Japanese Society Of Radiological Technology Research Articles

Encounter with the Journal of the Japanese Society of Radiological Technology (JSRT) and Its Digitization

Encounter with the Journal of the Japanese Society of Radiological Technology (JSRT) and Its Digitization

Aiming for Japanese Society of Radiological Technology (JSRT) That Spreads Its Wings to the World

Aiming for Japanese Society of Radiological Technology (JSRT) That Spreads Its Wings to the World

A deep unsupervised saliency model for lung segmentation in chest X-ray images

Medical image segmentation is a critical and laborious step in the diagnosis workflow by separating regions of organs or anomalies to be analyzed further by a physician, or classified automatically by an intelligent tool. Deep machine learning techniques have achieved impressive results in image segmentation tasks whenever immense amounts of labeled data are provided. However, having big amounts of manually supervised image data is labor intensive, and costly impractical. Unsupervised learning, contrariwise, can provide results using unlabeled images by reaching decisions, clustering, and deriving insights from the data only. We propose a deep unsupervised learning approach to perform lung segmentation in chest X-ray images, which is guided by heuristic saliency maps as a noisy labeling framework, which through a composite loss of reconstruction and regularization reach saliency estimation as final segmentation. Tests were performed using public chest X-ray images from the Japanese Society of Radiological Technology (JSRT), and from Montgomery County (MC) datasets. The proposed model based on saliency scored a dice coefficient of 0.87, against a state-of-the-art unsupervised method, which scored 0.68, on JSRT, and on the MC dataset 0.88, and 0.64, respectively. Qualitatively comparing, the saliency model worked better in reducing background noise. Considering that unsupervised models can be used as robust preparation steps, and as baselines for further different tasks, either guided by constant feedback from physicians, or as input to semi-supervised learning, the results have shown this to be a promise line of research to medical image segmentation problems.

A Patient-Specific Algorithm for Lung Segmentation in Chest Radiographs

Lung segmentation plays an important role in computer-aided detection and diagnosis using chest radiographs (CRs). Currently, the U-Net and DeepLabv3+ convolutional neural network architectures are widely used to perform CR lung segmentation. To boost performance, ensemble methods are often used, whereby probability map outputs from several networks operating on the same input image are averaged. However, not all networks perform adequately for any specific patient image, even if the average network performance is good. To address this, we present a novel multi-network ensemble method that employs a selector network. The selector network evaluates the segmentation outputs from several networks; on a case-by-case basis, it selects which outputs are fused to form the final segmentation for that patient. Our candidate lung segmentation networks include U-Net, with five different encoder depths, and DeepLabv3+, with two different backbone networks (ResNet50 and ResNet18). Our selector network is a ResNet18 image classifier. We perform all training using the publicly available Shenzhen CR dataset. Performance testing is carried out with two independent publicly available CR datasets, namely, Montgomery County (MC) and Japanese Society of Radiological Technology (JSRT). Intersection-over-Union scores for the proposed approach are 13% higher than the standard averaging ensemble method on MC and 5% better on JSRT.

Practical guidelines of online MR-guided adaptive radiotherapy.

The first magnetic resonance (MR)-guided radiotherapy system in Japan was installed in May 2017. Implementation of online MR-guided adaptive radiotherapy (MRgART) began in February 2018. Online MRgART offers greater treatment accuracy owing to the high soft-tissue contrast in MR-images (MRI), compared to that in X-ray imaging. The Japanese Society for Magnetic Resonance in Medicine (JSMRM), Japan Society of Medical Physics (JSMP), Japan Radiological Society (JRS), Japanese Society of Radiological Technology (JSRT), and Japanese Society for Radiation Oncology (JASTRO) jointly established the comprehensive practical guidelines for online MRgART. These guidelines propose the essential requirements for clinical implementation of online MRgART with respect to equipment, personnel, institutional environment, practice guidance, and quality assurance/quality control (QA/QC).The minimum requirements for related equipment and QA/QC tools, recommendations for safe operation of MRI system, and the implementation system are described. The accuracy of monitor chamber and detector in dose measurements should be confirmed because of the presence of magnetic field. The ionization chamber should be MR-compatible. Non-MR-compatible devices should be used in an area that is not affected by the static magnetic field (outside the five Gauss line), and their operation should be checked to ensure that they do not affect the MR image quality. Dose verification should be performed using an independent dose verification system that has been confirmed to be reliable through commissioning. This guideline proposes the checklists to ensure the safety of online MRgART. Successful clinical implementation of online MRgART requires close collaboration between physician, radiological technologist, nurse, and medical physicist.

