RESUMO
AIMS: With growing concerns over complications in diabetes sufferers, this study sought to develop an interpretable machine learning model to offer enhanced diagnostic and treatment recommendations. METHODS: We assessed coronary heart disease, diabetic nephropathy, diabetic retinopathy, and fatty liver disease using logistic regression, decision tree, random forest, and CatBoost algorithms. The SHAP algorithm was employed to elucidate the model's predictions, offering a more in-depth understanding of influential features. RESULTS: The CatBoost model notably outperformed other algorithms in AUC, achieving an average AUC of 90.47 % for the four complications. Through SHAP analysis and visualization, we provided clear and actionable insights into risk factors, enabling better complication risk assessment. CONCLUSIONS: We introduced an innovative, interpretable complication risk model for people with diabetes. This not only offers a potent tool for healthcare professionals but also empowers sufferers with clearer self-assessment capabilities, encouraging earlier preventive actions. Further studies will underscore the model's clinical applicability.
Assuntos
Diabetes Mellitus , Nefropatias Diabéticas , Retinopatia Diabética , Humanos , Retinopatia Diabética/diagnóstico , Retinopatia Diabética/epidemiologia , Retinopatia Diabética/etiologia , Algoritmos , Povo Asiático , China/epidemiologiaRESUMO
A dose reconstruction algorithm for electrionic portal imaging device(EPID)based on calibration and calculation is developed.The raw data of EPID in continuous acquisition mode are corrected for dark field and gain,and the gray level features of bright field are used to determine the field boundary.Subsequently,MU calibration,off-axis calibration and field size calibration are performed on the EPID data,and dose reconstruction is carried out based on the calibrated superimposed flux and the Monte Carlo model of the linac head.Nine cases of IMRT plans are selected for verification and measurement using EPID and MapCheck separately,and the passing rates between the two tools are compared under different gamma criteria(3%/3 mm and 2%/2 mm).For a planned case,the average passing rates of multiple cases verified by MapCheck under the two criteria were 99.02%±1.28%and 90.84%±4.49%,and the average passing rates of the EPID reconstruction models were 98.86%±1.19%and 91.39%±4.80%.Compared with MapCheck,the EPID reconstruction algorithm based on calibration and calculation has no significant difference in the passing rate of IMRT plan verification(P>0.05),which meets the clinical requirements of dose verification.
RESUMO
Objective:To investigate the prediction of the tumor proliferation antigen(Ki-67) expression status in breast cancer patients based on ultrasound radiomics combined with clinicopathologic features.Methods:Breast cancer patients who underwent 2D ultrasound and Ki-67 examination from January 2018 to February 2022 in Changzhou Second People′s Hospital, Nanjing Medical University were retrospectively analyzed. Among them, 427 patients from Chengzhong campus were randomly divided into training and validation sets in the ratio of 8∶2, and 229 patients from Yanghu campus were used as an independent external test set. Radiomics features were extracted from the region of interest of 2D ultrasound images, and the Mann-Whitney U test, recursive feature elimination, and minimum absolute shrinkage and selection operators were used to perform feature dimensionality reduction and to establish a radiomics score(Rad-score). Subsequently, single/multifactor logistic regression regression analyses were used to construct a joint prediction model based on Rad-score and clinicopathological features. Model performance and utility were assessed using the subject operating characteristic area under the curve (AUC), calibration curve, and decision curve analyses. Results:The AUCs of the joint model for predicting Ki-67 expression status in breast cancer in the training, validation, and test sets were 0.858, 0.797, and 0.802, respectively, which were superior to those of the radiomics (0.772, 0.731, and 0.713) and clinical models (0.738, 0.750, and 0.707). Calibration curve and decision curve analyses indicated that the joint model had good calibration and clinical value.Conclusions:A joint model based on ultrasound radiomics and clinicopathological features can effectively predict the Ki-67 expression status of breast cancer, which is expected to become a non-invasive tool for Ki-67 detection and provide clinicians with an important auxiliary diagnostic and therapeutic decision-making basis.
