Investigation of an implantable dosimeter for single-point water equivalent path length verification in proton therapy

Patient-specific deep learning for 3D protoacoustic image reconstruction and dose verification in proton therapy

BACKGROUND

Protoacoustic (PA) imaging has the potential to provide real-time 3D dose verification of proton therapy. However, PA images are susceptible to severe distortion due to limited angle acquisition. Our previous studies showed the potential of using deep learning to enhance PA images. As the model was trained using a limited number of patients' data, its efficacy was limited when applied to individual patients.

PURPOSE

In this study, we developed a patient-specific deep learning method for protoacoustic imaging to improve the reconstruction quality of protoacoustic imaging and the accuracy of dose verification for individual patients.

METHODS

Our method consists of two stages: in the first stage, a group model is trained from a diverse training set containing all patients, where a novel deep learning network is employed to directly reconstruct the initial pressure maps from the radiofrequency (RF) signals; in the second stage, we apply transfer learning on the pre-trained group model using patient-specific dataset derived from a novel data augmentation method to tune it into a patient-specific model. Raw PA signals were simulated based on computed tomography (CT) images and the pressure map derived from the planned dose. The reconstructed PA images were evaluated against the ground truth by using the root mean squared errors (RMSE), structural similarity index measure (SSIM) and gamma index on 10 specific prostate cancer patients. The significance level was evaluated by t-test with the p-value threshold of 0.05 compared with the results from the group model.

RESULTS

The patient-specific model achieved an average RMSE of 0.014 (p < 0.05 ${{{p}}}<{0.05}$ ), and an average SSIM of 0.981 (p < 0.05 ${{{p}}}<{0.05}$ ), out-performing the group model. Qualitative results also demonstrated that our patient-specific approach acquired better imaging quality with more details reconstructed when comparing with the group model. Dose verification achieved an average RMSE of 0.011 (p < 0.05 ${{{p}}}<{0.05}$ ), and an average SSIM of 0.995 (p < 0.05 ${{{p}}}<{0.05}$ ). Gamma index evaluation demonstrated a high agreement (97.4% [p < 0.05 ${{{p}}}<{0.05}$ ] and 97.9% [p < 0.05 ${{{p}}}<{0.05}$ ] for 1%/3  and 1%/5 mm) between the predicted and the ground truth dose maps. Our approach approximately took 6 s to reconstruct PA images for each patient, demonstrating its feasibility for online 3D dose verification for prostate proton therapy.

CONCLUSIONS

Our method demonstrated the feasibility of achieving 3D high-precision PA-based dose verification using patient-specific deep-learning approaches, which can potentially be used to guide the treatment to mitigate the impact of range uncertainty and improve the precision. Further studies are needed to validate the clinical impact of the technique.

Hybrid-supervised deep learning for domain transfer 3D protoacoustic image reconstruction

Objective. Protoacoustic imaging showed great promise in providing real-time 3D dose verification of proton therapy. However, the limited acquisition angle in protoacoustic imaging induces severe artifacts, which impairs its accuracy for dose verification. In this study, we developed a hybrid-supervised deep learning method for protoacoustic imaging to address the limited view issue.Approach. We proposed a Recon-Enhance two-stage deep learning method. In the Recon-stage, a transformer-based network was developed to reconstruct initial pressure maps from raw acoustic signals. The network is trained in a hybrid-supervised approach, where it is first trained using supervision by the iteratively reconstructed pressure map and then fine-tuned using transfer learning and self-supervision based on the data fidelity constraint. In the enhance-stage, a 3D U-net is applied to further enhance the image quality with supervision from the ground truth pressure map. The final protoacoustic images are then converted to dose for proton verification.Main results. The results evaluated on a dataset of 126 prostate cancer patients achieved an average root mean squared errors (RMSE) of 0.0292, and an average structural similarity index measure (SSIM) of 0.9618, out-performing related start-of-the-art methods. Qualitative results also demonstrated that our approach addressed the limit-view issue with more details reconstructed. Dose verification achieved an average RMSE of 0.018, and an average SSIM of 0.9891. Gamma index evaluation demonstrated a high agreement (94.7% and 95.7% for 1%/3 mm and 1%/5 mm) between the predicted and the ground truth dose maps. Notably, the processing time was reduced to 6 s, demonstrating its feasibility for online 3D dose verification for prostate proton therapy.Significance. Our study achieved start-of-the-art performance in the challenging task of direct reconstruction from radiofrequency signals, demonstrating the great promise of PA imaging as a highly efficient and accurate tool forinvivo3D proton dose verification to minimize the range uncertainties of proton therapy to improve its precision and outcomes.

