PBPetr Bruza

Petr Bruza

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Petr Bruza has a total of 337 co-authors. Here are the top co-authors by number of shared publications:

  • Brian W. Pogue

    Dartmouth College, Thayer School of Engineering, Hanover, New Hampshire, United States

    50 shared publications
  • David J. Gladstone

    Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire

    21 shared publications
  • Dalibor Panek

    Faculty of Biomedical Engineering, Czech Technical University in Prague, Nám. Sítná 3105, 27201 Kladno, Czech Republic

    18 shared publications
  • Brian W. Pogue

    Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire; Department of Medical Physics, University of Wisconsin-Madison, Madison, Wisconsin

    13 shared publications
  • Lesley A. Jarvis

    Department of Medicine Radiation Oncology Geisel School of Medicine Dartmouth College 03755 Hanover NH USA

    13 shared publications
  • Martinl Nikl

    Institute of Physics of the Czech Academy of Sciences, Cukrovarnicka 10/112, Prague, 16200, Czech Republic

    12 shared publications

Latest Publications

  • Open Access

    Time-of-flight fluorescence depth mapping using a spatiotemporal deep learning model

    by Shiru Wang, Arthur Pétusseau, Claudio Bruschini, Edoardo Charbon, Petr Bruza

    SignificanceFluorescence-guided surgery (FGS) utilizes molecular contrast agents to highlight critical structures or pathological tissues in real time. The premise of FGS is to enable precise surgical decision-making through accurate visualization and quantitative assessment of fluorophore distribution. However, strong effects of diffusion and absorption of fluorescent light in tissue confound fluorescence images, preventing accurate quantitative assessment of the concentration and distribution of fluorescent markers. These optical artifacts may lead to misinterpretation of tissue boundaries and compromised surgical precision, thereby diminishing the capabilities of FGS. Resolving topological depth maps of fluorophore distribution at the millimeter scale is an important first step in performing quantitative sub-surface fluorescence imaging.AimIn this study, we present a spatiotemporal deep learning architecture that utilizes picosecond single-photon avalanche diode (SPAD) sensor images to rapidly recover the depth topology of a fluorophore distribution embedded in diffuse media. The network is designed to work with wide-field, epi-illumination geometry and millimeter spatial resolution.ApproachA ConvLSTM-UNet deep learning network was developed for picosecond time-resolved image analysis. This network was trained on 5000 spatiotemporal maps simulated by the optical Monte Carlo method and convolved with the instrument response function (IRF) of the imaging system. The experimental setup utilized a SwissSPAD2 sensor synchronized with a 635 nm picosecond laser diode. Using only 10 selected temporal gates as input, the network could recover depth maps. Reconstruction accuracy was evaluated using mean error metrics across various depths and background concentrations of a fluorophore with a simulated decay time of 100 ps.ResultsA total of 75 different test fluorescence video data were evaluated. This set encompassed 15 unique inclusion shapes at five different depths. The network successfully reconstructed fluorescence topography up to 15 mm with a mean absolute error of less than 0.6 mm and mean depth variances below 0.5 mm. The inference time was ∼30 ms.ConclusionsIntegrating temporal and spatial deep learning networks enabled depth mapping from time-resolved fluorescence data. Utilizing real IRF proved the applicability of SPAD sensors for sub-surface fluorescence mapping.

  • Open Access

    Three‐dimensional Cherenkov emission surface mapping into the patient coordinate system for spatial delivery verification in EBRT

    by Alexander Geiersbach, Michael Jermyn, David Gladstone, Lesley Jarvis, Brian Pogue, Petr Bruza