CardioNet: Automatic Semantic Segmentation to Calculate the Cardiothoracic Ratio for Cardiomegaly and Other Chest Diseases.

Semantic segmentation for diagnosing chest-related diseases like cardiomegaly, emphysema, pleural effusions, and pneumothorax is a critical yet understudied tool for identifying the chest anatomy. A dangerous disease among these is cardiomegaly, in which sudden death is a high risk. An expert medical practitioner can diagnose cardiomegaly early using a chest radiograph (CXR). Cardiomegaly is a heart enlargement disease that can be analyzed by calculating the transverse cardiac diameter (TCD) and the cardiothoracic ratio (CTR). However, the manual estimation of CTR and other chest-related diseases requires much time from medical experts. Based on their anatomical semantics, artificial intelligence estimates cardiomegaly and related diseases by segmenting CXRs. Unfortunately, due to poor-quality images and variations in intensity, the automatic segmentation of the lungs and heart with CXRs is challenging. Deep learning-based methods are being used to identify the chest anatomy segmentation, but most of them only consider the lung segmentation, requiring a great deal of training. This work is based on a multiclass concatenation-based automatic semantic segmentation network, CardioNet, that was explicitly designed to perform fine segmentation using fewer parameters than a conventional deep learning scheme. Furthermore, the semantic segmentation of other chest-related diseases is diagnosed using CardioNet. CardioNet is evaluated using the JSRT dataset (Japanese Society of Radiological Technology). The JSRT dataset is publicly available and contains multiclass segmentation of the heart, lungs, and clavicle bones. In addition, our study examined lung segmentation using another publicly available dataset, Montgomery County (MC). The experimental results of the proposed CardioNet model achieved acceptable accuracy and competitive results across all datasets.

Automatic Tissue Attenuation-Based Contrast Enhancement of Low-Dynamic X-Ray Images

In medical imaging, X-ray images are an important part to diagnose the disease of the human body. The acquisition of X-ray images through X-rays, having a low-dynamic range and low contrast, such that the identification of bones and organs is more difficult. So that the contrast improvement is more needful for the X-ray images to identify the body organs. To improve the contrast of the low-dynamic X-ray images, a novel concept automatic tissue attenuation-based contrast enhancement (ATACE) method has been proposed. In this method, the X-ray image is decomposed into two components, i.e., detailed components and tissue components. The tissue component is modified using a removable factor and produced the removable component for enhancing the visible contrast of both dark and bright regions of the X-ray image. The automatic adjustment of the contrast parameter is calculated using local maximum and removable components to enhance the contrast of the X-ray image. The enhanced image is obtained using simple calculation with the help of removable components and contrast parameters. The proposed technique is applied to the X-ray dataset picked from the Japanese Society of Radiological Technology (JSRT). The parametric evaluation metrics of the proposed technique are achieved promising scores.

TAE-Seg: Generalized Lung Segmentation via Tilewise AutoEncoder Enhanced Network

Computer-aided diagnosis based on deep learning is progressively deployed for the analysis of medical images, yet the poor <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">robustness</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">generalization</i> of the model poses a challenge for clinical application. In addition, the lack of large amount of training data aggravates this problem. To mitigate this issue, we investigate and research from the transfer learning (i.e., pretraining-finetuning) perspective. More important, unlike the traditional setting that transfers knowledge from the natural image domain to the medical image domain, we find that the knowledge from a similar domain can further boost the model’s robustness and generalization. In this paper, we propose a generalized lung segmentation framework, which includes two parts: (i) an unsupervised tile-wise autoencoder pretraining architecture for learning meaningful and transferable knowledge; (ii) a reconstruction network regularized segmentation model for finetuning. The experiments are conducted on several chest X-ray datasets. Our results show an accurate lung segmentation with 96.95%, 97.19%, and 95.77% Dice coefficient index on the Montgomery County (MC) chest X-ray dataset, Shenzhen (SH) chest X-ray dataset, and Japanese Society of Radiological Technology (JSRT) database, respectively. Moreover, quantitative and qualitative results demonstrate the superior model robustness of the proposed method to data corruption and high generalization performance on unseen datasets, especially when the training data is limited.