3Din vivodose verification in prostate proton therapy with deep learning-based proton-acoustic imaging

Dose delivery uncertainty is a major concern in proton therapy, adversely affecting the treatment precision and outcome. Recently, a promising technique, proton-acoustic (PA) imaging, has been developed to provide real-timein vivo3D dose verification. However, its dosimetry accuracy is limited due to the limited-angle view of the ultrasound transducer. In this study, we developed a deep learning-based method to address the limited-view issue in the PA reconstruction. A deep cascaded convolutional neural network (DC-CNN) was proposed to reconstruct 3D high-quality radiation-induced pressures using PA signals detected by a matrix array, and then derive precise 3D dosimetry from pressures for dose verification in proton therapy. To validate its performance, we collected 81 prostate cancer patients' proton therapy treatment plans. Dose was calculated using the commercial software RayStation and was normalized to the maximum dose. The PA simulation was performed using the open-source k-wave package. A matrix ultrasound array with 64 × 64 sensors and 500 kHz central frequency was simulated near the perineum to acquire radiofrequency (RF) signals during dose delivery. For realistic acoustic simulations, tissue heterogeneity and attenuation were considered, and Gaussian white noise was added to the acquired RF signals. The proposed DC-CNN was trained on 204 samples from 69 patients and tested on 26 samples from 12 other patients. Predicted 3D pressures and dose maps were compared against the ground truth qualitatively and quantitatively using root-mean-squared-error (RMSE), gamma-index (GI), and dice coefficient of isodose lines. Results demonstrated that the proposed method considerably improved the limited-view PA image quality, reconstructing pressures with clear and accurate structures and deriving doses with a high agreement with the ground truth. Quantitatively, the pressure accuracy achieved an RMSE of 0.061, and the dose accuracy achieved an RMSE of 0.044, GI (3%/3 mm) of 93.71%, and 90%-isodose line dice of 0.922. The proposed method demonstrates the feasibility of achieving high-quality quantitative 3D dosimetry in PA imaging using a matrix array, which potentially enables the online 3D dose verification for prostate proton therapy.

On-line range verification for proton beam therapy using spherical ionoacoustic waves with resonant frequency

In contrast to conventional X-ray therapy, proton beam therapy (PBT) can confine radiation doses to tumours because of the presence of the Bragg peak. However, the precision of the treatment is currently limited by the uncertainty in the beam range. Recently, a unique range verification methodology has been proposed based on simulation studies that exploit spherical ionoacoustic waves with resonant frequency (SPIREs). SPIREs are emitted from spherical gold markers in tumours initially introduced for accurate patient positioning when the proton beam is injected. These waves have a remarkable property: their amplitude is linearly correlated with the residual beam range at the marker position. Here, we present proof-of-principle experiments using short-pulsed proton beams at the clinical dose to demonstrate the feasibility of using SPIREs for beam-range verification with submillimetre accuracy. These results should substantially contribute to reducing the range uncertainty in future PBT applications.

