Affect Image Quality Research Articles

Fourier Neural Operator Network for Fast Photoacoustic Wave Simulations

Simulation tools for photoacoustic wave propagation have played a key role in advancing photoacoustic imaging by providing quantitative and qualitative insights into parameters affecting image quality. Classical methods for numerically solving the photoacoustic wave equation rely on a fine discretization of space and can become computationally expensive for large computational grids. In this work, we applied Fourier Neural Operator (FNO) networks as a fast data-driven deep learning method for solving the 2D photoacoustic wave equation in a homogeneous medium. Comparisons between the FNO network and pseudo-spectral time domain approach were made for the forward and adjoint simulations. Results demonstrate that the FNO network generated comparable simulations with small errors and was orders of magnitude faster than the pseudo-spectral time domain methods (~26× faster on a 64 × 64 computational grid and ~15× faster on a 128 × 128 computational grid). Moreover, the FNO network was generalizable to the unseen out-of-domain test set with a root-mean-square error of 9.5 × 10−3 in Shepp–Logan, 1.5 × 10−2 in synthetic vasculature, 1.1 × 10−2 in tumor and 1.9 × 10−2 in Mason-M phantoms on a 64 × 64 computational grid and a root mean squared of 6.9 ± 5.5 × 10−3 in the AWA2 dataset on a 128 × 128 computational grid.

Visible–Infrared Dual-Sensor Fusion for Single-Object Tracking

Visual object tracking is a classical yet rapidly evolving task in computer vision. Traditional RGB trackers in the visible spectrum suffer from poor lighting conditions and bad weather, which significantly affect image quality. To tackle this drawback, RGB trackers are extended from single modality (RGB) to dual modalities [RGB-thermal (RGB-T)] via visible–infrared dual-sensor fusion. Compared with visible cameras, thermal infrared cameras are insensitive to lightning conditions and have a strong ability to penetrate haze and smog. Inspired by the complementary merits of visible and thermal information, we propose a visible–infrared (RGB-T) tracker that fuses the visible modality and infrared modality on the pixel level, feature level, and decision level. First, for the pixel-level fusion, a fused histogram is derived from a four-channel RGB-T image that concatenates an aligned RGB and thermal image pair. With the fused histogram, a foreground mask is derived for guiding correlation filter learning by segmenting the foreground target from the neighbor background. Second, for the feature-level fusion, handcrafted features, such as histogram of gradient (HOG) and color name (CN), are extracted from the RGB and thermal images, respectively, and then concatenated along the channel dimension for feature fusion. Third, for the decision-level fusion, the channel weight of each feature channel in both modalities is estimated for weighting per-channel filter response in localization. To demonstrate the effectiveness of our visible and infrared fusion strategy, we perform an ablation study to evaluate each level of fusion independently on the RGBT234 dataset. We also perform extensive experiments on the RGBT234 and VTUAV_ST datasets to evaluate the performance of the proposed RGB-T tracker compared with state-of-the-art trackers. It is worth noting that our three-level visible–infrared fusion framework is generic and flexible and our implementation is far from optimal. This framework is open-source and easy to be combined with any correlation filter-based RGB-T tracker for further performance gain.

Assessment of the knowledge level of radiographers and CT technologists regarding computed tomography parameters in Iran

ObjectiveTo evaluate the knowledge of CT technologists and radiographers concerning CT parameters, in order to keep CT users be aware of all CT parameters and their impact on image quality and patient dose while keeping their knowledge up to date. MethodsThis cross-sectional descriptive-analytic study was conducted in Zanjan province, Iran, using a valid and reliable online questionnaire with 53 questions distributed to 153 radiology personnel. Each questionnaire included questions about demographic characteristics and CT parameters, including kVp, mAs, image noise, slice thickness, pitch and automatic tube current modulation (ATCM), and the effects of these parameters on image quality and radiation dose. ResultsA response rate of 61.0% (94/153) was achieved. This study enrolled 30 radiographers and 64 CT technologists. There was no significant difference in the mean scores of the two groups (52.16 ± 9.90 vs. 56.09 ± 12.00, t ​= ​−1.545, P ​= ​0.126). CT personnel had the highest and lowest knowledge scores in image reconstruction (t ​= ​−28.33, P ​= ​0.001) and mAs (t ​= ​4.27, P ​= ​0.590), respectively, whereas radiographers had the highest and lowest knowledge scores in the fields of image noise (t ​= ​9.85, P ​= ​0.015) and kVp (t ​= ​−12.73, P ​= ​0.007), respectively. ConclusionsThe total knowledge scores of CT technologists and radiographers for various scan parameters affecting image quality and dose were insufficient. Thus, it is recommended that regular retraining courses be held to enhance and update technologists’ knowledge.

