Table of contents

Presumably, intensity-modulated proton radiotherapy (IMPT) is the most powerful form of proton radiotherapy. In the current state of the art, IMPT beam configurations (i.e. the number of beams and their directions) are, in general, chosen subjectively based on prior experience and practicality. Beam configuration optimization (BCO) for IMPT could, in theory, significantly enhance IMPT's therapeutic potential. However, BCO is complex and highly computer resource-intensive. Some algorithms for BCO have been developed for intensity-modulated photon therapy (IMRT). They are rarely used clinically mainly because the large number of beams typically employed in IMRT renders BCO essentially unnecessary. Moreover, in the newer form of IMRT, volumetric modulated arc therapy, there are no individual static beams. BCO is of greater importance for IMPT because it typically employs a very small number of beams (2-4) and, when the number of beams is small, BCO is critical for improving plan quality. However, the unique properties and requirements of protons, particularly in IMPT, make BCO challenging. Protons are more sensitive than photons to anatomic changes, exhibit variable relative biological effectiveness along their paths, and, as recently discovered, may spare the immune system. Such factors must be considered in IMPT BCO, though doing so would make BCO more resource intensive and make it more challenging to extend BCO algorithms developed for IMRT to IMPT. A limited amount of research in IMPT BCO has been conducted; however, considerable additional work is needed for its further development to make it truly effective and computationally practical. This article aims to provide a review of existing BCO algorithms, most of which were developed for IMRT, and addresses important requirements specific to BCO for IMPT optimization that necessitate the modification of existing approaches or the development of new effective and efficient ones.

Objective. Monolithic scintillator crystals coupled to silicon photomultiplier (SiPM) arrays are promising detectors for PET applications, offering spatial resolution around 1 mm and depth-of-interaction information. However, their timing resolution has always been inferior to that of pixellated crystals, while the best results on spatial resolution have been obtained with algorithms that cannot operate in real-time in a PET detector. In this study, we explore the capabilities of monolithic crystals with respect to spatial and timing resolution, presenting new algorithms that overcome the mentioned problems. Approach. Our algorithms were tested first using a simulation framework, then on experimentally acquired data. We tested an event timestamping algorithm based on neural networks which was then integrated into a second neural network for simultaneous estimation of the event position and timestamp. Both algorithms are implemented in a low-cost field-programmable gate array that can be integrated in the detector and can process more than 1 million events per second in real-time. Results. Testing the neural network for the simultaneous estimation of the event position and timestamp on experimental data we obtain 0.78 2D FWHM on the (x, y) plane, 1.2 depth-of-interaction FWHM and 156 coincidence time resolution on a $25\,\mathrm{mm}\,{\rm{\times }}\,25\,\mathrm{mm}\,{\rm{\times }}\,8\,\mathrm{mm}\,{\rm{\times }}$ LYSO monolith read-out by 64 $3\,\mathrm{mm}{\rm{\times }}3\,\mathrm{mm}$ Hamamatsu SiPMs. Significance. Our results show that monolithic crystals combined with artificial intelligence can rival pixellated crystals performance for time-of-flight PET applications, while having better spatial resolution and DOI resolution. Thanks to the use of very light neural networks, event characterization can be done on-line directly in the detector, solving the issues of scalability and computational complexity that up to now were preventing the use of monolithic crystals in clinical PET scanners.

Objective. Arterial dispersion ultrasound vibrometry (ADUV) relies on the use of guided waves in arterial geometries for shear wave elastography measurements. Both the generation of waves through the use of acoustic radiation force (ARF) and the techniques employed to infer the speed of the resulting wave motion affect the spectral content and accuracy of the measurement. In particular, the effects of the shape and location of the ARF beam in ADUV have not been widely studied. In this work, we investigated how such variations of the ARF beam affect the induced motion and the measurements in the dispersive modes that are excited. Approach. The study includes an experimental evaluation on an arterial phantom and an in vivo validation of the observed trends, observing the two walls of the waveguide, simultaneously, when subjected to variations in the ARF beam extension (F/N) and focus location. Main results. Relying on the theory of guided waves in cylindrical shells, the shape of the beam controls the selection and nature of the induced modes, while the location affects the measured dispersion curves (i.e. variation of phase velocity with frequency or wavenumber, multiple modes) across the waveguide walls. Significance. This investigation is important to understand the spectral content variations in ADUV measurements and to maximize inversion accuracy by tuning the ARF beam settings in clinical applications.

