Fiber-Tip Surface-Micromachined Optical Ultrasound Transducer (SMOUT) Probe for Acoustic Detection Induced by Ultrahigh Dose Rate (UHDR) Electron Beam

Abstract

Online dose mapping is essential for safe and accurate beam delivery in FLASH radiation therapy (RT). While ionizing radiation acoustic imaging (iRAI) enables real-time, deep tissue dose measurement and beam localization for RT, previous piezoelectric acoustic sensors have suffered from electronic noise and neutron damage from ultra-high dose rate radiation (UHDR) beam. Described here is an innovative single-element full-optical surface-micromachined optical ultrasound transducer (SMOUT) mounted on an optical fiber tip to detect acoustic waves induced by UHDR beam. The design and fabrication of the fiber-tip SMOUT probe (diameter = 400 μm) have been demonstrated, and its optical and acoustic characteristics, as well as transmittance to the FLASH electron beam and imaging capability have been characterized. Experimental results show that the fiber-tip SMOUT probe has optical resonance wavelength (ORW) ~ 773 nm, acoustic central frequency ~ 300 kHz and −6 dB acoustic bandwidth ~ 160 kHz, and dose rate independent single pulse noise equivalent dose ~ 5.19 cGy. More importantly, the fiber-tip SMOUT probe exhibits negligible disturbance to FLASH irradiation. After 1-D scanning the fiber-tip SMOUT probe, the 2-D dose distribution of the FLASH electron beam can be mapped with high consistency. These findings highlight the potential of SMOUT based iRAI for mapping the dose distribution under UHDR beam, which is an important step toward practical implementation of iRAI-based online dosimetry in FLASH RT.

Graphical Abstract

I. Introduction

Flash radiation therapy (RT), a novel approach delivering ultra-high dose rate (UHDR) radiation beams (>40 Gy/s), has emerged as a transformative advancement in the field [1]. Preclinical and clinical pilot studies have demonstrated that FLASH RT can substantially reduce normal tissue toxicity while maintaining, or even enhancing, tumor control, with a phenomenon termed the “FLASH effect” [2]. This capability holds promise for expanding the therapeutic window and revolutionizing clinical radiation oncology. However, the unique physiobiological mechanisms of FLASH RT, coupled with rapid dose delivery, present significant challenges for real-time dosimetry and treatment verification [3].

Existing clinical dosimetry devices, including diodes, ionization chambers and scintillators [4] [5] [6], are not optimized for the ultra-short irradiation timescales and UHDR characteristic of FLASH RT. Most conventional dosimetry systems are also only available for point measurements, lack volumetric resolution, and may suffer from response saturation and dose-rate dependence, rendering them unsuitable for accurate and real-time monitoring of FLASH dose delivery [3] [7] [8] [9]. Efforts have been made to optimize the current dosimetry methods for FLASH RT [10] [11] [12], however, the lack of real-time, pulse-by-pulse, in-vivo volumetric mapping of dose deposition still hinders rigorous evaluation of biological responses and the safe clinical translation of FLASH RT [13]. Therefore, there is a critical need for dosimetric technologies that can provide immediate and volumetric feedback during FLASH RT—enabling precise verification of dose distribution and supporting robust investigation of the FLASH effect in tissue.

Ionizing radiation acoustic imaging (iRAI) is a novel imaging technique, which detects acoustic waves generated by the rapid thermal expansion of tissue following absorption of ionizing radiation pulses, such as X-rays, electron or proton beams [14] [15] [16] [17]. Previous studies have shown the promise of iRAI for real-time volumetric measurement of deep-tissue dose and localization of beam during both conventional and FLASH RT [18] [19]. However, the piezoelectric transducers have struggled with electronic noise and neutron damage from UHDR beam [20]. To address this issue, a new surface-micromachined optical ultrasound transducer (SMOUT) has been developed [21] [22] [23] [24] [25], whose acoustic signal is full-optically readout and naturally resistant to electromagnetic interference and neutrons from UHDR beam. Moreover, the SMOUT element can be miniaturized while maintaining sensitivity, which is promising to be integrated as a much smaller and more compact array package, compared with the current piezoelectric transducer array.