Rethinking low-cost microscopy workflow: Image enhancement using deep based Extended Depth of Field methods

Microscopic techniques in low-to-middle income countries are constrained by the lack of adequate equipment and trained operators. Since light microscopy delivers crucial methods for the diagnosis and screening of numerous diseases, several efforts have been made by the scientific community to develop low-cost devices such as 3D-printed portable microscopes. Nevertheless, these devices present some drawbacks that directly affect image quality: the capture of the samples is done via mobile phones; more affordable lenses are usually used, leading to poorer physical properties and images with lower depth of field; misalignments in the microscopic set-up regarding optical, mechanical, and illumination components are frequent, causing image distortions such as chromatic aberrations. This work investigates several pre-processing methods to tackle the presented issues and proposed a new workflow for low-cost microscopy. Additionally, two new deep learning models based on Convolutional Neural Networks are also proposed (EDoF-CNN-Fast and EDoF-CNN-Pairwise) to generate Extended Depth of Field (EDoF) images, and compared against state-of-the-art approaches. The models were tested using two different datasets of cytology microscopic images: public Cervix93 and a new dataset that has been made publicly available containing images captured with μSmartScope. Experimental results demonstrate that the proposed workflow can achieve state-of-the-art performance when generating EDoF images from low-cost microscopes.

Image dataset for benchmarking automated fish detection and classification algorithms

Multiparametric video-cabled marine observatories are becoming strategic to monitor remotely and in real-time the marine ecosystem. Those platforms can achieve continuous, high-frequency and long-lasting image data sets that require automation in order to extract biological time series. The OBSEA, located at 4 km from Vilanova i la Geltrú at 20 m depth, was used to produce coastal fish time series continuously over the 24-h during 2013–2014. The image content of the photos was extracted via tagging, resulting in 69917 fish tags of 30 taxa identified. We also provided a meteorological and oceanographic dataset filtered by a quality control procedure to define real-world conditions affecting image quality. The tagged fish dataset can be of great importance to develop Artificial Intelligence routines for the automated identification and classification of fishes in extensive time-lapse image sets.

An Investigation of Signal Preprocessing for Photoacoustic Tomography.

Photoacoustic tomography (PAT) is increasingly being used for high-resolution biological imaging at depth. Signal-to-noise ratios and resolution are the main factors that determine image quality. Various reconstruction algorithms have been proposed and applied to reduce noise and enhance resolution, but the efficacy of signal preprocessing methods which also affect image quality, are seldom discussed. We, therefore, compared common preprocessing techniques, namely bandpass filters, wavelet denoising, empirical mode decomposition, and singular value decomposition. Each was compared with and without accounting for sensor directivity. The denoising performance was evaluated with the contrast-to-noise ratio (CNR), and the resolution was calculated as the full width at half maximum (FWHM) in both the lateral and axial directions. In the phantom experiment, counting in directivity was found to significantly reduce noise, outperforming other methods. Irrespective of directivity, the best performing methods for denoising were bandpass, unfiltered, SVD, wavelet, and EMD, in that order. Only bandpass filtering consistently yielded improvements. Significant improvements in the lateral resolution were observed using directivity in two out of three acquisitions. This study investigated the advantages and disadvantages of different preprocessing methods and may help to determine better practices in PAT reconstruction.

Predicting the image quality of respiratory-gated and breath-hold 3D MRCP from the breathing curve: a prospective study.