Objective: Cervical cancer is one of the two biggest killers of women and early detection of cervical precancerous lesions can effectively improve the survival rate of patients. Manual diagnosis by combining colposcopic images and clinical examination results is the main clinical diagnosis method at present. Developing an intelligent diagnosis algorithm based on artificial intelligence is an inevitable trend to solve the objectification of diagnosis and improve the quality and efficiency of diagnosis. Approach: A colposcopic multimodal fusion convolutional neural network (CMF-CNN) was proposed for the classification of cervical lesions. Mask region convolutional neural network was used to detect the cervical region while the encoding network EfficientNet-B3 was introduced to extract the multimodal image features from the acetic image and iodine image. Finally, Squeeze-and-Excitation, Atrous Spatial Pyramid Pooling, and convolution block were also adopted to encode and fuse the patient's clinical text information. Main results: The experimental results showed that in 7106 cases of colposcopy, the accuracy, macro F1-score, macro-areas under the curve of the proposed model were 92.70%, 92.74%, 98.56%, respectively. They are superior to the mainstream unimodal image classification models. Significance: CMF-CNN proposed in this paper combines multimodal information, which has high performance in the classification of cervical lesions in colposcopy, so it can provide comprehensive diagnostic aid.

Objective: The diversity in technical configuration between clinically available radiosurgery systems, results in accordingly diverse treatment times for the same physical dose prescription, spanning from several min to more than 1 h. This, combined with evidence supporting the impact of dose delivery temporal pattern on the bio-effectiveness of low-LET radiation treatments, challenges the 'acute exposure' assumption adopted clinically to estimate the biological outcome of a given treatment scheme under the concept of biologically effective dose (BED). Approach: In this work, the treatment plans of 30 patients underwent CyberKnife radiosurgery for vestibular schwannoma (VS), prescribing a marginal dose of 13 Gy to the tumor, were retrospectively reviewed and the corresponding dose distributions were resolved in the temporal domain. For this purpose, the dose delivery timeline for each treatment was calculated based on relevant treatment plan data and technical specifications of the CyberKnife system, while dosimetry data were independently acquired on a CT-based digital model of each patient using an in-house developed dose calculation algorithm. Main results: Results showed that CyberKnife delivers highly inhomogeneous dose rate distributions in the temporo-spatial domain. This influences the delivered BED levels due to alterations in the sublethal damage repair (SLR) occurring within the treatment session. Using a BED framework involving SLR effects, it was shown that each physical dose iso-surface is associated with a BED_slr range. For the patient cohort studied, a typical range of 2%, with respect to the mean BED_slr value was found at 1σ. Significance: The marginal BED_slr delivered to the tumor by the prescription dose iso-surface deteriorates with treatment time, involving both beam-on time and beam-off gaps. For treatment time, T, between 21 and 50 min, this can be expressed by ${{BED}}_{{slr}}({{Gy}}_{2.47})=\left(-0.35\pm 2.8 \% \right)\bullet T\left(\min \right)+(76.74\pm 0.4 \% ).$ Compared to the acute exposure approach, a BED 'loss' of 21% is associated with the delivery of 13 Gy to the VS-tumor in 35 min.

We have developed a new type of detector array for monitoring of radiation beams in radiotherapy. The detector has parallel-plane architecture with multiple large-area uniform thin-film electrodes. At least one of the electrodes is resistive and has multiple signal readouts spread out along its perimeter. The integral dose deposited in the detector gives rise to multiple signals that depend on the distribution of radiation with respect to resistive electrode array (REA) geometry. The purpose of the present study was to experimentally determine basic detector response to MLC collimated x-ray fields. Two detector arrays have been characterized: circular and rectangular. The current and electrostatic potential distribution within the resistive electrode are governed by the Laplace and continuity equations with boundary conditions at the border with the readouts. Measurements for pencil beams showed that signal strength depends primarily on the distances between the location of the pencil beam and the readouts. Measurements for larger irregular MLC showed that signals as a function of time are quasi-linear with respect to MLC position and are proportional to the MLC area. Derivation of clinically relevant radiation beam parameters from REA signals, such as MLC position, MLC gap size and monitor unit per MLC segment relies on the detector response model with empirical model parameters. An approximate analytical detector response model was proposed and used to fit experiment data.