RESUMO
Objective:To evaluate the feasibility of 3D reconstruction techniques based on multi-depth cameras for daily patient positioning in radiotherapy.Methods:Through region of interest (ROI) extraction, filtering, registration, splicing and other processes, multi-depth cameras (Intel RealSense D435i) were used to fuse point clouds in real-time manner to obtain the real optical 3D surface of patients. The reconstructed surface was matched with the external contour of the localization CT to complete the positioning. In this article, the feasibility of the system was validated by using multiple models. Clinical feasibility of 5 patients with head and neck radiotherapy, 10 cases of chest radiotherapy and 5 cases of pelvic radiotherapy was also validated. The data of each group were analyzed by paired t-test. Results:The system running time was 0.475 s, which met the requirement of real-time monitoring. The six-dimensional registration errors in the model experiment were (1.00±0.74) mm, (1.69±0.69) mm, (1.36±0.87) mm, 0.15°±0.14°, 0.25°±0.20°, 0.13°±0.13° in the x, y, z, rotational, pitch and roll directions, respectively. In the actual patient positioning, the mean positioning errors were (0.77±0.51) mm, (1.24±0.67) mm, (0.94±0.76) mm, 0.61°±0.41°, 0.69°±0.55°, and 0.52°±0.35° in the x, y, z, rotational, pitch and roll directions, respectively. The translational error was less than 2.8 mm, and the positioning error was the largest in the pelvic region. Conclusions:Real-time 3D reconstruction techniques based on multi-depth cameras is applicable for patient positioning during radiotherapy. The method is accurate in positioning and can detect the small movement of the patient's position, which meets the requirements of radiotherapy.
RESUMO
Radiation therapy (RT), as a crucial part of current cancer treatment, has caused great concern since it brings therapeutic efficacy along with the risk of chronic complications. With an increase in age, patients treated with RT are subjected to a high incidence of vascular diseases in the neck and peripheral heart primarily due to the artery stenosis induced by radiation-caused vascular injury. To gain a deeper understanding of artery stenosis, its hazards, clinical presentation, pathogenesis, preventive recommendations, and treatment method was reviewed.
RESUMO
Objective:To construct a cycle dual-task network based on cycleGAN to implement 3D CT synthesis from single-view projection for adaptive radiotherapy of thoracic tumor and then evaluate image quality and dose accuracy.Methods:A total of 45 thoracic tumor patients admitted to the Affiliated Changzhou No.2 People's Hospital of Nanjing Medical University were collected, and 991 cases were also selected from public dataset as pretrained dataset. Multi-view projections were acquired by ASTRA algorithm. The public dataset was divided into a training set of 800 cases, a validation set of 160 cases and a test set of 31 cases. The dataset obtained from patients in our hospital was divided into a training set of 40 cases and a test set of 5 cases. The network included synthetic CT model and multi-view projection prediction model and achieved the dual-task training. The final test only used the synthetic CT model to acquire the predicted CT images and deliver image quality [mean absolute error (MAE), peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM)] and dose evaluation.Results:Image quality evaluation metrics for synthetic CT showed high image synthesis accuracy with MAE of 0.05±0.01, PSNR of 19.08±1.69, SSIM of 0.75±0.04, respectively. The dose distribution calculated on synthetic CT was also close to the actual dose distribution. The mean 3%/3 mm γ pass rate for synthetic CT was 93.1%.Conclusions:A dual-task cycle network modified on cycleGAN has been implemented to rapidly and accurately predict 3D CT from single-view projection, which can be applied to the workflow of adaptive radiotherapy for thoracic cancer. Both image generation quality and dosimetric evaluation demonstrate that synthetic CT can meet the clinical requirements for radiotherapy.
RESUMO
Image-guided radiation therapy (IGRT) is a visual image-guided radiotherapy technique that has many advantages such as increasing the dose of tumor target area and reducing the dose of normal organ exposure. Cone beam CT (CBCT) is one of the most commonly used medical images in IGRT, and the rapid and accurate targeting of CBCT and the segmentation of dangerous organs are of great significance for radiotherapy. The current research method mainly includes partitioning method based on registration and segmentation method based on deep learning. This study reviews the CBCT image segmentation method, existing problems and development directions.