To compare the image quality of breath-hold magnetic resonance cholangiopancreatography (BH-MRCP) and respiratory-gating MRCP (RG-MRCP), and to explore breathing curve-based factors and patient-related data affecting image quality. A total of 126 participants who underwent RG-MRCP and BH-MRCP on a 3-T magnetic resonance (MR) scanner were enrolled from May to December 2021. The images were evaluated by three radiologists on a 5-point scale. Respiratory parameters were extracted from the breathing curves. The Wilcoxon test was used to compare the image quality between the two MRCPs. Logistic regression analyzes were performed to identify age, sex, abdominal pain, and breathing predictor variables of better image quality. BH-MRCP performed better in visualizing intrahepatic bile ducts and overall image quality than RG-MRCP (p < 0.01). Factors predicting relatively good image quality included lower standard deviation of the respiratory amplitude (SDamp)-minimum-peak (odds ratio = 0.16, p < 0.01) for RG-MRCP and lower SDamp (OR = 0.69, p < 0.01) for BH-MRCP. BH-MRCP had significantly better overall image quality than RG-MRCP. Respiratory conditions exerted a significant impact on MRCP image quality, and parameters derived from the breathing curve could help predict the image quality of both sequences. • Both breath-hold (BH) and respiratory-gating (RG) MRCP demonstrate satisfying image quality. • BH-GRASE-MRCP is significantly better than RG-MRCP at the group level, but not for every individual. • Respiratory conditions exert a significant impact on the image quality, and the breathing curve can help predict the image quality.

New challenges and opportunities for low-field MRI

In this manuscript we deal with recent advances in low-field Magnetic Resonance Imaging (MRI). The development of low-cost MRI solutions allowing portability and trustable diagnosis is a hot topic worldwide by these days. We analyze basic technical issues of recent examples of fixed-field instruments operating at low-field. Then we discuss pros and cons of the pre-polarized approach, from both physical and technical perspectives. Permanent magnet and electromagnet technology are confronted. Finally, magnetic field-cycling is introduced as an alternative MRI technique, where field-dependent experiments can be exploded for the development of new contrast mechanisms that are not feasible for fixed-field MRI instruments. As field cycled machines usually deals with switched currents in electromagnets, magnetic field instability and inhomogeneity are the main limiting factors affecting image quality. We finalize this manuscript discussing how it turns possible to overcome these limitations, thus opening new possibilities for the development of cost effective MRI technology.

Optimisasi dengan Audit Dosis Radiasi pada Pemeriksaan Thorak Proyeksi AP Pasien Dewasa

Optimization is one of the radiation protection efforts by giving radiation doses to patients as low as possible without reducing the quality of clinical images or what is often referred to as ALARA (As Low as Reasonably Achievable). A Diagnostic Reference Level (DRL) is an investigative indicator to optimize radiation protection for diagnostic and interventional radiology patients. The typical radiation dose value in AP thoracic examination patients at Gading Pluit Hospital is higher than the Indonesian DRL (I-DRL) value, so it is necessary to optimize the dose. Optimization of radiation dose aims to protect patients by evaluating exposure factors without compromising the quality of clinical images. AP chest image acquisition was performed using clinical exposure factors and their variations on the general radiographic X-Ray Listem REX 525-R. Evaluation of dose and image quality was carried out by a reading doctor/radiologist using a blind test. Optimizing results show that a lower exposure factor can reduce the radiation dose in AP thoracic examination of adult patients without affecting image quality. Exposure factor 80 kVp; 150 mA; 3.75 mAs is recommended to replace the exposure factor used before optimization, namely 80 kVp; 200 mA; 7 mAs. Keywords: optimization, DRL, radiation protection.

Iron-based coupling media for MRI-guided ultrasound surgery.