A novel range-verification method using ionoacoustic wave generated from spherical gold markers for particle-beam therapy: a simulation study

This study proposes a novel alternative range-verification method for proton beam with acoustic waves generated from spherical metal markers. When proton beam is incident on metal markers, most of the resulting pressure waves are confined in the markers because of the large difference in acoustic impedance between the metal and tissue. However, acoustic waves with frequency equal to marker’s resonant frequency escape this confinement; the marker briefly acts as an acoustic transmitter. Herein, this phenomenon is exploited to measure the range of the proton beam. We test the proposed strategy in 3-D simulations, combining the dose calculations with modelling of acoustic-wave propagation. A spherical gold marker of 2.0 mm diameter was placed in water with a 60 MeV proton beam incident on it. We investigated the dependence of pressure waves on the width of beam pulse and marker position. At short beam pulse, specific high-frequency acoustic waves of 1.62 MHz originating from the marker were observed in wave simulations, whose amplitude correlated with the distance between the marker and Bragg peak. Results indicate that the Bragg peak position can be estimated by measuring the acoustic wave amplitudes from the marker, using a single detector properly designed for the resonance frequency.

In vivo range verification in particle therapy

Exploitation of the full potential offered by ion beams in clinical practice is still hampered by several sources of treatment uncertainties, particularly related to the limitations of our ability to locate the position of the Bragg peak in the tumor. To this end, several efforts are ongoing to improve the characterization of patient position, anatomy, and tissue stopping power properties prior to treatment as well as to enable in vivo verification of the actual dose delivery, or at least beam range, during or shortly after treatment. This contribution critically reviews methods under development or clinical testing for verification of ion therapy, based on pretreatment range and tissue probing as well as the detection of secondary emissions or physiological changes during and after treatment, trying to disentangle approaches of general applicability from those more specific to certain anatomical locations. Moreover, it discusses future directions, which could benefit from an integration of multiple modalities or address novel exploitation of the measurable signals for biologically adapted therapy.

Time-resolved diode dosimetry calibration through Monte Carlo modeling for in vivo passive scattered proton therapy range verification

Purpose

Our group previously introduced an in vivo proton range verification methodology in which a silicon diode array system is used to correlate the dose rate profile per range modulation wheel cycle of the detector signal to the water‐equivalent path length (WEPL) for passively scattered proton beam delivery. The implementation of this system requires a set of calibration data to establish a beam‐specific response to WEPL fit for the selected ‘scout’ beam (a 1 cm overshoot of the predicted detector depth with a dose of 4 cGy) in water‐equivalent plastic. This necessitates a separate set of measurements for every ‘scout’ beam that may be appropriate to the clinical case. The current study demonstrates the use of Monte Carlo simulations for calibration of the time‐resolved diode dosimetry technique.

Methods

Measurements for three ‘scout’ beams were compared against simulated detector response with Monte Carlo methods using the Tool for Particle Simulation (TOPAS). The ‘scout’ beams were then applied in the simulation environment to simulated water‐equivalent plastic, a CT of water‐equivalent plastic, and a patient CT data set to assess uncertainty.

Results

Simulated detector response in water‐equivalent plastic was validated against measurements for ‘scout’ spread out Bragg peaks of range 10 cm, 15 cm, and 21 cm (168 MeV, 177 MeV, and 210 MeV) to within 3.4 mm for all beams, and to within 1 mm in the region where the detector is expected to lie.

Conclusion

Feasibility has been shown for performing the calibration of the detector response for three ‘scout’ beams through simulation for the time‐resolved diode dosimetry technique in passive scattered proton delivery.

Nuclear physics in particle therapy: a review

Charged particle therapy has been largely driven and influenced by nuclear physics. The increase in energy deposition density along the ion path in the body allows reducing the dose to normal tissues during radiotherapy compared to photons. Clinical results of particle therapy support the physical rationale for this treatment, but the method remains controversial because of the high cost and of the lack of comparative clinical trials proving the benefit compared to x-rays. Research in applied nuclear physics, including nuclear interactions, dosimetry, image guidance, range verification, novel accelerators and beam delivery technologies, can significantly improve the clinical outcome in particle therapy. Measurements of fragmentation cross-sections, including those for the production of positron-emitting fragments, and attenuation curves are needed for tuning Monte Carlo codes, whose use in clinical environments is rapidly increasing thanks to fast calculation methods. Existing cross sections and codes are indeed not very accurate in the energy and target regions of interest for particle therapy. These measurements are especially urgent for new ions to be used in therapy, such as helium. Furthermore, nuclear physics hardware developments are frequently finding applications in ion therapy due to similar requirements concerning sensors and real-time data processing. In this review we will briefly describe the physics bases, and concentrate on the open issues.

