Physical and Engineering Sciences in Medicine最新文献

Electron beam dosimetry is sensitive to the surface contour of the patient. Over 10% difference between Treatment Planning System (TPS) and independent monitor-unit (IMU) calculations have been reported in the literature. Similar results were observed in our clinic between Radformation ClearCalc IMU and Eclipse TPS electron Monte Carlo (eMC) algorithm (v.16.1). This paper presents data measured under 3D printed spherical and cylindrical phantoms to validate the eMC algorithm in the presence of curved geometries. Measurements were performed with multiple detectors and compared to calculations made in Eclipse for the 6, 9 and 12 MeV electron energies. This data is used to create curvature correction factors (CCFs), defined as the ratio of the detector reading with the curved-surface phantom to a flat phantom at the same depth. The mean difference between the TPS calculated and measured CCFs using the NACP, Diode E, microSilicon, and microDiamond detectors were 1.3, 0.9, 0.7 and 0.7% respectively, with maximum differences of 4.5, 2.3, 1.9, and 1.8% respectively. Applying CCFs to previous failing patient IMU calculations improved agreement to the TPS. CCFs were implemented in our clinic for patient-specific IMU calculations with the assistance of a ESAPI script.

The purpose of this work was to determine percentage depth dose (PDD) curves for kilovoltage x-rays from the WOmed-T105 unit, with open-ended steel applicators and beam qualities ranging from 0.5 to 4.2 mm Al. Measurements were made with parallel plate chambers in a water phantom, with extrapolation based on a fifth order polynomial used to estimate the surface dose. Measurements were also made with parallel plate chambers in a plastic water phantom, with thin plastic sheets used to obtain detailed measurements at shallow depths (less than 1 mm). Monte Carlo simulations were performed using the EGSnrc package, with two different sources as input: a SpekPy simulation of the x-ray beam and a full simulation of the x-ray tube, treatment head and applicators. Results showed that all four methods (two measurements and two simulations) agreed within the measurement uncertainty at depths greater than 2 mm. At shallow depths, significant differences were noted. At depths less than 0.1 mm, the full Monte Carlo simulation and the solid water measurements showed a sharp spike in surface dose which is attributed to electron contamination, which was not seen in the SpekPy Monte Carlo simulation or the extrapolated water measurements. At depths between 0.1 mm and 2 mm, beyond the range of contaminant electrons, the extrapolated water measurements underestimate the dose by up to 13% compared to the full Monte Carlo simulation and the solid water measurements, attributed to fluorescent photons generated in the applicators. This work demonstrates that for open-ended applicators, measurement of depth doses in water with extrapolation of surface dose has the potential to significantly underestimate the dose at shallow depths between the surface and 2 mm, even after eliminating electron contamination from the beam.

The cervical vertebral maturation (CVM) method is essential to determine the timing of orthodontic and orthopedic treatment. In this paper, a target detection model called DC-YOLOv5 is proposed to achieve fully automated detection and staging of CVM. A total of 1800 cephalometric radiographs were labeled and categorized based on the CVM stages. We introduced a model named DC-YOLOv5, optimized for the specific characteristics of CVM based on YOLOv5. This optimization includes replacing the original bounding box regression loss calculation method with Wise-IOU to address the issue of mutual interference between vertical and horizontal losses in Complete-IOU (CIOU), which made model convergence challenging. We incorporated the Res-dcn-head module structure to enhance the focus on small target features, improving the model's sensitivity to subtle sample differences. Additionally, we introduced the Convolutional Block Attention Module (CBAM) dual-channel attention mechanism to enhance focus and understanding of critical features, thereby enhancing the accuracy and efficiency of target detection. Loss functions, precision, recall, mean average precision (mAP), and F1 scores were used as the main algorithm evaluation metrics to assess the performance of these models. Furthermore, we attempted to analyze regions important for model predictions using gradient Class Activation Mapping (CAM) techniques. The final F1 scores of the DC-YOLOv5 model for CVM identification were 0.993, 0.994 for mAp0.5 and 0.943 for mAp0.5:0.95, with faster convergence, more accurate and more robust detection than the other four models. The DC-YOLOv5 algorithm shows high accuracy and robustness in CVM identification, which provides strong support for fast and accurate CVM identification and has a positive effect on the development of medical field and clinical diagnosis.

Performance testing of gamma cameras and single photon computed tomography/computed tomography (SPECT/CT) systems is not subject to regulatory requirements across states and territories in Australia. Internationally recognised testing standards from organisations such as the National Electrical Manufacturers Association (NEMA) describe methodologies for recommended tests. However, variations exist in suggested quality control (QC) schedules from professional bodies such as the Australia and New Zealand Society of Nuclear Medicine (ANZSNM). In this study, a survey was conducted to benchmark current QC programs across a selected sample of eight standalone and networked Australian public hospitals. Vendor-specific flood-field uniformity (intrinsic or extrinsic/system) verification without photomultiplier (PMT) tuning and CT QC were performed at all sites. Weekly and monthly PMT tuning followed by intrinsic flood-field verifications were performed at most sites. At least half of the sites performed monthly centre of rotation (COR) offset verifications. SPECT/CT alignment calibrations and verifications were undertaken by service engineers at all sites, and periodic verifications were performed by local staff at varying frequencies. Variations were observed for other periodic QC tests such as spatial resolution and planar sensitivity. Similarly, variations were observed for tests specific to whole-body systems and SPECT systems. Most sites checked daily and periodic QC results against pass/fail criteria set by vendors. Additional analyses of the QC results, including trend analysis and periodic reviews, were not common practice. The lack of regulatory requirements is likely to have led to variations in QC tests that are generally either harder to perform or are more labour intensive.