In this study, we examine the effects of a recently developed, iron-based coupling medium (IBCM) on guidance magnetic resonance (MR) scans during transcranial, magnetic-resonance-guided, focused ultrasound surgery (tMRgFUS) procedures. More specifically, this study tests the hypotheses that the use of the IBCM will (a) not adversely affect image quality, (b) remove aliasing from small field-of-view scans, and (c) decouple image quality from the motion state of the coupling fluid. An IBCM, whose chemical synthesis and characterization are reported elsewhere, was used as a coupling medium during tMRgFUS procedures on gel phantoms. Guidance magnetization-prepared rapid-gradient-echo (MP-RAGE), TSE, and GRE scans were acquired with fields of view of 28 and 18cm. Experiments were repeated with the IBCM in several distinct flow states. GRE scans were used to estimate temperature time courses as a gel target was insonated. IBCM performance was measured by computing (i) the root mean square difference (RMSD) of TSE and GRE pixel values acquired using water and the IBCM, relative to the use of water; (ii) through-time temperature uncertainty for GRE scans; and (iii) Bland-Altman analysis of the temperature time courses. Finally, guidance TSE and GRE scans of a human volunteer were acquired during a separate sham tMRgFUS procedure. As a control, all experiments were repeated using a water coupling medium. Use of the IBCM reduced RMSD in TSE scans by a factor of 4 or more for all fields of view and nonstationary motion states, but did not reduce RMSD estimates in MP-RAGE scans. With the coupling media in a stationary state, the IBCM altered estimates of temperature uncertainty relative to the use of water by less than 0.2°C. However, with a high flow state, the IBCM reduced temperature uncertainties by the statistically significant amounts (at the 0.01 level) of 0.5°C (28cm field of view) and 5°C (18cm field of view). Bland-Altman analyses found a 0.1°C ± 0.5°C difference between temperature estimates acquired using water and the IBCM as coupling media. Finally, scans of a human volunteer using the IBCM indicate more conspicuous grey/white matter contrast, a reduction in aliasing, and a less than 0.2°C change in temperature uncertainty. The use of an IBCM during tMRgFUS procedures does not adversely affect image quality for TSE and GRE scans, can decouple image quality from the motion state of the coupling fluid, and can remove aliasing from scans where the field of view is set to be much smaller than the spatial extent of the coupling fluid.

Is the postoperative pedicle screw position after dorsal instrumentation with or without intraoperative cone beam CT imaging worse in patients with obesity than in normal-weight patients?

BackgroundIntraoperative cone beam CT (CBCT) imaging in dorsal instrumentation facilitates pedicle screw positioning. However, in patients with obesity, the benefit may be reduced due to artifacts that affect image quality. The purpose of this study was to evaluate whether intraoperative CBCT leads to an improved postoperative screw position compared to conventional fluoroscopy independent of body weight.MethodsA total of 71 patients (18 patients with a BMI > 30 kg/m2, 53 patients with a BMI < 30 kg/m2) who underwent dorsal instrumentation with intraoperative CBCT imaging were included in study groups one (SG1) and two (SG2). Two control groups (CG1 and CG2) were randomly sampled to include 22 patients with a BMI > 30 kg/m2 and 60 patients with a BMI < 30 kg/m2 who underwent dorsal instrumentation without intraoperative CBCT imaging. The pedicle screw position in postoperative computed tomography was assessed using the Gertzbein–Robbins classification.ResultsIn SG1 (BMI > 30 kg/m2), a total of 107 (83.6%) pedicle screws showed no relevant perforation (type A + B), and 21 (16.4%) pedicle screws showed relevant perforation (type C − E). In SG2 (BMI < 30 kg/m2), 328 (90.9%) screws were classified as type A + B, and 33 (9.1%) screws were classified as type C − E. In CG1 (BMI > 30 kg/m2), 102 (76.1%) pedicle screws showed no relevant perforation (type A + B), and 32 (23.9%) pedicle screws showed relevant perforation (type C − E). In CG2 (BMI < 30 kg/m2), 279 (76.9%) screws were classified as type A + B, and 84 (23.1%) screws were classified as type C − E. There were significant differences between the values of SG1 and SG2 (p = 0.03) and between the values of SG2 and CG2 (p < 0.0001).ConclusionCBCT imaging in dorsal instrumentation can lead to an improved pedicle screw position among both patients with obesity and normal-weight patients. However, patients with obesity showed significantly worse pedicle screw positions postoperatively after dorsal instrumentation with intraoperative CBCT imaging than normal-weight patients.