Immobilization masks are used to prevent patient movement during head and neck (H&N) radiotherapy. Motion restriction is beneficial both during treatment, as well as in the pre-treatment simulation phase, where magnetic resonance imaging (MRI) is often used for target definition. However, the shape and size of the immobilization masks hinder the use of regular, close-fitting MRI receive arrays. In this work, we developed a mask-compatible 8-channel H&N array that consists of a single-channel baseplate, on which the mask can be secured, and a flexible 7-channel anterior element that follows the shape of the mask. The latter uses high impedance coils to achieve its flexibility and radiolucency. A fully-functional prototype was manufactured, its radiolucency was characterized, and the gain in imaging performance with respect to current clinical setups was quantified. Dosimetry measurements showed an overall dose change of −0.3%. Small, local deviations were up to −2.7% but had no clinically significant impact on a full treatment plan, as gamma pass rates (3%/3 mm) only slightly reduced from 97.9% to 97.6% (clinical acceptance criterion: ≥95%). The proposed H&N array improved the imaging performance with respect to three clinical setups. The H&N array more than doubled (+123%) and tripled (+246%) the signal-to-noise ratio with respect to the clinical MRI-simulation and MR-linac setups, respectively. G-factors were also lower with the proposed H&N array. The improved imaging performance resulted in a clearly visible signal-to-noise ratio improvement of clinically used TSE and DWI acquisitions. In conclusion, the 8-channel H&N array improves the imaging performance of MRI-simulation and MR-linac acquisitions, while dosimetry suggests that no clinically significant dose changes are induced.

Identifying tumour infiltration zones during tumour resection in order to excise as much tumour tissue as possible without damaging healthy brain tissue is still a major challenge in neurosurgery. The detection of tumour infiltrated regions so far requires histological analysis of biopsies taken from at expected tumour boundaries. The gold standard for histological analysis is the staining of thin cut specimen and the evaluation by a neuropathologist. This work presents a way to transfer the histological evaluation of a neuropathologist onto optical coherence tomography (OCT) images. OCT is a method suitable for real time in vivo imaging during neurosurgery however the images require processing for the tumour detection. The method demonstrated here enables the creation of a dataset which will be used for supervised learning in order to provide a better visualization of tumour infiltrated areas for the neurosurgeon. The created dataset contains labelled OCT images from two different OCT-systems (wavelength of 930 nm and 1300 nm). OCT images corresponding to the stained histological images were determined by shaping the sample, a controlled cutting process and a rigid transformation process between the OCT volumes based on their topological information. The histological labels were transferred onto the corresponding OCT images through a non-rigid transformation based on shape context features retrieved from the sample outline in the histological image and the OCT image. The accuracy of the registration was determined to be 200 ± 120 μm. The resulting dataset consists of 1248 labelled OCT images for each of the two OCT systems.

Objective. To develop and test the feasibility of a novel Single ProjectIon DrivEn Real-time Multi-contrast (SPIDERM) MR imaging technique that can generate real-time 3D images on-the-fly with flexible contrast weightings and a low latency. Approach. In SPIDERM, a 'prep' scan is first performed, with sparse k-space sampling periodically interleaved with the central k-space line (navigator data), to learn a subject-specific model, incorporating a spatial subspace and a linear transformation between navigator data and subspace coordinates. A 'live' scan is then performed by repeatedly acquiring the central k-space line only to dynamically determine subspace coordinates. With the 'prep'-learned subspace and 'live' coordinates, real-time 3D images are generated on-the-fly with computationally efficient matrix multiplication. When implemented based on a multi-contrast pulse sequence, SPIDERM further allows for data-driven image contrast regeneration to convert real-time contrast-varying images into contrast-frozen images at user's discretion while maintaining motion states. Both digital phantom and in-vivo experiments were performed to evaluate the technical feasibility of SPIDERM. Main results. The elapsed time from the input of the central k-space line to the generation of real-time contrast-frozen 3D images was approximately 45 ms, permitting a latency of 55 ms or less. Motion displacement measured from SPIDERM and reference images showed excellent correlation (${R}^{2}\geqslant 0.983$). Geometric variation from the ground truth in the digital phantom was acceptable as demonstrated by pancreas contour analysis (Dice ≥ 0.84, mean surface distance ≤ 0.95 mm). Quantitative image quality metrics showed good consistency between reference images and contrast-varying SPIDREM images in in-vivo studies (mean ${\rm{NMRSE}}=0.141,{\rm{PSNR}}=30.12,{\rm{SSIM}}=0.88$). Significance. SPIDERM is capable of generating real-time multi-contrast 3D images with a low latency. An imaging framework based on SPIDERM has the potential to serve as a standalone package for MR-guided radiation therapy by offering adaptive simulation through a 'prep' scan and real-time image guidance through a 'live' scan.