In this article, we report a novel full-optical fiber-tip SMOUT probe for detecting the acoustic waves induced by UHDR electron beam. First, the design, fabrication, and assembly of the fiber-tip SMOUT probe are presented. Second, its optical and acoustic performances are characterized. The performances, including radiation transparency, sensitivity and noise-resistance, are then evaluated under a UHDR electron beam. Finally, the fiber-tip SMOUT probe is scanned across a line to mimic a linear array to map the dose distribution of UHDR electron beam in water. This study represents an important step toward practical implementation of iRAI based online dosimetry in FLASH RT.

II. Methods

A. SMOUT Design and Fabrication

As shown in Fig. 1(a), the SMOUT is composed of a Fabry-Perot (F-P) cavity bounded by a top and a bottom distributed Bragg reflector (DBR) layer [21] [23]. When the light is incident onto the SMOUT, the optical reflectivity is wavelength-dependent, due to the interferences between the two DBRs. The interference of beams reflected at interfaces with different refractive index is destructive at certain wavelength with local minimum reflectivity, which is defined as the optical resonance wavelength (ORW) of the SMOUT, denoted as λ₀ in Fig. 1(b). The ORW is determined by the distance between the top and bottom DBRs, as well as the thickness and refractive indices of the thin films forming the DBRs [24]. When an ultrasound wave impinges on the SMOUT, vibration of the top membrane alters the cavity length, resulting in a shift of the reflectance spectrum (blue and red dashed curves in Fig. 1(b)). The quadrature point of the spectrum slope, λ_bias, is selected as the wavelength to interrogate the received ultrasound signal, where the intensity of the reflected light is almost linearly modulated by the impinged ultrasound (Fig. 1(b)). Alternatively, the wavelength located symmetrically on the opposite side of λ₀ (marked as a black dot on the green curve in Fig. 1(b)) can also serve as λ_bias, though the resulting signal amplitude will have the opposite polarity.

Fig. 1.

(a) The cross-section diagram of a SMOUT whose top diaphragm (DBR + parylene) is vibrated by the impinging ultrasound wave. DBR: distributed Bragg reflector. (b) The SMOUT reflectance spectrum shifted by the top diaphragm vibration, where λ₀ and λ_bias is interrogation optical wavelength, and respectively. (c) Photo of a representative 400-μm-ϕ SMOUT coated with 18-μm-thick parylene and sealed by water-proof glue.

The SMOUT is fabricated based on a six-step process [22] [24]. First, multiple pairs of silicon dioxide/nitride (SiO/SiN) layers are consecutively deposited on a double-side-polished borosilicate wafer substrate as the bottom DBR by plasma-enhanced chemical vapor deposition (PECVD). Second, a sacrificial layer is sputtered onto the bottom DBR and patterned by photolithography. Third, the top DBR is deposited onto the sacrificial material by PECVD. Fourth, after the wafer is diced into individual elements, the sacrificial layer is wet etched to release the top DBR as a suspended diaphragm. Next, a Parylene layer is coated on the top DBR to suppress the unwanted high-order modes of vibrations, thereby enhancing sensitivity at the fundamental mode, while also providing mechanical protection against damage and contamination. Finally, after the SMOUT is mounted onto the fiber tip, the etching hole is sealed by water-proof glue to complete the fabrication. Fig. 1(c) shows a representative 400-μm-ϕ SMOUT coated with 18-μm-thick Parylene and sealed with glue, with overall dimensions of approximately 2.0 × 1.7 × 0.3 mm³.