To propose a style transfer model for multi-contrast magnetic resonance imaging (MRI) images with a cycle-consistent generative adversarial network (CycleGAN) and evaluate the image quality and prognosis prediction performance for glioblastoma (GBM) patients from the extracted radiomics features. Style transfer models of T1 weighted MRI image (T1w) to T2 weighted MRI image (T2w) and T2w to T1w with CycleGAN were constructed using the BraTS dataset. The style transfer model was validated with the Cancer Genome Atlas Glioblastoma Multiforme (TCGA-GBM) dataset. Moreover, imaging features were extracted from real and synthesized images. These features were transformed to rad-scores by the least absolute shrinkage and selection operator (LASSO)-Cox regression. The prognosis performance was estimated by the Kaplan-Meier method. For the accuracy of the image quality of the real and synthesized MRI images, the MI, RMSE, PSNR, and SSIM were 0.991 ± 2.10 $\times {10}^{-4}$ , 2.79 ± 0.16, 40.16 ± 0.38, and 0.995 ± 2.11 $\times {10}^{-4}$ , for T2w, and .992 ± 2.63 $\times {10}^{-4}$ , 2.49 ± 6.89 $\times {10}^{-2}$ , 40.51 ± 0.22, and 0.993 ± 3.40 $\times {10}^{-4}$ for T1w, respectively. The survival time had a significant difference between good and poor prognosis groups for both real and synthesized T2w (p < 0.05). However, the survival time had no significant difference between good and poor prognosis groups for both real and synthesized T1w. On the other hand, there was no significant difference between the real and synthesized T2w in both good and poor prognoses. The results of T1w were similar in the point that there was no significant difference between the real and synthesized T1w. It was found that the synthesized image could be used for prognosis prediction. The proposed prognostic model using CycleGAN could reduce the cost and time of image scanning, leading to a promotion to build the patient's outcome prediction with multi-contrast images.

Manual segmentation poses a time-consuming challenge for disease quantification, therapy evaluation, treatment planning, and outcome prediction. Convolutional neural networks (CNNs) hold promise in accurately identifying tumor locations and boundaries in PET scans. However, a major hurdle is the extensive amount of supervised and annotated data necessary for training. To overcome this limitation, this study explores semi-supervised approaches utilizing unlabeled data, specifically focusing on PET images of diffuse large B-cell lymphoma (DLBCL) and primary mediastinal large B-cell lymphoma (PMBCL) obtained from two centers. We considered 2-[¹⁸F]FDG PET images of 292 patients PMBCL (n = 104) and DLBCL (n = 188) (n = 232 for training and validation, and n = 60 for external testing). We harnessed classical wisdom embedded in traditional segmentation methods, such as the fuzzy clustering loss function (FCM), to tailor the training strategy for a 3D U-Net model, incorporating both supervised and unsupervised learning approaches. Various supervision levels were explored, including fully supervised methods with labeled FCM and unified focal/Dice loss, unsupervised methods with robust FCM (RFCM) and Mumford-Shah (MS) loss, and semi-supervised methods combining FCM with supervised Dice loss (MS + Dice) or labeled FCM (RFCM + FCM). The unified loss function yielded higher Dice scores (0.73 ± 0.11; 95% CI 0.67-0.8) than Dice loss (p value < 0.01). Among the semi-supervised approaches, RFCM + αFCM (α = 0.3) showed the best performance, with Dice score of 0.68 ± 0.10 (95% CI 0.45-0.77), outperforming MS + αDice for any supervision level (any α) (p < 0.01). Another semi-supervised approach with MS + αDice (α = 0.2) achieved Dice score of 0.59 ± 0.09 (95% CI 0.44-0.76) surpassing other supervision levels (p < 0.01). Given the time-consuming nature of manual delineations and the inconsistencies they may introduce, semi-supervised approaches hold promise for automating medical imaging segmentation workflows.