    Background: Cherenkov imaging provides the ability to visualize beam delivery during radiation therapy treatments in real‐time informing the therapy team about possible deviations from the treatment plan. Current clinical practice with Cherenkov imaging relies on the real‐time view and post‐treatment qualitative assessment of Cherenkov images by therapists, physicists and clinicians who may then perform secondary treatment checks. Purpose: While Cherenkov images provide new insights into beam location, the limitation is that quantifying the observed deviation is difficult due to the lack of spatial registration of the Cherenkov emission in patient's coordinate space. This work addresses this problem with the first fusion of Cherenkov images with the three‐dimensional patient surface to enable quantitative spatial beam verification in patient coordinate space. By combining multiple camera viewpoints with CT‐sim derived surface coordinates, this system is capable of generating virtual beam's eye view images for objective measurement of beam delivery deviations on patient surfaces. This approach enables automated error detection thresholds and represents the first step toward real‐time, quantitative optical treatment verification and dosimetry. Methods: Cherenkov images from multiple geometrically calibrated cameras were projected onto Computed Tomography (CT) surface datasets to provide a Cherenkov surface image map on the patient. A raycasting algorithm was utilized to virtually map the Cherenkov surface image into the beam's eye view. To test the ability to quantitatively measure beam versus patient surface deviations from the virtual beam's eye view, a series of 10 × 10 cm 2 6X fields with simulated spatial shifts were delivered to both flat and anthropomorphic phantoms. Virtual beam's eye view images were generated for each shift, and the magnitudes of Cherenkov map shifts were measured with Iterative Closest Point (ICP) and Hausdorff distance metrics with respect to a reference field. Results: Iterative Closest Point algorithms achieved an average x and y accuracy of 0.5 ± 0.2 mm and 0.3 ± 0.2 mm for the flat phantom and 0.5 mm ± 0.7 mm and 0.3 mm ± 0.2 mm for the anthropomorphic torso phantom. Camera projection error adds up to an additional 0.9 mm of error. Conclusion: Multi‐view surface mapping of Cherenkov images onto the patient surface mesh enabled quantitative assessment of beam versus topology shifts between fractions, using the Cherenkov emission outline as a beam extent surrogate. We demonstrated the utility and metrics of Cherenkov 3D mapping towards future applications utilizing daily or real‐time surfaces obtained from, for example, cone‐beam CT or surface‐guided radiotherapy (SGRT) systems.

  • Open Access

    Emulating Clinical Workflow of Scintillator Array Dosimetry for FLASH Pencil-Beam Scanning Proton Therapy

    by Roman Vasyltsiv, Joseph Harms, Megan Clark, David J. Gladstone, Brian W. Pogue, Rongxiao Zhang, Petr Bruza


  • Improving Cherenkov-based radiotherapy dose estimates in diverse patient populations via skin luminance imaging

    by S M Decker, J M Andreozzi, D Hernandez, D A Alexander, V Wickramasinghe, R L Hachadorian, I M Oraiqat, E Chen, I R Washington, J Gui, et al.

    Objective. Cherenkov imaging is an emerging technology that detects light naturally emitted from patient tissue during radiation treatment. The initial intensity of Cherenkov light is proportional to radiation dose, but its absorption is highly dependent on patient skin pigmentation, with increasing melanin attenuating more Cherenkov photons per dose. This effect must be calibrated per patient before Cherenkov emission can serve as an accurate surrogate for dose. In this study, we present the first attempt at in vivo Cherenkov imaging of a diverse patient cohort and calibration for the effect of skin pigmentation towards quantitative Cherenkov dosimetry. Approach. A multi-institutional collaboration was designed to increase the diversity of our patient imaging cohort. Cherenkov imaging was completed with a time-gated, iCMOS camera, and color background images were taken with an RGB camera module under standardized lighting. Under an institutional review board-approved retrospective protocol, skin pigmentation was assessed per patient by calculating the relative luminance of their treated area from the color images. Additionally, 2D dose maps were generated by projecting the exponentially-weighted dose from the surface to 5 mm into the body, representative of Cherenkov emission. Main results. Of N 6MV = 23 and N 15MV = 20 breast patients imaged, encompassing a variety of skin pigmentations, Cherenkov intensity varied nearly 5X for the same dose delivered across the examined cohort. Plotting Cherenkov intensity per unit dose revealed a direct correlation with relative luminance, providing a linear calibration factor based on skin pigmentation. Including this calibration factor significantly improved Cherenkov-to-dose linearity, from R 2 = 0.79–0.96 for 6 MV and R 2 = 0.19–0.91 for 15 MV (p < 0.05). Significance. This study marks the first assessment of Cherenkov imaging in a diverse, representative patient population. It addresses an integral factor towards achieving quantitative in vivo Cherenkov dosimetry and demonstrates significant mitigation of the effect of skin pigmentation, while preserving the non-contact, real-time benefits of Cherenkov imaging.