B. SMOUT Reflectivity Spectrum

Before mounting the SMOUT onto the tip of the optical fiber, its reflectivity spectrum is characterized to ensure that the interrogation wavelength λ_bias is available. As shown in Fig. 2(a), continuous-wave (CW) white light is directed from Port 1 to 2 of a multimode fiber circulator, collimated by a fiber collimator, and then focused by an objective lens onto the SMOUT. The light reflected from the SMOUT travels along reverse path from Port 1 to 3 and is recorded by an optical spectrometer. The reflectivity spectrum of the SMOUT is determined from three sequential measurements. First, with no reflector in place, the recorded signal is designated as the background (S_background), which is due to internal reflections within optical components. Second, the measurement is repeated by using a mirror as an ideal reflector. The light reflected by the mirror only is S_mirror − S_background, which represents the light incident onto the SMOUT. Third, the measurement is repeated by using the SMOUT as the reflector. The light reflected by the SMOUT only is S_SMOUT − S_background. Accordingly, the SMOUT reflectivity (R_SMOUT) is calculated as

$$R_{\mathit{\text{SMOUT}}}=\frac{S_{\mathit{\text{SMOUT}}}-S_{\mathit{\text{background}}}}{S_{\mathit{\text{mirror}}}-S_{\mathit{\text{background}}}}$$ (1)

Fig. 2.

(a) Diagram of the setup to characterize the reflectivity spectrum of the SMOUT. CW: continuous wave. (b) A measured reflectivity spectrum of the SMOUT.

Fig. 2(b) shows the reflectivity spectrum of the representative SMOUT in Fig. 1(c). There are two ORWs (~ 668 and 772.58 nm) existing in the spectrum, each allowing the interrogation of the ultrasound signal received by the SMOUT at two quadrature points (662, 671, 768, and 777 nm). For simplicity, only one ORW (772.58 nm) and its corresponding λ_bias (777 nm) are indicated in Fig. 2(b), with 777 nm selected for ultrasound readout in this study.

C. SMOUT Assembly onto Optical Fiber Tip

Ultraviolet (UV) epoxy is used to attach the 400-μm-ϕ SMOUT to the tip of the 200-μm-ϕ optical fiber. Precise alignment is essential to maximize the coupling of the interrogation light and also the ultrasound detection sensitivity. To achieve this, a setup resembling an inverted microscope is built, with the objective lens and CCD camera positioned beneath a transparent stage and oriented upward (Fig. 3(a)). First, the SMOUT substrate is flipped and placed on the transparent stage with the SMOUT element facing downward. Second, the optical fiber, positioned above the stage with its tip facing downward and coated in uncured UV epoxy, is gradually lowered until it contacts the SMOUT’s glass substrate. Third, white light is coupled into the opposite end of the optical fiber, so that the optical fiber (indicated by the white light spot) and the SMOUT can be observed simultaneously. The stage is then translated horizontally to align the SMOUT concentrically with the optical fiber. Finally, the epoxy is cured under UV illumination to permanently bond the SMOUT and fiber, and the etching hole is sealed with waterproof glue. The photo of the assembled SMOUT on the optical fiber tip is shown in Fig. 3(b).

Fig. 3.

(a) A simplified diagram of the SMOUT assembly onto optical fiber tip with an inverted microscope. (b) Photo of the side view of the SMOUT mounted on the tip of the optical fiber.

III. Acoustic Testing and Results

The acoustic center frequency (f_c) and bandwidth (BW) of the assembled fiber-tip SMOUT probe are characterized using the photoacoustic (PA) testing setup shown in Fig. 4(a). A xenon flash lamp (Hamamatsu L7684, pulse width ~ few μs) is used as the light source for PA excitation. The flash lamp, instead of a pulsed laser, is selected as the light source, because its pulse width (a few μs) is relatively close to that of the FLASH radiation source, whereas the pulse width of typical pulsed lasers is on the order of a few ns. This allows the flash lamp to better emulate the FLASH source, thereby providing more reliable estimates of the fiber-tip SMOUT probe’s acoustic performance (e.g. center frequency (f_c) and bandwidth (BW)) under the FLASH radiation. The light from the flash lamp is coupled through a light guide onto a piece of black tape. Due to the large divergence of the flashlight exiting the guide, the resulting energy density on the black tape is relatively low (a few mJ/cm† per pulse). A representative PA signal received by the SMOUT, along with its frequency spectrum obtained via fast Fourier transform (FFT), is shown in Fig. 4(b). The results indicate an f_c of approximately 300 kHz and a 6-dB BW of about 160 kHz. The noise equivalent pressure (NEP), which is defined as the minimal detectable pressure of the fiber-tip SMOUT probe, is characterized to estimate its acoustic sensitivity. A transmitting-receiving setup is built in water, using a piece of 6.5-mm-thick lead zirconate titanate (PZT) (f_c ≈ 300 kHz) and a 0.2-mm needle hydrophone (NH0200, Precision Acoustics) as the acoustic transmitter and receiver, respectively. The peak pressure received by the hydrophone is 19.8 kPa. The hydrophone is then replaced by the fiber-tip SMOUT probe at the same location to ensure that the acoustic pressure arriving at the hydrophone and the SMOUT is the same. Based on the peak-to-peak amplitude of the signal and the root mean square (RMS) of the noise, the NEP of the fiber-tip SMOUT probe is determined to be around 16 Pa and 8 Pa with one and four times signal averaging, respectively, over a frequency range of 1 kHz to 500 kHz.