Low-Frequency Data Embedding for DFT-Based Image Steganography

As sharing digital media is getting more prominent, there has been a rise in the development of techniques for protecting digital media against misuse. However, the amount of such techniques has been less prominent for printed goods. The data hiding method presented in this paper is suited for the print domain. The method uses the Discrete Fourier Transform to embed data into images and Gray Component Replacement to maintain high image quality by masking embedding artifacts. This research examines whether the method's performance can be optimized using low-frequency ranges for embedding. The aim is to evaluate the usage of low-frequency ranges in order to achieve higher robustness. Different frequency ranges are tested to determine how they affect image quality and detection rates. Tests are conducted in the digital domain. The results show that the method can maintain image quality regardless of the frequency range, and that low-frequency ranges lead to more consistent detection rates. Future research will be done on images in the print domain.

Optimum Parameters for Imaging and Dose Calculation in Lung SBRT

<h3>Purpose/Objective(s)</h3> Efficiency, set up accuracy, patient comfort and accurate dose calculations are critical requirements in radiation therapy especially in stereotactic body radiotherapy (SBRT). There are several CT imaging parameters such as kV, mAs, image reconstruction algorithm filters, field of view and pitch that affect image quality used for volume delineation and dose calculation. Modifying the imaging parameters may lead to variation in Hounsfield Units (HU) that may affect the dose calculations. We investigated the relationship between the image reconstruction algorithms (Body (B) and Lung filter (L)), HU and dose distribution. We hypothesize that variation in the imaging parameters may lead to better image quality with clinically insignificant dose to target and organs at risk (OARs). <h3>Materials/Methods</h3> Ten lung SBRT patients were imaged on a 16 slice CT scanner. The standard clinical scan (C) is a B- reconstruction filter-based scan with low pitch. For selected patients, in additional to standard scan, sharp scans were acquired using greater pitch and L-filter algorithm to see the effect of density and dose calculation. Clinical workflow includes fusing the images acquired with B and L filter. Clinically B filter images are used for planning and treatment. For this study purposes, an additional plan was generated on L filtered images using appropriate HU to density curve. CT density phantom was scanned on a 16 slice CT scanner with standard Body (B) filter as well as standard images of CT density phantom were reconstructed with lung (L) filter algorithm to generate HU to physical density table for a treatment planning system and used for dose calculation. Image qualities for the two filter were evaluated visually by an attending radiation oncologist. The dosimetric differences in the target coverages were evaluated using the HU curve generated for B and L filter. <h3>Results</h3> For clinical cases, using lung SBRT protocols, planning target volume (PTV) coverage V100% were recorded. In addition to the target coverage mean physical density was recorded for the B and L filter and C-clinical scans. Average difference in the PTV physical density was 2.4%, 0.3% and 2.1% for B-C, L-C and B-L, where B stands for – B Filter density curve on the sharp scan, L – L Filter density curve on the sharp scan, C – Clinical scan. No significant differences were noted in PTV coverages for B-C 1.7% (p < 0.0004), L-C 2.03% (p < 0.0007) and B-L -0.29% (p < 0.0005) based plans for V100%. <h3>Conclusion</h3> The L filter high pitch combination improved image contrast and thus helped in contouring. No significant difference in the PTV coverage was observed when B filter HU curve was applied to sharp quality scan used for dose calculation. Thus, images acquired with L filter can be used clinically with existing B filter HU curve. Overall Differences in dosimetric parameters are insignificant compared to the small uncertainties in HU numbers.

Performance Evaluation of Indian Food Image Classification system using Transfer Learning with MobileNetV3

Food image classification and recognition is an emerging research area due to its growing importance in the medical and health industries. As India is growing digitally rapidly, an automated Indian food image classification system will help in the development of diet tracking, calorie estimation, and many other health and food consumption-related applications. In recent years many deep learning techniques evolved. Deep learning is a robust and low-cost method for extracting information from food images, though, challenges lie in extracting information from real-world food images due to various factors affecting image quality such as photos from different angles and positions, several objects appearing in the photo, etc. We used Non-Standard dataset for Indian Food images of 13 different classes consist more than 3500 images. In this paper, we use CNN as our base model to build our system, which gives an accuracy of 47% of the system. After that, we deployed the transfer learning technique with MobileNetV3 for improvement in accuracy, which resulted in an improvement in accuracy of up to 90%.