  • Quantum-CMOS image sensor assessment for Cherenkov imaging during radiotherapy

    by Jeremy E. Hallett, Petr Bruza, Brian W. Pogue

    Cherenkov emission from human tissues during linac-delivered radiotherapy provides a detectable optical signal that shows the real-time dose delivery. The challenge of this modality is that the linac pulses producing Cherenkov light have short durations (2–4 μs) with low duty cycle (0.1%), and the signal reaches the camera with near single photon per pixel intensity. In this study, the ability to use an ultra-low noise (94%. Additional advantages include low susceptibility to stray x-ray radiation. This recent development of ultra-low noise qCMOS technology is a unique option for Cherenkov imaging, showing the ability to image signals well above the noise without need for image intensifiers or gain amplification.

  • Open Access

    Enhanced porphyrin-based hypoxia imaging by temporal oversampling of delayed fluorescence signal

    by Marien I. Ochoa, Arthur F. Petusseau, Matthew S. Reed, Petr Brůža, Brian W. Pogue

    Understanding tissue oxygen content, especially the lack of oxygen, is key for assessing proper tissue function. Of special interest are tumors, which often grow with chaotic neo-vasculature, subsequently creating both chronic and transient hypoxia conditions. The administration of 5-ALA1,2 as a precursor of protoporphyrin IX (PpIX) is widely adopted in neurosurgery to guide glioma resection,1,3 as well as in photodynamic therapy (PDT).4 – 6 Although nearly all PpIX imaging applications capture the prompt fluorescence (PF) component of PpIX, a more rarely explored property of PpIX is its weaker delayed fluorescence (DF) component.7 – 9 The DF component10 of PpIX arises from emissions originating from the first excited singlet state, which has been replenished through reverse intersystem crossing from the excited triplet state where oxygen is significantly quenched. Investigations of PpIX DF imaging as a marker of hypoxia have been reported for surgical guidance,8,11,12 PDT,7 and burn injuries13,14 where the DF component is isolated using time-gated cameras. From a biology standpoint, the use of techniques such as pressure-enhanced sensing of tissue oxygen (PRESTO)11,12 where tissue is palpated to create transient hypoxia and hence increase DF signal output has been proposed. However, the PRESTO effect can vary depending on the application; therefore, finding a complementary solution that can globally increase DF signal output is desired. Despite the utility of the technique to obtain real-time imaging of hypoxia, PpIX DF intensity is orders of magnitude lower than the standard PF, and this is why it has been largely ignored. Detection of this signal becomes a challenge as external light sources (e.g., room lights) can overwhelm the signal and make it inapplicable to open surgery settings where room lights are needed. Even when the white light is time-gated, aiding to see the FOV, achieving the minimum level of background light required for DF imaging remains a limitation. In this paper, the ability to increase the DF intensity and maximize signal gathering by variation of both excitation and detection time windows is explored. Of importance is the speed of acquisition to image fast oxygen kinetics in real time. A key driver in this work was the realization that recovery of the full-emission lifetime15 decay was likely not required, but rather, simply capturing the maximum signal intensity of the DF signal was the goal. This realization steered this work toward overdriving the frequency of capture, to maximize DF emission signal collection immediately after the excitation pulse. We achieved this using a faster frequency to overdrive the sampling rate as well as optimized pulse-width and gate-width settings. This technical advancement was explored here as a way to substantially increase the signal-to-background ratio of the images while keeping real-time frame rates.


  • First Retrospective QA of FAST-01 Clinical Treatment Fields Using High-Speed Quantitative Scintillation Imaging

    by Megan Clark, Zhiyan Xiao, Austin Sloop, Yongbin Zhang, Eunsin Lee, Roman Vasyltsiv, David Gladstone, Rongxiao Zhang, Petr Bruza, Anthony E. Mascia


  • Open Access

    Optimized scintillation imaging in low dose rate and bright room light conditions

    by Alexander P Niver, Petr Bruza, Brian W Pogue

    Objective. To develop a robust method for non-contact surface dosimetry during Total Body Irradiation (TBI) that uses an optimally paired choice of scintillator material with camera photocathode and can work insensitively to the normal ambient room lighting conditions (∼500 Lux). Approach. This goal was approached by assessing the emission contrast of scintillator signal to background room ratio (SBR) detected by the camera, in the challening conditions of low dose rate TBI with high room lights. A total of 9 fast-response scintillators, 3 wavelength shifters, and 2 camera photocathodes were systematically tested to determine the optimal combination. The effects of room lights on the scintillator signal and the background signal were assessed to avoid signal saturation while retaining accurate dose measurement. A bandpass wavelength filter was then applied to reduce the effects on room lights and scintillator signal. Main Results. One scintillator (EJ262) combined with a blue-green sensitive photocathode camera and a 500 nm band pass filter produced the greatest available scintillator SBR of 95 with maximal room lights on. The caveat is that this design rejects all patient Cherenkov light, which can be useful for visualizing the patient treatment. Another option which retained the Cherenkov signal but produced less available scintillator signal was found with another scintillator (EJ-260) and a red photocathode camera with SBR of 35, but a narrow bandpass filter is required to make it work in ambient room lights, which addition will also remove most of the Cherenkov signal. Significance. Non-contact scintillator imaging can be used for surface dosimetry in TBI with appropriate pairing of scintillator emission spectrum and camera photocathode sensitivity or optical filtering range.