Fig. 4.

(a) A diagram of the PA setup to characterize the acoustic f_c and BW of the fiber-tip SMOUT probe, using a piece of black tape as the target upon flashlight illumination. (b) Representative PA signal and its spectrum received by the fiber-tip SMOUT probe, indicating f_c ≈ 300 kHz, and −6dB BW ≈ 160 kHz.

IV. FLASH Experiment

A. Radiation Transparency of Fiber-tip SMOUT Probe

The radiation transparency of the fiber-tip SMOUT probe is a critical parameter for future clinical implantation of iRAI in online dose monitoring. To evaluate its radiation transparency, the fiber-tip SMOUT probe was placed into the path of 6-MeV FLASH electron beam, generated by a modified clinical linear accelerator (LINAC) (Clinac, Varian Medical Systems) with a 2 cm × 2 cm copper cutout collimator. 6 MeV is the lower bound of the typical FLASH electron beam energy. As the beam energy increases, its penetration depth in a material also increases and the proportion of energy lost in the SMOUT decreases. In this case, testing at 6 MeV provides a conservative, worst-case assessment of the transparency of the fiber-tip SMOUT probe to the electron beam. As shown in Figs. 5(a) and 6(a), the electron fluence passing through the fiber-tip SMOUT probe was recorded in both traverse and axial planes via radio-chromic film (EBT3, Ashland Inc.), which is a standard method to monitor the delivered dose of the radiation beam. During the measurement, the fiber-tip SMOUT probe was placed close to the center of the field and parallel to the beam path. The distance between the film placed on the transverse plane and the collimator is 1 cm, which is the same as that between the top edge of the film placed on the axial plane and the collimator. After exposure, the film is loaded into an Epson Expression 10000XL scanner (Epson America, Inc.) and digitized into a 2D map of the dose distribution with FilmQA Pro software (Ashland Inc.).

Fig. 5.

(a) Diagram of the transverse setup, and (b) the measured dose distribution in XY plane. (c) The comparisons between the two cross-section profiles over the whole depth and around the shadow along lines A and B in (b).

Fig. 6.

(a) Diagram of the axial cross-section setup, and (b) the measured dose distribution in XZ plane. (c) The comparisons between the two cross-section profiles over the whole depth and around the shadow along lines A and B in (b).