Dose ratio proton radiography using the proximal side of the Bragg peak

PURPOSE

In recent years, there has been a movement toward single-detector proton radiography, due to its potential ease of implementation within the clinical environment. One such single-detector technique is the dose ratio method in which the dose maps from two pristine Bragg peaks are recorded beyond the patient. To date, this has only been investigated on the distal side of the lower energy Bragg peak, due to the sharp falloff. The authors investigate the limits and applicability of the dose ratio method on the proximal side of the lower energy Bragg peak, which has the potential to allow a much wider range of water-equivalent thicknesses (WET) to be imaged. Comparisons are made with the use of the distal side of the Bragg peak.

METHODS

Using the analytical approximation for the Bragg peak, the authors generated theoretical dose ratio curves for a range of energy pairs, and then determined how an uncertainty in the dose ratio would translate to a spread in the WET estimate. By defining this spread as the accuracy one could achieve in the WET estimate, the authors were able to generate lookup graphs of the range on the proximal side of the Bragg peak that one could reliably use. These were dependent on the energy pair, noise level in the dose ratio image and the required accuracy in the WET. Using these lookup graphs, the authors investigated the applicability of the technique for a range of patient treatment sites. The authors validated the theoretical approach with experimental measurements using a complementary metal oxide semiconductor active pixel sensor (CMOS APS), by imaging a small sapphire sphere in a high energy proton beam.

RESULTS

Provided the noise level in the dose ratio image was 1% or less, a larger spread of WETs could be imaged using the proximal side of the Bragg peak (max 5.31 cm) compared to the distal side (max 2.42 cm). In simulation, it was found that, for a pediatric brain, it is possible to use the technique to image a region with a square field equivalent size of 7.6 cm(2), for a required accuracy in the WET of 3 mm and a 1% noise level in the dose ratio image. The technique showed limited applicability for other patient sites. The CMOS APS demonstrated a good accuracy, with a root-mean-square-error of 1.6 mm WET. The noise in the measured images was found to be σ = 1.2% (standard deviation) and theoretical predictions with a 1.96σ noise level showed good agreement with the measured errors.

CONCLUSIONS

After validating the theoretical approach with measurements, the authors have shown that the use of the proximal side of the Bragg peak when performing dose ratio imaging is feasible, and allows for a wider dynamic range than when using the distal side. The dynamic range available increases as the demand on the accuracy of the WET decreases. The technique can only be applied to clinical sites with small maximum WETs such as for pediatric brains.

Evolution of technology to optimize the delivery of proton therapy: the third generation

The evolution of proton therapy technology will lead to a new generation of systems that allow for greater accuracy and precision of the dose delivery and will be more compact. We envision that over the next 10-15 years, the quality of deliverable proton dose distributions in the patient will be pushed nearly toward the physical limit of proton therapy. Those future proton therapy systems will fit into treatment rooms of similar size as today's conventional radiation treatment rooms. At the same time, due to technological advancements, the cost of proton therapy will come down to the cost of advanced photon therapy. We discuss some of the technologies that will put these speculative improvements within reach.

In vivo proton range verification: a review

Protons are an interesting modality for radiotherapy because of their well defined range and favourable depth dose characteristics. On the other hand, these same characteristics lead to added uncertainties in their delivery. This is particularly the case at the distal end of proton dose distributions, where the dose gradient can be extremely steep. In practice however, this gradient is rarely used to spare critical normal tissues due to such worries about its exact position in the patient. Reasons for this uncertainty are inaccuracies and non-uniqueness of the calibration from CT Hounsfield units to proton stopping powers, imaging artefacts (e.g. due to metal implants) and anatomical changes of the patient during treatment. In order to improve the precision of proton therapy therefore, it would be extremely desirable to verify proton range in vivo, either prior to, during, or after therapy. In this review, we describe and compare state-of-the art in vivo proton range verification methods currently being proposed, developed or clinically implemented.