RESUMO
Objective:To develop the real-time radiotherapy monitoring system of three-dimensional (3D) point cloud by using depth camera and verify its feasibility.Methods:Taking the depth camera coordinate system as the world coordinate system, the conversion relationship between the simulation CT coordinate system and the world coordinate system was obtained from the calibration module. The patient's simulation CT point cloud was transformed into the world coordinate system through the above relationship, and registered with the patient's surface point cloud obtained in real-time manner by the depth camera to calculate the six-dimensional (6D) error, and complete the positioning verification and fractional internal position error monitoring in radiotherapy. Mean and standard deviation of 6D calculation error, Hausdorff distance of point cloud after registration and the running time of each part of the program were calculated to verify the feasibility of the system. Fifteen real patients were selected to calculate the 6D error between the system and cone beam CT (CBCT).Results:In the phantom experiment, the errors of the system in the x, y and z axes were (1.292±0.880)mm, (1.963±1.115)mm, (1.496±1.045)mm, respectively, and the errors in the rotation, pitch and roll directions were 0.201°±0.181°, 0.286°±0.326°, 0.181°±0.192°, respectively. For real patients, the translational error of the system was within 2.6 mm, the rotational error was approximately 1°, and the program run at 1-2 frames/s. The precision and speed met the radiotherapy requirement. Conclusion:The 3D point cloud radiotherapy real-time monitoring system based on depth camera can automatically complete the positioning verification before radiotherapy, real-time monitoring of body position during radiotherapy, and provide error visual feedback, which has potential clinical application value.
RESUMO
In the current clinical diagnosis, medical images have become an important basis for diagnosis, and different modes of medical images provide different tissue information and functional information. Single-mode images can only provide single diagnostic information, by which difficult and complicated diseases cannot be diagnosed, and comprehensive and accurate diagnostic results can be obtained only with the help of multiple diagnostic information. The multimodal fusion technology fuses multiple modes of medical images into single-mode images, and thus the single-mode images contain complementary information between multiple modes of images, so that sufficient information for clinical diagnosis can be obtained in a single image. In this paper, the multimodal medical image fusion methods are sorted into two types, namely the traditional fusion method and the fusion method based on deep learning.
RESUMO
@#<b>Objective</b> To investigate the dosimetric effect of truncated regions in computed tomography (CT) images on the targets and organs at risk in volumetric modulated arc therapy (VMAT) for middle thoracic esophageal cancer. <b>Methods</b> CT images of 15 patients with middle thoracic esophageal cancer were selected. Circle masks were used to make the volume of the truncated region account for 10%, 20%, 30%, and 40% of the arm volume, and the corresponding truncated CT images were obtained. The real CT was denoted as CT0. Two radiotherapy plans were made on CT0. One plan was VMAT_1F with full arcs, and the other one was VMAT_3F with arm avoidance. The plans were transplanted to four truncated CT, respectively, and the dosimetric differences between different plans were compared using Wilcoxon signed-rank test. <b>Results</b> Compared with VMAT_1F in CT0, <i>D</i><sub>mean</sub> and <i>V</i><sub>5</sub> of the lung decreased in VMAT_3F, but <i>D</i><sub>max</sub> of the spinal cord, <i>D</i><sub>mean</sub> of the heart, and <i>V</i><sub>20</sub> of the lung increased. In VMAT_3F, there was no statistically significant difference between the dosimetric parameters in the four truncated CT and those in CT0 (all <i>P</i> > 0.05). In VMAT_1F, except for homogeneity index and <i>D</i><sub>max</sub> of the spinal cord, the dosimetric parameters in four truncated CT were significantly different from those in CT0 (<i>P</i> < 0.05). The dosimetric difference increased with the increase in truncated region-to-volume ratio. <b>Conclusion</b> Complete CT data should be collected in clinical practice, and the radiation field avoiding the truncated regionshould be set if necessary to reduce the influence of the truncated region on dosimetry.
RESUMO
Surface guided radiation therapy (SGRT) is a method of radiation therapy with non-invasive and non-radiation image guidance technology, which uses continuous real-time imaging to monitor the whole course of treatment. This paper summarizes the characteristics, representative products, application in clinical research and treatment, and quality control of SGRT. This emerging technology plays an increasingly important role in delivering more precise, safe, and comfortable radiotherapy to patients.
RESUMO
With the improvement of the complexity of medical image synthesis and the demand for the accuracy of clinical radiotherapy, deep learning algorithm plays an increasingly important role in pseudo CT image synthesis and analysis. This paper classifies and analyzes the pseudo CT image synthesis technology based on deep learning method in terms of the types of image modes, and describes the ongoing application in radiotherapy.