Figs. 5(b) and 6(b) show the mapped dose distribution of the transverse and axial planes, respectively. In both images, there exists a faint shadow (indicating lower dose deposition) caused by the fiber-tip SMOUT probe (pointed by the white arrow). The dose impacted by the fiber-tip SMOUT probe is investigated by comparing the cross-sectional profiles along the two white dashed lines A and B (Figs. 5(b) and 6(b)). Line A was extracted from the location where the fiber-tip SMOUT was positioned, 5 mm away from the center of the beam. Line B was extracted from a location also 5 mm from the center of the beam but situated in the direction opposite to the fiber-tip SMOUT. In Line A, the dose deposition was partially blocked by fiber-tip SMOUT probe. Line B presents the adjacent region without the interference of fiber-tip SMOUT probe. The maximum dose difference on axial plane induced by fiber-tip SMOUT probe is 4.1% with a total affected depth around 6.5 mm (Fig. 6(c)). The depth beyond 6.5 mm does not show significant difference due to the scattering of the electron beam. The similarity between the two profiles is quantified by three parameters. First, Pearson product-moment correlation coefficient R (∈ [–1,1]), which describes the shape/linear correlation between the two curves, and R closer to 1 means more consistency (insensitive to translation and scaling). Second, cosine similarity cos θ (∈ [–1,1]), which describes the direct alignment of the two curves, and cos θ closer to 1 means better alignment. Third, normalized root mean square error (NRMSE), which describes the average magnitude of errors between the two curves, and an NRMSE closer to 0 means better overlapping. For the transverse setup, the Pearson R = 0.9998 and 0.8519, NRMSE = 0.0107 and 0.2062, cosine similarity = 0.9999 and 0.9999 over the whole depth and around the shadow, respectively (Fig. 5(c)). For the axial setup, the Pearson R = 0.9995 and 0.7225, NRMSE = 0.0112 and 0.3562, cosine similarity = 0.9997 and 0.9998 over the whole depth and around the shadow, respectively (Fig. 6(c)). Therefore, the disturbance from the fiber-tip SMOUT probe to the dose delivery in both transverse and axial planes is almost negligible.

B. iRAI Experiment

1). Setup

Fig. 7 shows the schematic diagram and a photo of the iRAI setup incorporating the fiber-tip SMOUT probe. A tank filled with water is used to mimic the dose deposition in tissue (due to similar physical properties with soft tissue). An 18-MeV UHDR electron beam with 0.5 Gy per pulse generated from modified LINAC (with the same 2 cm × 2 cm copper cutout collimator) was horizontally hit to the left surface of the water tank (Fig. 7(a)). To avoid interference of the tank material, a 5 cm × 5 cm square was removed from the left wall of the tank and replaced with plastic film with 0.1-mm thickness. The fiber-tip SMOUT probe was immersed in the water with its detection surface perpendicularly to the incident FLASH electron beam to ensure optimal acoustic reception and consistent angular sensitivity. The fiber was mounted on to an optical rod driven by a two-axis translational stage to scan around X and Z axes. The radiation-induced acoustic signal was read out by interrogating the fiber-tip SMOUT probe with a near-infrared (NIR) continuous-wave (CW) laser at 777 nm (785HP, BroadSweeper, SUPERLUM). The interrogation laser is coupled into Port 1 of a multimode-fiber circulator, transmitted through Port 2 to the fiber-tip SMOUT probe, and partially reflected back after modulation by the incident acoustic wave. The modulated light traveled back through Ports 2 and 3, where it is detected by a photodetector (PD), pre-amplified, and recorded on a high-speed digital oscilloscope (MSO54B, Tektronix). The synchronization between FLASH electron beam delivery and data acquisition was precisely controlled via trigger from the LINAC.

Fig. 7.

(a) Schematic diagram and (b) photo of the iRAI setup with LINAC and the fiber-tip SMOUT probe. The SMOUT scanning along Z and X axis is indicated by the white solid and dashed arrow, respectively. PD: photodetector.

2). A-line Signal

The fiber-tip SMOUT probe is scanned along the Z axis (white solid arrow in Fig. 7(a)), with 1-cm increment over a 9-cm range to detect the radiation-induced acoustic (RA) signal along the dose depth inside the water and system noise under different dose rate. Fig. 8(a) shows a representative time-domain signal detected by the fiber-tip SMOUT probe after 40−500 kHz bandpass filtering without averaging. RA1 at 0 μs is the RA signal excited from the water close to the fiber-tip SMOUT probe upon FLASH illumination. RA2 at ~ 40 μs is that from the water close to the tank wall to the SMOUT after propagation through water. Signals between RA1 and RA2 originate from source points located between the fiber-tip SMOUT probe and the tank wall.

Fig. 8.