Objective. Dual energy computed tomography (DECT) has been shown to provide additional image information compared to conventional CT and has been used in clinical routine for several years. The objective of this work is to present a DECT implementation for a Small Animal Radiation Research Platform (SARRP) and to verify it with a quantitative analysis of a material phantom and a qualitative analysis with an ex-vivo mouse measurement. Approach. For dual energy imaging, two different spectra are required, but commercial small animal irradiators are usually not optimized for DECT. We present a method that enables dual energy imaging on a SARRP with sequential scanning and an Empirical Dual Energy Calibration (EDEC). EDEC does not require the exact knowledge of spectra and attenuation coefficients; instead, it is based on a calibration. Due to the SARRP geometry and reconstruction algorithm, the calibration is done using an artificial CT image based on measured values. The calibration yields coefficients to convert the measured images into material decomposed images. Main results. To analyze the method quantitatively, the electron density and the effective atomic number of a material phantom were calculated and compared with theoretical values. The electron density showed a maximum deviation from the theoretical values of less than 5% and the atomic number of slightly more than 6%. For use in mice, DECT is particularly useful in distinguishing iodine contrast agent from bone. A material decomposition of an ex-vivo mouse with iodine contrast agent was material decomposed to show that bone and iodine can be distinguished and iodine-corrected images can be calculated. Significance. DECT is capable of calculating electron density images and effective atomic number images, which are appropriate parameters for quantitative analysis. Furthermore, virtual monochromatic images can be obtained for a better differentiation of materials, especially bone and iodine contrast agent.

Objective. Time-of-flight-positron emission tomography would highly benefit from a coincidence time resolution (CTR) below 100 ps: improvement in image quality and patient workflow, and reduction of delivered dose are among them. This achievement proved to be quite challenging, and many approaches have been proposed and are being investigated for this scope. One of the most recent consists in combining different materials with complementary properties (e.g. high stopping power for 511 keV $\gamma$-ray and fast timing) in a so-called heterostructure, metascintillator or metapixel. By exploiting a mechanism of energy sharing between the two materials, it is possible to obtain a fraction of fast events which significantly improves the overall time resolution of the system. Approach. In this work, we present the progress on this innovative technology. After a simulation study using the Geant4 toolkit, aimed at understanding the optimal configuration in terms of energy sharing, we assembled four heterostructures with alternating plates of BGO and EJ232 plastic scintillator. We fabricated heterostructures of two different sizes (3 × 3 × 3 mm³ and  3 × 3 × 15 mm³), each made up of plates with two different thicknesses of plastic plates. We compared the timing of these pixels with a standard bulk BGO crystal and a structure made of only BGO plates (layered BGO). Main results. CTR values of 239 ± 12 ps and 197 ± 10 ps FWHM were obtained for the 15 mm long heterostructures with 100 µm and 200 µm thick EJ232 plates (both with 100 µm thick BGO plates), compared to 271 ± 14 ps and 303 ± 15 ps CTR for bulk and layered BGO, respectively. Significance. Significant improvements in timing compared to standard bulk BGO were obtained for all the configurations tested. Moreover, for the long pixels, depth of interaction (DOI) collimated measurements were also performed, allowing to validate a simple model describing light transport inside the heterostructure.

Glioblastoma (GBM) is a severe malignant brain tumor with bad prognosis, and overall survival (OS) time prediction is of great clinical value for customized treatment. Recently, many deep learning (DL) based methods have been proposed, and most of them build deep networks to directly map pre-operative images of patients to the OS time. However, such end-to-end prediction is sensitive to data inconsistency and noise. In this paper, inspired by the fact that clinicians usually evaluate patient prognosis according to previously encountered similar cases, we propose a novel multimodal deep KNN based OS time prediction method. Specifically, instead of the end-to-end prediction, for each input patient, our method first search its K nearest patients with known OS time in a learned metric space, and the final OS time of the input patient is jointly determined by the K nearest patients, which is robust to data inconsistency and noise. Moreover, to take advantage of multiple imaging modalities, a new inter-modality loss is introduced to encourage learning complementary features from different modalities. The in-house single-center dataset containing multimodal MR brain images of 78 GBM patients is used to evaluate our method. In addition, to demonstrate that our method is not limited to GBM, a public multi-center dataset (BRATS2019) containing 211 patients with low and high grade gliomas is also used in our experiment. As benefiting from the deep KNN and the inter-modality loss, our method outperforms all methods under evaluation in both datasets. To the best of our knowledge, this is the first work, which predicts the OS time of GBM patients in the strategy of KNN under the DL framework.