RESUMO
Objective:To develop a 3D visualization technology-assisted patient positioning system for radiotherapy and compare it with traditional patient positioning method for breast and pelvic radiotherapy.Methods:A total of 40 patients receiving radiotherapy in Changzhou No.2 People′s Hospital from June 2020 to April 2021 were selected for this study, including 20 patients with breast cancer and 20 patients with pelvic cancer.3D visualization reconstruction was carried out using the CT data of the patients for positioning. Then the 3D visualization models were integrated with the real treatment environment and were then shifted to the isocentral positions of accelerators through interactive operations. Based on this, the patients were actually positioned. Every week, all of the patients were firstly treated with traditional positioning, followed by 3D visualization-guided positioning. As a result, 240 times of positioning data of all patients were collected in three weeks. They were compared with the data of cone-beam CT(CBCT)-guided positioning, which served as the gold standard.Results:The absolute positioning errors of 3D visualization-guided positioning along x, y and z axes were (1.92±1.23), (2.04±1.16), and (1.77±1.37)mm, respectively for patients with breast cancer and were (2.07±1.08), (1.33±0.88), and (1.99±1.25)mm, respectively for patients with pelvic cancer. Compared with traditional positioning method , the accuracy of 3D visualization-guided positioning along x、 y, and z axes was increased by 38.83%, 52.40% and 33%, respectively for patients with breast cancer and was improved by 36.84%, 54.04% and 52.58% for patients with pelvic cancer, with all differences being statistically significant along y and z axes ( t=2.956-5.734, P< 0.05). Meanwhile, the error distribution of the two positioning method was statistically significant along in y axis for patients with breast cancer( χ2=7.481, P<0.05) and was statistically significant along each axis for patients with pelvic cancer( χ2=5.900, 6.415, 7.200, P<0.05). Conclusions:The positioning method guided by 3D visualization technology can effectively improve the positioning accuracy of patients with breast cancer and patients with pelvic cancer and is of value in potential clinical application.
RESUMO
Magnetic resonance imaging (MRI) is a technology with no radiation and high resolution of soft tissues. Therefore, MRI-guided radiotherapy has become a hot spot in the field of radiotherapy. It is of great importance to accurately delineate the targets in radiation oncology. Currently, the delineation of targets is mostly completed by manual segmentation, which is time-consuming, subjective and inconsistent. Automatic segmentation can improve the efficiency and consistency without sacrificing the accuracy of segmentation. In this article, the automatic segmentation methods of MRI applied in radiotherapy were reviewed. The goals, challenges and methods of automatic segmentation for different radiotherapy sites including prostate, nasopharyngeal carcinoma, brain tumors and other organs were analyzed and discussed.
RESUMO
Objective:To explore the volume resolution of prostate motion target by four-dimensional (4D) ultrasound.Methods:The prostate ultrasound model was selected, and the group comparison study was conducted using 4D ultrasound to outline the prostate target under different motion amplitudes (A) and motion period (T). The simulated A value was set as 0.5 mm, 1 mm, 2 mm, 3 mm, 4 mm, and 5 mm, respectively. The T value was set as 1 s, 2 s, 3 s, and 4 s, respectively. The volume of the target of the model prostate was calculated, and the static ultrasound image of the target was used as the control group to analyze the difference between two groups.Results:When the model was still, the size of the target of ultrasound was consistent with that of CT scan ( P>0.05). When the A values were 0.5 mm and 1 mm, there was no statistical difference between the volume in period 1-4 s and the volume in the target at rest (all P>0.05). When the A values were 2 mm and 3 mm, and the T values were 1 s, 2 s and 3 s there was statistical difference between the volume of target and that of of static ultrasonic target (all P<0.05). When the A value was 2 mm and the T value was 4 s, there was no statistical difference between the target volume and the static target volume ( P=0.710). The range within the group was 6.7 cm 3, and the standard deviation was 1.15 cm 3. When the A value was 3 mm and the T value was 4 s, the volume repeatability of the target was poor, and the range within the group was 14.4 cm 3; when the A values were 4 mm and 5 mm, and the T values were 1-4 s, the range within the group was 3.27-17.63 cm 3 and 6.51-21.02 cm 3, respectively. The volume repeatability of the target under each period was extremely poor, which could not meet the clinical requirements. Conclusion:4D ultrasound can provide reliable reference data for patients′ target delineation within 1-4 s of motion cycle and within 1 mm of motion amplitude, which exerts on effect upon the original position of probe.