(a) A representative time-domain signal received by the fiber-tip SMOUT probe, where RA1 is directly excited from the water close to the SMOUT upon FLASH illumination, and RA2 is that from the water close to the tank wall (on the side of LINAC) and then propagated to the SMOUT. (b) The peak-peak amplitude of RA1 signal and background noise vs. FLASH dose depth by scanning the fiber-tip SMOUT probe from 1 cm to 10 cm along Z axis (white solid arrow in Fig. 7(a)).

The peak-peak amplitudes of RA1 and the background noise along the Z axis are plotted in Fig. 8(b). The amplitude of RA1 reaches the highest in the proximal region (around 2cm depth), where the 18-MeV FLASH electron beam releases the highest energy. As the fiber-tip SMOUT probe is positioned deeper along the beam direction, the RA1 amplitude continues to drop due to the reduction of electron energy release. The overall changes of the RA1 amplitude detected by the fiber-tip SMOUT probe is consistent with the typical percent depth dose (PDD) of 18-MeV electron beam in water [26] [27]. The background noise (RMS value of the waveforms later than 100 μs where the ultrasound signal has died out) remains low and consistent across the 1–10 cm depth with 51.9 mV (mean) ± 19.1 mV (std). Based on the 0.5 Gy dose at the entrance (dose depth = 1 cm), it can be concluded that the minimum detectable dose of the fiber-tip SMOUT probe at the entrance (noise-equivalent dose) is around 5.19 cGy.

3). 2D Imaging

To investigate the imaging capability of the fiber-tip SMOUT probe in mapping the dose during FLASH RT, the probe was placed at 10-cm dose depth and scanned along X-axis (white dashed arrow in Fig. 7(a)) with a 3-mm step size over a 78-mm range to mimic a linear array. An 18-MeV FLASH electron beam with a 2 cm × 2 cm cooper cutout was used to generate a dose gradient inside the water. The cross-section dose distribution along the depth of the electron beam was recorded by EBT3 radio-chromic film.

The iRAI image (Fig. 9(a)) was reconstructed based on the SMOUT-acquired A-line signals and the Delay-and-Sum (DAS) algorithm. Fig. 9(b) presents the reference dose distribution inside water measured by the radio-chromic film, which was located at the same plane as that in Fig. 9(a). To assess the depth-resolved performance of the fiber-tip SMOUT probe, the central-axis profile along the white dash-dotted line in the iRAI image is compared with that in reference (Fig. 9(c)), which can be treated as the PDD curve of the FLASH beam. The two profiles closely match with each other from 0 – 35 mm, with modest deviations in the far field (35–80 mm). Over the whole range, the two profiles have Pearson R = 0.9740, RMSE = 9.67%, and cosine similarity = 0.9913, revealing their strong shape agreement and minimal systematic bias. These results indicate that the fiber-tip SMOUT probe achieves high agreement in clinically relevant depths and minor deviations in the far field with the reference PDD. Besides 1-D comparison, a gamma-index analysis was used to evaluate the 2D consistency between the iRAI and reference dose. An index of 80.06% with criteria of 5mm/5% DTA/DD indicates that the two mappings also have relatively good consistency.

Fig. 9.

Measured 2D dose distribution of the FLASH electron beam in water based on (a) the iRAI image of the fiber-tip SMOUT probe and (b) reference radio-chromic film. (c) The central-axis profile along the white dash-dotted line in the iRAI image (a) is compared with that in (b), which can be treated as the PDD curve of the FLASH beam.

V. Conclusion and Discussion

Ionizing radiation acoustic imaging has shown great potential for addressing the dosimetry challenges in UHDR radiation therapy by using piezoelectric transducers to detect the radiation-indued acoustic signal. To address electronic noise and neutron damage induced by UHDR beam under FLASH RT, an innovative all-optical ionizing-radiation acoustic imaging system based on fiber-tip SMOUT probe has been demonstrated. The fiber-tip SMOUT probe is designed, fabricated, and characterized with optical and acoustic measurements. More importantly, the key factors, including radiation transparency, noise resistance, sensitivity and imaging capability, were evaluated with a UHDR electron beam generated from a modified LINAC.