Scientific Research Group Report: Nationwide Survey on Radiation Exposure of Pediatric CT Examination in Japan (2018)

To understand the latest pediatric computed tomography (CT) exposure required for the revision of national DRLs. A questionnaire was sent to 409 facilities where the members of the Japanese Society of Radiological Technology and the Japanese Society of Pediatric Radiology are enrolled. We investigated the imaging conditions, CTDIvol, and DLP of the pediatric head, chest, and abdominal CT examinations. In all, 43 facilities (11%) responded to our survey. multi detector-row CT (MDCT) systems were available in all surveyed facilities. More than 98% of the MDCT systems had more than 64 detector rows. The CTDIvol of all CT protocols was lower than the NDRL due to the progress of updating to MDCTs with radiation exposure reduction functions such as an iterative reconstruction, but the DLP of head and abdominal CT protocols of some age group were higher than NDRL. It is necessary to review the imaging protocol with the attending physician and radiologist and consider further optimization of medical exposure.

Lung Segmentation using Active Shape Model to Detect the Disease from Chest Radiography.

Background: Some parametric models are used to diagnose problems of lung segmentation more easily and effectively.Objective: The present study aims to detect lung diseases (nodules and tuberculosis) better using an active shape model (ASM) from chest radiographs.Material and Methods: In this analytical study, six grouping methods, including three primary methods such as physicians, Dice similarity, and correlation coefficients) and also three secondary methods using SVM (Support Vector Machine) were used to classify the chest radiographs regarding diaphragm congestion and heart reshaping. The most effective method, based on the evaluation of the results by a radiologist, was found and used as input data for segmenting the images by active shape model (ASM). Several segmentation parameters were evaluated to calculate the accuracy of segmentation. This work was conducted on JSRT (Japanese Society of Radiological Technology) database images and tuberculosis database images were used for validation.Results: The results indicated that the ASM can detect 94.12 ± 2.34 % and 94.38 ± 3.74 % (mean± standard deviation) of pulmonary nodules in left and right lungs, respectively, from the JRST radiology datasets. Furthermore, the ASM model detected 88.33 ± 6.72 % and 90.37 ± 5.48 % of tuberculosis in left and right lungs, respectively.Conclusion: The ASM segmentation method combined with pre-segmentation grouping can be used as a preliminary step to identify areas with tuberculosis or pulmonary nodules. In addition, this presented approach can be used to measure the size and dimensions of the heart in future studies.

Characteristic evaluation of a simple hand-made semiconductor dosimeter for mammography

Daily quality control of mammography equipment is important for providing optimal images and determining appropriate doses. To popularise dosimeters, the Measurement Division, the Japanese Society of Radiological Technology hold a seminar on making simple dosimeters. At this seminar, a hand-made dosimeter for mammography (HD-M) can be made at low cost. However, HD-Ms employ semiconductors, and their energy responses are subject to significant variations. This investigation involved the determination of precautions when using HD-Ms, examining their energy response characteristics and measurable energy ranges. HD-M has four types of selectors for response correction. When the selector and the tube voltage were equal, the HD-M readings matched that of the ionisation chamber within 5%. However, in case of target filter combinations and measuring tube voltages that the selector does not support, the HD-M readings differed by up to 53% from the ionisation chamber values. HD-M may use different measurement circuits and semiconductor elements depending on the time of the seminar. In this study, it was clarified that the correction factor k, which is the average value of the ratio of the measured value of the ionisation chamber dosimeter to the measured value of HD-M, changes from 0.62 to 1.53 depending on irradiation conditions such as the combination of target filters and the tube voltage. It was demonstrated that HD-M functions sufficiently as a dosimeter for daily management by determining the correction factor using the method proposed in this study.