RESUMO
Medical images can provide clinicans with accurate and comprehensive patients’ information. Morphological or functional abnormalities caused by various diseases can be manifested in many aspects. Although MR images and CT images can highlight the medical image data of different tissue structures of patients, single MR images or CT images cannot fully reflect the complexity of diseases. Using MR image to predict CT image is one of the cross-modal prediction of medical images. In this paper, the methods of MR image prediction for CTmage are classified into four categoriesincluding registration based on atlas, based on image segmentationmethod, based on learning method and based on deep learning method. In our research, we concluded that the method based on deep learning should bemore promoted in the future by compering the existing problems and future development of MR image predicting CT image method.
RESUMO
Objective: To generate virtual non-contrast (VNC) computed tomography (CT) from intravenous enhanced CT through convolutional neural networks (CNN) and compare calculated dose among enhanced CT, VNC, and real non-contrast scanning. Method: 50 patients who accepted non-contrast and enhanced CT scanning before and after intravenous contrast agent injections were selected, and two sets of CT images were registered. A total of 40 and 10 groups were used as training and test datasets, respectively. The U-Net architecture was applied to learn the relationship between the enhanced and non-contrast CT. VNC images were generated in the test through the trained U-Net. The CT values of non-contrast, enhanced and VNC CT images were compared. The radiotherapy treatment plans for esophageal cancer were designed, and dose calculation was performed. Dose distributions in the three image sets were compared. Results: The mean absolute error of CT values between enhanced and non-contrast CT reached 32.3 ± 2.6 HU, and that between VNC and non-contrast CT totaled 6.7 ± 1.3 HU. The average CT values in enhanced CT of great vessels, heart, lungs, liver, and spinal cord were all significantly higher than those of non-contrast CT (p < 0.05), with the differences reaching 97, 83, 42, 40, and 10 HU, respectively. The average CT values of the organs in VNC CT showed no significant differences from those in non-contrast CT. The relative dose differences of the enhanced and non-contrast CT were -1.2, -1.3, -2.1, and -1.5% in the comparison of mean doses of planned target volume, heart, great vessels, and lungs, respectively. The mean dose calculated by VNC CT showed no significant difference from that by non-contrast CT. The average γ passing rate (2%, 2 mm) of VNC CT image was significantly higher than that of enhanced CT image (0.996 vs. 0.973, p < 0.05). Conclusion: Designing a treatment plan based on enhanced CT will enlarge the dose calculation uncertainty in radiotherapy. This paper proposed the generation of VNC CT images from enhanced CT images based on U-Net architecture. The dose calculated through VNC CT images was identical with that obtained through real non-contrast CT.
RESUMO
Medical images play an important role in clinical diagnosis and treatment. During the radiotherapy, CT can be available for the location and definition of the target volume. The medical images from multiple modalities are used to obtain the information on pathological body from many angles. However, obtaining multiple-modality medical images could be more resource-consuming, and difficult to guarantee the consistency of patients′ state. Medical image translation between multiple modalities can achieve the predication from one modality to another. The studies on medical images from multiple modalities such as CT, ultrasound, MRI and PET are reviewed in detail in this paper, , with discussions provided about characteristics of multiple modalities and challenges faced, as well as the research areas to be developed.
RESUMO
Image-guided techniques are critical to improving the accuracy of radiotherapy for tumors. Ultrasound images have been gradually applied in the set-up verification of clinical radiotherapy and adaptive radiotherapy because of the real-time, reproducible and non-radiative characteristics. In this paper, the application of ultrasound image-guided technology in radiotherapy was classified and analyzed, and the latest research progress was introduced.
RESUMO
Objective An improved method for obtaining pseudo-computed tomography (CT ps) based on ultrasound deformation field.Methods The three-dimensional image data of computed tomography and ultrasound for three postoperative cervical cancer patients were selected,including the CT (CTsim) and ultrasound (USsim) images obtained during the simulated positioning stage,and the cone beam CT (CBCT) and ultrasound images obtained during the positioning verification stage of the treatment one week later.Binary masks of the OROI and OROW were created and applied in ultrasound image registration;thus,the deformation field was obtained.The deformation field was applied to CTsim images and different pseudo-CT images were obtained.Similarities between these pseudo-CT images and those of CBCT were compared,and registration accuracies between pseudo-CT images under different binary masks and CTsim were discussed.Results The averages of the correlation coefficient between pseudo-CT based on OROI,OROW,no binary mask and CBCT were 0.95,0.82 and 0.64 respectively.The average of the normalized mean square Error were 0.12,0.42 and 0.57 respectively.Conclusion The pseudo-CT based on OROI binary mask matches the best with CTsim and achieves the highest similarity with CBCT.