RETRACTED ARTICLE: Improvement of seismic data quality and recognition of fault discontinuities through seismic data conditioning applications: a case study of Issaran oil field, Gulf of Suez, Egypt

Obtaining high-quality seismic imaging in shallow heterogeneous carbonate reservoirs with complicated structural regimes, such as the Issaran field, is difficult. Issaran field is a heavy oil shallow heterogeneous fractured carbonate reservoirs located in the Gulf of Suez of Egypt. It has many geological factors that affect image quality and pose numerous challenges. In addition, the seismic data was acquired more than 12 years ago, with narrow azimuth and short offsets. As a result, the fault zones are not sharply defined. Furthermore, the seismic data was processed about 10 years ago. The signal-to-noise ratio is relatively poor due to the random and coherent noises. A robust data conditioning workflow for noise suppression and fault discontinuity sharpening is used to improve the post-stack seismic data quality. In this context, a steering dataset was generated, and a dip-steered median filter (DSMF), a dip-steered diffusion filter, and a fault enhancement filter (FEF) were applied to sharpen the discontinuities. Structural attributes were evaluated, to investigate how the newly applied data conditioning affects the clarity of fault patterns. Furthermore, multiple physical attributes were extracted, including instantaneous phase, instantaneous frequency, and RMS amplitude to better understand the reservoir stratigraphic heterogeneity. The application of DSMF is useful in removing the residual random noises. The FEF-similarity attribute revealed small-scale faults with a 50-foot vertical throw. The physical attributes proved that the Issaran carbonate facies is controlled by structure. Moreover, the RMS amplitude attribute helped in distinguishing between porous and non-porous dolomite facies. As imaging quality has significantly improved, the applied seismic data conditioning workflow is beneficial for the field development in Issaran field. It is also suggested that this data conditioning workflow be applied in other heterogeneous carbonate reservoirs with complex structures around the world.

Advanced pediatric neuroimaging.

Advanced magnetic resonance neuroimaging techniques play an important adjunct role to conventional MRI sequences for better depiction and characterization of a variety of brain disorders. In this article we briefly review the basic principles and clinical utility of a select number of these techniques, including clinical functional MRI for presurgical planning, clinical diffusion tensor imaging and related techniques, dynamic susceptibility contrast perfusion imaging using gadolinium injection, and arterial spin labeling perfusion imaging. The article focuses on general principles of clinical MRI acquisition protocols, relevant factors affecting image quality, and a general framework for obtaining images for each of these techniques. We also present relevant advances for acquiring these types of imaging sequences in a clinical setting.

Evaluations of the performances of PET and MRI in a simultaneous PET/MRI instrument for pre-clinical imaging

BackgroundPET/MRI is an attractive imaging modality due to the complementary nature of MRI and PET. Obtaining high quality small animal PET/MRI results is key for the translation of novel PET/MRI agents and techniques to the radiology clinic. To obtain high quality imaging results, a hybrid PET/MRI system requires additional considerations beyond the standard issues with separate PET and MRI systems. In particular, researchers must understand how their PET system affects the MR acquisitions and vice versa. Depending on the application, some of these effects may substantially influence image quality. Therefore, the goal of this report is to provide guidance, recommendations, and practical experiments for implementing and using a small animal PET/MRI instrument.ResultsVarious PET and MR image quality parameters were tested with their respective modality alone and in the presence of both systems to determine how the combination of PET/MRI affects image quality. Corrections and calibrations were developed for many of these effects. While not all image characteristics were affected, some characteristics such as PET quantification, PET SNR, PET spatial resolution, PET partial volume effects, and MRI SNR were altered by the presence of both systems.ConclusionsA full exploration of a new PET/MRI system before performing small animal PET/MRI studies is beneficial and necessary to ensure that the new instrument can produce highly accurate and precise PET/MR images.