The use of Al₂O₃:C-based optically stimulated luminescent dosimeters (OSLDs) in diagnostic X-ray is a challenge because of their energy dependence (ED) and variability of element sensitivity factors (ESFs). This study aims to develop a method to determine ED and ESFs of Landauer nanoDot™ OSLDs for clinical X-ray and investigate the uncertainties associated with ESF and ED correction factors. An area of 2 × 2 cm² at the central axis of the X-ray field was used to establish the ESFs. A total of 80 OSLDs were categorized into "controlled" (n = 40) and "less-controlled" groups (n = 40). The ESFs of the OSLDs were determined using an 80 kVp X-ray beam quality in free-air geometry. The OSLDs were cross-calibrated with an ion chamber to establish the average calibration coefficient and ESFs. The OSLDs were then irradiated at tube potentials ranging from 50 to 150 kVp to determine their ED. The uniformity of the X-ray field was ± 1.5% at 100 cm source-to-surface distance. The batch homogeneities of user-defined ESFs were 2.4% and 8.7% for controlled and less-controlled OSLDs, respectively. The ED of OSLDs ranged from 1.125 to 0.812 as tube potential increased from 50 kVp to 150 kVp. The total uncertainty of OSLDs, without ED correction, could be as high as 16%. After applying ESF and ED correction, the total uncertainties were reduced to 6.3% in controlled OLSDs and 11.6% in less-controlled ones. OSLDs corrected with user-defined ESF and ED can reduce the uncertainty of dose measurements in diagnostic X-rays, particularly in managing less-controlled OSLDs.

Plasmonic photothermal therapy (PPTT) involves the use of nanoparticles and near-infrared radiation to attain a temperature above 50 °C within the tumor for its thermal damage. PPTT is largely explored for superficial tumors, and its potential to treat deeper subsurface tumors is dealt feebly, requiring the assessment of thermal damage for such tumors. In this paper, the extent of thermal damage is numerically analyzed for PPTT of invasive ductal carcinoma (IDC) situated at 3-9 mm depths. The developed numerical model is validated with suitable tissue-tumor mimicking phantoms. Tumor (IDC) embedded with gold nanorods (GNRs) is subjected to broadband near-infrared radiation. The effect of various GNRs concentrations and their spatial distributions [viz. uniform distribution, intravenous delivery (peripheral distribution) and intratumoral delivery (localized distribution)] are investigated for thermal damage for subsurface tumors situated at various depths. Results show that lower GNRs concentrations lead to more uniform internal heat generation, eventually resulting in uniform temperature rise. Also, the peripheral distribution of nanoparticles provides a more uniform spatial temperature rise within the tumor. Overall, it is concluded that PPTT has potential to induce thermal damage for subsurface tumors, at depths of upto 9 mm, by proper choice of nanoparticle distribution, dose/concentration and irradiation parameters based on the tumor location. Moreover, intravenous administration of nanoparticles seems a good choice for shallower tumors, while for deeper tumors, uniform distribution is required to attain the necessary thermal damage. In the future, the algorithm may be extended further, involving 3D patient-specific tumors and through mice model-based experiments.

The ECG is a crucial tool in the medical field for recording the heartbeat signal over time, aiding in the identification of various cardiac diseases. Commonly, the interpretation of ECGs necessitates specialized knowledge. However, this paper explores the application of machine learning algorithms and deep learning algorithm to autonomously identify cardiac diseases in diabetic patients in the absence of expert intervention. Two models are introduced in this study: The MLP model effectively distinguishes between individuals with heart diseases and those without, achieving a high level of accuracy. Subsequently, the deep CNN model further refines the identification of specific cardiac conditions. The PTB-Diagnostic ECG dataset commonly used in the field of biomedical signal processing and machine learning, particularly for tasks related to electrocardiogram (ECG) analysis. a widely recognized dataset in the field, is employed for training, testing, and validation of both the MLP and CNN models. This dataset comprises a diverse range of ECG recordings, providing a comprehensive representation of cardiac conditions. The proposed models feature two hidden layers with weights and biases in the MLP, and a three-layer CNN, facilitating the mapping of ECG data to different disease classes. The experimental results demonstrate that the MLP and deep CNN based models attain accuracy levels of up to 90.0% and 98.35%, and sensitivity 97.8%, 95.77%, specificity 88.9%, 96.3% F1-Score 93.13%, 95.84% respectively. These outcomes underscore the efficacy of deep learning approaches in automating the diagnosis of cardiac diseases through ECG analysis, showcasing the potential for accurate and efficient healthcare solutions.

Manual contouring of organs at risk (OAR) is time-consuming and subject to inter-observer variability. AI-based auto-contouring is proposed as a solution to these problems if it can produce clinically acceptable results. This study investigated the performance of multiple AI-based auto-contouring systems in different OAR segmentations. The auto-contouring was performed using seven different AI-based segmentation systems (Radiotherapy AI, Limbus AI version 1.5 and 1.6, Therapanacea, MIM, Siemens AI-Rad Companion and RadFormation) on a total of 42 clinical cases with varying anatomical sites. Volumetric and surface dice similarity coefficients and maximum Hausdorff distance (HD) between the expert's contours and automated contours were calculated to evaluate their performance. Radiotherapy AI has shown better performance than other software in most tested structures considered in the head and neck, and brain cases. No specific software had shown overall superior performance over other software in lung, breast, pelvis and abdomen cases. Each tested AI system was able to produce comparable contours to the experts' contours of organs at risk which can potentially be used for clinical use. A reduced performance of AI systems in the case of small and complex anatomical structures was found and reported, showing that it is still essential to review each contour produced by AI systems for clinical uses. This study has also demonstrated a method of comparing contouring software options which could be replicated in clinics or used for ongoing quality assurance of purchased systems.