frac{I_{\text {target}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document} Normalized Target Intensity (NTI): The intensity at the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {target}}$$\end{document}) normalized to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {peak}}^{\text {CT}}$$\end{document}. 11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {NTI} = \frac{I_{\text {target}}}{I_{\text {peak}}^{\text {CT}}} \end{aligned}$$\end{document} Focal Positioning Error (FPE): The Euclidean distance between the location of the peak intensity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {peak}}$$\end{document}) and the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {target}}$$\end{document}). 12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {FPE} = \Vert \textbf{x}_{\text {peak}} - \textbf{x}_{\text {target}}\Vert _2 \end{aligned}$$\end{document} Focal Positioning Error (FPE): The Euclidean distance between the location of the peak intensity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {peak}}$$\end{document}) and the intended target location (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textbf{x}_{\text {target}}$$\end{document}). 12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {FPE} = \Vert \textbf{x}_{\text {peak}} - \textbf{x}_{\text {target}}\Vert _2 \end{aligned}$$\end{document} Dice Similarity Coefficient (DSC): Measures the volumetric overlap of the focal volume. The focal volume was defined as the region where the acoustic pressure exceeds 50% of the peak pressure (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|p_{\text {peak}}|$$\end{document}) in each simulation. 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {DSC} = \frac{2 \times \left| V_{\text {sim}} \cap V_{\text {CT}} \right| }{ \left| V_{\text {sim}} \right| + \left| V_{\text {CT}} \right| } \end{aligned}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {CT}}$$\end{document} are the binary masks of the focal volumes for the test simulation and the CT-reference simulation, respectively. Dice Similarity Coefficient (DSC): Measures the volumetric overlap of the focal volume. The focal volume was defined as the region where the acoustic pressure exceeds 50% of the peak pressure (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|p_{\text {peak}}|$$\end{document}) in each simulation. 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {DSC} = \frac{2 \times \left| V_{\text {sim}} \cap V_{\text {CT}} \right| }{ \left| V_{\text {sim}} \right| + \left| V_{\text {CT}} \right| } \end{aligned}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\text {CT}}$$\end{document} are the binary masks of the focal volumes for the test simulation and the CT-reference simulation, respectively. Pearson Correlation Coefficient (PCC): Calculated on the pressure values within a volumetric region of interest (ROI) encompassing the main focal lobe. 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {PCC} = \frac{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})(p_{\text {CT}, i} - \bar{p}_{\text {CT}}) }{ \sqrt{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})^2 } \sqrt{ \sum _{i=1}^{M} (p_{\text {CT}, i} - \bar{p}_{\text {CT}})^2 } } \end{aligned}$$\end{document} where M is the number of voxels in the ROI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {CT}}$$\end{document} are the pressure values from the test and reference simulations, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{p}$$\end{document} denotes the mean pressure within the ROI. Pearson Correlation Coefficient (PCC): Calculated on the pressure values within a volumetric region of interest (ROI) encompassing the main focal lobe. 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \text {PCC} = \frac{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})(p_{\text {CT}, i} - \bar{p}_{\text {CT}}) }{ \sqrt{ \sum _{i=1}^{M} (p_{\text {sim}, i} - \bar{p}_{\text {sim}})^2 } \sqrt{ \sum _{i=1}^{M} (p_{\text {CT}, i} - \bar{p}_{\text {CT}})^2 } } \end{aligned}$$\end{document} where M is the number of voxels in the ROI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {sim}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\text {CT}}$$\end{document} are the pressure values from the test and reference simulations, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{p}$$\end{document} denotes the mean pressure within the ROI.