  • Open Access

    Noise & mottle suppression methods for cumulative Cherenkov images of radiation therapy delivery

    by Jeremy E Hallett, Petr Bruza, Michael Jermyn, Ke Li, Brian W Pogue

    Purpose. Cherenkov imaging during radiotherapy provides a real time visualization of beam delivery on patient tissue, which can be used dynamically for incident detection or to review a summary of the delivered surface signal for treatment verification. Very few photons form the images, and one limitation is that the noise level per frame can be quite high, and mottle in the cumulative processed images can cause mild overall noise. This work focused on removing or suppressing noise via image postprocessing. Approach. Images were analyzed for peak-signal-to-noise and spatial frequencies present, and several established noise/mottle reduction algorithms were chosen based upon these observations. These included total variation minimization (TV-L1), non-local means filter (NLM), block-matching 3D (BM3D), alpha (adaptive) trimmed mean (ATM), and bilateral filtering. Each were applied to images acquired using a BeamSite camera (DoseOptics) imaged signal from 6x photons from a TrueBeam linac delivering dose at 600 MU min−1 incident on an anthropomorphic phantom and tissue slab phantom in various configurations and beam angles. The standard denoised images were tested for PSNR, noise power spectrum (NPS) and image sharpness. Results. The average peak-signal-to-noise ratio (PSNR) increase was 17.4% for TV-L1. NLM denoising increased the average PSNR by 19.1%, BM3D processing increased it by12.1% and the bilateral filter increased the average PSNR by 19.0%. Lastly, the ATM filter resulted in the lowest average PSNR increase of 10.9%. Of all of these, the NLM and bilateral filters produced improved edge sharpness with, generally, the lowest NPS curve. Conclusion. For cumulative image Cherenkov data, NLM and the bilateral filter yielded optimal denoising with the TV-L1 algorithm giving comparable results. Single video frame Cherenkov images exhibit much higher noise levels compared to cumulative images. Noise suppression algorithms for these frame rates will likely be a different processing pipeline involving these filters incorporated with machine learning.


  • Open Access

    Robust Real-time Segmentation of Bio-Morphological Features in Human Cherenkov Imaging during Radiotherapy via Deep Learning

    by Shiru Wang, Yao Chen, Lesley A. Jarvis, Yucheng Tang, David J. Gladstone, Kimberley S. Samkoe, Brian W. Pogue, Petr Bruza, Rongxiao Zhang

    Cherenkov imaging enables real-time visualization of megavoltage X-ray or electron beam delivery to the patient during Radiation Therapy (RT). Bio-morphological features, such as vasculature, seen in these images are patient-specific signatures that can be used for verification of positioning and motion management that are essential to precise RT treatment. However until now, no concerted analysis of this biological feature-based tracking was utilized because of the slow speed and accuracy of conventional image processing for feature segmentation. This study demonstrated the first deep learning framework for such an application, achieving video frame rate processing. To address the challenge of limited annotation of these features in Cherenkov images, a transfer learning strategy was applied. A fundus photography dataset including 20,529 patch retina images with ground-truth vessel annotation was used to pre-train a ResNet segmentation framework. Subsequently, a small Cherenkov dataset (1,483 images from 212 treatment fractions of 19 breast cancer patients) with known annotated vasculature masks was used to fine-tune the model for accurate segmentation prediction. This deep learning framework achieved consistent and rapid segmentation of Cherenkov-imaged bio-morphological features on another 19 patients, including subcutaneous veins, scars, and pigmented skin. Average segmentation by the model achieved Dice score of 0.85 and required less than 0.7 milliseconds processing time per instance. The model demonstrated outstanding consistency against input image variances and speed compared to conventional manual segmentation methods, laying the foundation for online segmentation in real-time monitoring in a prospective setting.