Purpose. Real-time three-dimensional (3D) magnetic resonance (MR) imaging is challenging because of slow MR signal acquisition, leading to highly under-sampled k-space data. Here, we proposed a deep learning-based, k-space-driven deformable registration network (KS-RegNet) for real-time 3D MR imaging. By incorporating prior information, KS-RegNet performs a deformable image registration between a fully-sampled prior image and on-board images acquired from highly-under-sampled k-space data, to generate high-quality on-board images for real-time motion tracking. Methods. KS-RegNet is an end-to-end, unsupervised network consisting of an input data generation block, a subsequent U-Net core block, and following operations to compute data fidelity and regularization losses. The input data involved a fully-sampled, complex-valued prior image, and the k-space data of an on-board, real-time MR image (MRI). From the k-space data, under-sampled real-time MRI was reconstructed by the data generation block to input into the U-Net core. In addition, to train the U-Net core to learn the under-sampling artifacts, the k-space data of the prior image was intentionally under-sampled using the same readout trajectory as the real-time MRI, and reconstructed to serve an additional input. The U-Net core predicted a deformation vector field that deforms the prior MRI to on-board real-time MRI. To avoid adverse effects of quantifying image similarity on the artifacts-ridden images, the data fidelity loss of deformation was evaluated directly in k-space. Results. Compared with Elastix and other deep learning network architectures, KS-RegNet demonstrated better and more stable performance. The average (±s.d.) DICE coefficients of KS-RegNet on a cardiac dataset for the 5- , 9- , and 13-spoke k-space acquisitions were 0.884 ± 0.025, 0.889 ± 0.024, and 0.894 ± 0.022, respectively; and the corresponding average (±s.d.) center-of-mass errors (COMEs) were 1.21 ± 1.09, 1.29 ± 1.22, and 1.01 ± 0.86 mm, respectively. KS-RegNet also provided the best performance on an abdominal dataset. Conclusion. KS-RegNet allows real-time MRI generation with sub-second latency. It enables potential real-time MR-guided soft tissue tracking, tumor localization, and radiotherapy plan adaptation.

Objective. In charged-particle therapy, a ripple filter (RiFi) is used for broadening the Bragg peak in the beam direction. A conventional RiFi consists of plates with a fine ridge and groove structure. The construction of the RiFi has been a time-consuming and costly task. In this study, we developed a simple RiFi made of multi-layered metal mesh (mRiFi), with which the Bragg peak is broadened due to structural randomness, similar to what occurs for the already proposed RiFi with porous material. Approach. The mRiFi was constructed by stacking commercially available metal meshes at random positions and angles. The mRiFi was inexpensive to fabricate due to its high availability and low machining accuracy. The Bragg peak width modulated by the mRiFi can be uniquely determined by the wire material, wire diameter, wire-to-wire spacing of the metal mesh, and the number of mesh sheets. We fabricated four mRiFis consisting of 10, 20, 30, and 40 layers of stainless steel meshes with a wire diameter of 0.1 mm and a wire-to-wire spacing of 0.508 mm. Main results. Using the mRiFis consisting of 10, 20, 30, and 40 mesh sheets, we succeeded in broadening the Bragg peak following the normal distribution with the respective standard deviation σ values of 0.83, 1.15, 1.41, and 1.56 mm in water in experimental planar-integrated depth dose measurements with 140.3 MeV u⁻¹ carbon-ion beams. The effect of range broadening with the mRiFi was independent of its lateral position, and the measurement of the surface dose using radiochromic films showed no severe inhomogeneity with a homogeneity index greater than 0.3 caused by the mRiFis. Significance. The developed mRiFi can be used as a RiFi in charged-particle therapy. The mRiFi has three advantages: high supply stability of the material for manufacturing it, easy fabrication, and low cost.