As demonstrated by the results, a fiber-tip SMOUT probe with 400-μm diameter to minimize the beam interference was custom designed for iRAI application under UHDR electron beam. To maximize the sensitivity of detecting the acoustic signals induced by the 4-μs radiation pulse, the central frequency and the bandwidth of the fiber-tip SMOUT probe were set to 300 kHz and 160 kHz, respectively. The transverse and axial disturbance of fiber-tip SMOUT probe to dose distribution evaluated with radiographic film under UHDR electron beam has shown maximum dose difference of 3.6% and 4.1%, respectively, which is lower than the clinical acceptable variation of 5% [28] and can be further reduced in the future. The high similarity of the beam profiles with and without the SMOUT present further demonstrated the effective radiation transparency of fiber-tip SMOUT probe. A consistent system noise of 51.9 mV (mean) ± 19.1 mV (std) was observed under different levels of radiation dose, which demonstrates its dose rate independent anti-noise performance. The ability to capture single-pulse RA signal with a noise-equivalent dose of 5.19 cGy further demonstrates high sensitivity and effective noise/interference rejection. These characteristics are advantageous for FLASH radiotherapy dosimetry, where only a limited number of radiation pulses (e.g., <10) are delivered and pulse by pulse dose monitoring is critical for efficacy and safety. By scanning the single-element fiber-tip SMOUT probe to mimic a linear array, its imaging capability of dose mapping was demonstrated. A gamma index of 80.06% using the DTA/DD 5mm/5% criteria was achieved compared with the referenced film, indicating that the iRAI mapped dose distribution based on fiber-tip SMOUT probe reproduces the overall features of dose distribution observed on the film. Another important point is the universality of proposed probe for different radiation beam structures. For a 4-μs pulse, the central frequency of the generated acoustic wave is approximately 300 kHz. Although the probe described in this manuscript has a central frequency of 300 kHz, the acoustic wave generated by pulsed radiation beams is broadband, allowing it to work with different pulse durations. However, the sensitivity of the proposed system may decrease when used with other pulse durations. Therefore, a pre-calibration of the probe’s sensitivity is necessary for precise dose mapping.

Although the key performances of fiber-tip SMOUT probe have been demonstrated via proof-of-concept experiments, there are still several limitations that could be addressed by future development. The current method of scanning with the single-element fiber-tip SMOUT probe is limited by a long acquisition time, requiring a large number of radiation pulses to capture a dose image. In the meantime, although no radiation damage and sensitivity degeneration was noticed during the experiment, the long-term radiation effect, including the cumulative radiation damage, was not fully evaluated. Additionally, the accuracy of the mapped dose distribution based on this scanning technique does not yet meet clinical criteria. These discrepancies are primarily due to sparse spatial sampling—using a 3-mm sampling interval and only 26 positions in the delay-and-sum (DAS) iRAI reconstruction, which leads to undersampling of high spatial frequencies near the beam penumbra. Moreover, due to air sealed inside the SMOUT cavity, the reflectivity spectrum and ORW will be affected by the ambient temperature. In principle, this temperature-induced variation could be mitigated by actively controlling the gas pressure inside the cavity (e.g., via a microfluidic channel connected to the SMOUT element). The detailed analysis and quantitative characterization of the dependence of the SMOUT spectrum and ORW on ambient temperature, as well as potential mitigation strategies, will be investigated in our future work. Furthermore, to achieve high-accuracy mapping of the dose delivery within a limited number of FLASH pulses or even with a single frame per pulse, a 2D array of the fiber-tip SMOUT probes will be developed. Each element in the array will be equipped with an independent optoelectronic channel for parallel data acquisition. In addition to hardware improvements, future work will target physics-informed wavefield inversion or PSF-constrained deconvolution to reduce blur and sidelobes, as well as calibration of sound speed and frequency-dependent attenuation. Optimization of time-gain compensation (TGC) and amplitude-to-dose mapping will also be pursued. These refinements are expected to enhance image resolution, improve quantitative accuracy, and increase the robustness and reproducibility of 3D FLASH dose imaging.