Deep learning for fast photoacoustic wave simulations

Photoacoustic tomography involves absorption of pulsed light and subsequent generation of ultrasound, which when detected using an array of sensors can produce clinically useful images. Simulation tools for photoacoustic wave propagation have played a key role in advancing photoacoustic imaging by providing quantitative and qualitative insights into parameters affecting image quality. Classical methods for numerically solving the photoacoustic wave equation rely on a fine discretization of space and can become computationally expensive for large computational grids. In this work, we apply Fourier Neural Operator (FNO) networks as a fast data-driven deep learning method for solving the 2D photoacoustic wave equation in a homogeneous medium. Comparisons between the FNO network and pseudo-spectral time domain approach demonstrated that the FNO network generated comparable simulations with small errors and was an order of magnitude faster. Moreover, the FNO network was generalizable and could generate simulations not observed in the training data.

A deep neural networks-based image reconstruction algorithm for a reduced sensor model in large-scale tomography system

Image reconstruction for soft-field tomography is a highly nonlinear and ill-posed inverse problem. Owing to the highly complicated nature of soft-field, the reconstructed images are always poor in quality. One of the factors that affect image quality is the number of sensors in a tomography system. It is commonly assumed that increasing the number of sensors in a tomography system will improve the ill-posed condition in image reconstruction and hence improve image quality. However, as the number of sensors increases, challenges such as more complicated and expensive hardware, slower data acquisition rates, longer image reconstruction times, and larger sensitivity matrices will arise, resulting in a greater ill-posed condition. Since deep learning (DL) is capable of expressing complex nonlinear functions, the majority of research efforts have been directed toward developing a robust DL-based inverse solver for image reconstruction. However, no study has been conducted to solve the inverse problem and improve the quality of the reconstructed image using a reduced sensor model for a large-scale tomography system. This paper proposed an image reconstruction algorithm based on Deep Neural Networks (DNN) to investigate its feasibility in solving the ill-posed inverse problem caused by the reduced sensor model for a large-scale tomography system. The proposed DNN model is based on a supervised, feed-forward, fully connected, backpropagation network. It comprises an input layer, three hidden layers and an output layer. Also, it was trained using large data samples obtained from COMSOL simulation. The relationship between the scattered electromagnetic field measurement and the corresponding true electromagnetic field distribution vector is determined. During the image reconstruction process, the untrained scattered electromagnetic field measurement samples are used as inputs to the trained DNN model, and the model output is an estimate of the electromagnetic field distribution. The results show that the proposed DNN can accurately describe the distribution of electromagnetic field and boundary shape of phantom compared to traditional algorithms (LBP, FBP, Noser and Tikhonov), regardless of the size and number of phantoms within the monitoring area. Hence, the proposed DNN is more robust and has a high degree of generalization.

Implementation of a comprehensive set of optimised CBCT protocols and validation through imaging quality and dose audit.

Cone-beam computed tomography (CBCT) for radiotherapy treatment verification has increased in frequency; therefore, it is crucial to optimise image quality and radiation dose to patients. The aim of this study was to implement optimised CBCT protocols for the Varian TrueBeams for most tumour sites in adult patients. A combination of patient size-specific CBCT protocols from the literature and developed in-house was used. Scans taken before and after optimisation were compared by senior radiographers and physicists to evaluate how changes affected image quality and clinical usability for online image registration. The change in dose for each new CBCT protocol was compared to the Varian default. A clinical audit was performed following implementation to evaluate the changes in imaging dose for all patients receiving a CBCT during that period. Ten CBCT protocols were introduced including head and neck and patient-size-specific thorax and pelvis/abdomen protocols. Scans from 102 patients with images before and after optimisation were assessed, none of the scans showed image quality changes compromising clinical usability and for some image quality was improved. Between November 2020 and June 2021, 1185 patients had CBCTs using the new protocols. The imaging dose was reduced for 52% of patients, remained the same for 37% and increased for 12%. This study showed that substantial dose reductions and image quality improvements can be achieved with simple changes in the default settings of the Varian TrueBeam CBCT without affecting the radiographers' confidence in online image registration. This study represents a comprehensive assessment and optimisation of CBCT protocols for most sites, validated on a large cohort of patients.