In summary, an all-optical ionizing radiation acoustic sensor based on fiber-tip SMOUT probe has been developed to address the limitations of piezoelectric transducers in iRAI under UHDR beam. Despite the fact that both the system resolution and elements for transducer array can be improved in the future, the current experiments performed on phantom with UHDR beam have demonstrated the system capabilities for mapping the dose deposition during FLASH RT with high radiation transparency, high noise resistance, and high sensitivity. This new development presents an important step for practical implementation of iRAI based online dosimetry in FLASH RT.

Biographies

Cheng Fang (M’20) received his B.E. degree in Information Engineering from Zhejiang University, Hangzhou, Zhejiang, China, in 2011. He received his MEng. and Ph.D. degree in Electrical Engineering from the Texas A&M University, College Station, TX, USA, in 2013 and 2023, respectively. He continues as a Postdoc Fellow with research interests in robotic sensing, optoacoustic imaging, and microfabrication technologies.

Borui Li received his B.S. degree from Changchun University of Science and Technology in 2022. He received his M.S. degree in Electrical and Computer Engineering from University of Michigan in 2024. He then began his doctoral studies at Texas A&M University with research interests in micro-optical devices fabrication and applications for biomedical imaging. He is currently a Ph.D. student at Texas A&M University.

Dale W. Litzenberg joined the Department of Radiation Oncology in 1997 after completing his PhD in Physics at the University of Michigan. He completed his post-doctoral training in the department and joined the faculty in 2002. His primary research interests include FLASH radiation therapy using ultra high dose rates to reduce toxicity to healthy tissues, and the implementation and integration of new technologies. He was the first to implement and fully integrate a six degree of freedom patient-positioning system with daily image guidance for targeting moving tumors. He was an early pioneer in studying the effect of magnetic fields on radiation therapy beams, and of real-time electromagnetic tracking of moving tumors.

Issam El Naqa is the founding chair of the department of Machine Learning at Moffitt Cancer Center, Tampa, Fl. He is an internationally recognized authority in the fields of machine learning, data analytics, and oncology outcomes modeling and has published extensively in these areas with more than 220 + peer-reviewed journal publications and 4 edited textbooks. He has been a member and fellow of several academic and professional societies including AAPM and IEEE. His research has been funded by several federal and private grants in Canada and the USA and served on national and international study sections. He acts as a peer-reviewer and editorial board member for several leading international journals in his areas of expertise.

Wei Zhang received his Ph.D. degree in Biomedical Engineering from the Chinese Academy of Medical Sciences & Peking Union Medical College in 2018. He then began his postdoctoral training in the Department of Biomedical Engineering at the University of Michigan. He is currently a Research Assistant Professor in the Department of Biomedical Engineering at the University of Michigan. His current research focuses on the development of ionizing radiation acoustic imaging and photoacoustic imaging.

Xueding Wang is the Jonathan Rubin Collegiate Professor at the Department of Biomedical Engineering, University of Michigan, holding an adjunct Professor position at the Department of Radiology. His extensive experience spans imaging system development and the adaptation of novel diagnostic technologies to both laboratory research and clinical management. His research, supported by NIH, NSF, DoD, and other funding agencies, has yielded over 150 peer-reviewed publications. Dr. Wang has made significant contributions to the clinical applications of photoacoustic imaging in various medical fields including arthritis, cancer, and eye diseases. He has received numerous awards, including the Sontag Foundation Fellow of the Arthritis National Research Foundation in 2005 and the Distinguished Investigator Award of the Academy of Radiology Research in 2013. Dr. Wang also contributes to the academic community as an editorial board member of several scientific journals and as a steering committee member of the Journal of Lightwave Technology.

Jun Zou (S’98–M’02–SM’15) received the Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign in 2002. In 2004, he joined the Department of Electrical and Computer Engineering, Texas A&M University, where he is currently a Full Professor and directs the Micro Imaging and Sensing Devices and Systems (MISDS) Lab. His current research interests lie in the development ofmicro and nano opto-electro-mechanical devices and systems for biomedical imaging, robotics, and artificial intelligence applications. He is a Senior Member of the SPIE.