Exploring Optimality of Piecewise Polynomial Interpolation Functions for Lung Field Modeling in 2D Chest X-Ray Images

In this paper, a landmark based approach, using five different interpolating polynomials (linear, cubic convolution, cubic spline, PCHIP, and Makima) for modeling of lung field region in 2D chest X-ray images have been presented. Japanese Society of Radiological Technology (JSRT) database which is publicly available has been used for evaluation of the proposed method. Selected radiographs are anatomically landmarked using 17 and 16 anatomical landmark points to represent left and right lung field regions, respectively. Local, piecewise polynomial interpolation is then employed to create additional semilandmark points to form the lung contour. Jaccard similarity coefficients and Dice coefficients have been used to find accuracy of the modeled shape through comparison with the prepared ground truth. With the optimality condition of three intermediate semilandmark points, PCHIP interpolation method with an execution time of 5.04873 s is found to be the most promising candidate for lung field modeling with an average Dice coefficient (DC) of 98.20 and 98.54% (for the left and right lung field, respectively) and with the average Jaccard similarity coefficient (JSC) of 96.47 and 97.13% for these two lung field regions. While performance of Makima and cubic convolution is close to the PCHIP with the same optimality condition, i.e., three intermediate semilandmark points, the optimality condition for the cubic spline method is of at least seven intermediate semilandmark points which, however, does not result in better performance in terms of accuracy or execution time.

Multi-label segmentation and detection of COVID-19 abnormalities from chest radiographs using deep learning

Due to COVID-19, demand for Chest Radiographs (CXRs) have increased exponentially. Therefore, we present a novel fully automatic modified Attention U-Net (CXAU-Net) multi-class segmentation deep model that can detect common findings of COVID-19 in CXR images. The architectural design of this model includes three novelties: first, an Attention U-net model with channel and spatial attention blocks is designed that precisely localize multiple pathologies; second, dilated convolution applied improves the sensitivity of the model to foreground pixels with additional receptive fields valuation, and third a newly proposed hybrid loss function combines both area and size information for optimizing model. The proposed model achieves average accuracy, DSC, and Jaccard index scores of 0.951, 0.993, 0.984, and 0.921, 0.985, 0.973 for image-based and patch-based approaches respectively for multi-class segmentation on Chest X-ray 14 dataset. Also, average DSC and Jaccard index scores of 0.998, 0.989 are achieved for binary-class segmentation on the Japanese Society of Radiological Technology (JSRT) CXR dataset. These results illustrate that the proposed model outperformed the state-of-the-art segmentation methods.

Deep LF-Net: Semantic lung segmentation from Indian chest radiographs including severely unhealthy images

A chest radiograph, commonly called chest x-ray (CxR), plays a vital role in the diagnosis of various lung diseases. Automated segmentation of the lungs is an important step to design a computer-aided diagnostic tool for examination of a CxR. Precise lung segmentation is considered extremely challenging because of variance in the shape of the lung caused by health issues, age, and gender. The proposed work investigates the use of an efficient deep convolutional neural network for accurate segmentation of lungs from CxR. We attempt an end to end DeepLabv3+ network which integrates DeepLab architecture, encoder-decoder, and dilated convolution for semantic lung segmentation with fast training and high accuracy. We experimented with the different pre-trained base networks: Resnet18 and Mobilenetv2, associated with the Deeplabv3+ model for performance analysis. The proposed approach does not require any pre-processing technique on chest x-ray images before being fed to a neural network. We construct a dataset of CxR images with corresponding lung mask of the Indian population that contain healthy and unhealthy CxRs of clinically confirmed patients of tuberculosis, chronic obstructive pulmonary disease, interstitial lung disease, pleural effusion, and lung cancer. The proposed method is tested on 688 images of our Indian CxR dataset including images with severe abnormal findings to validate its robustness. We also experimented on commonly used benchmark datasets such as Japanese Society of Radiological Technology; Montgomery County, and Shenzhen, China for state-of-the-art comparison. The performance of our method is tested against techniques described in the literature and achieved the superior performance.

RU-Net for Heart Segmentation from CXR

Cardiovascular disease is one of the top causes of death in the world. In order to release heavy workload for doctor, automated segmentation methods using deep learning are proposed by researchers. Due to limitation of medical images, we proposed a novel model RU-Net based on the combination of U-Net and Residual Network for heart segmentation. We replaced Res path from direct skip connection from encoder to decoder. We use Jaccard similarity coefficient to compare the result of our method and U-Net with public dataset called Japanese Society of Radiological Technology (JSRT). The experiment result demonstrates the accuracy of our method.

View Identification Assisted Fully Convolutional Network for Lung Field Segmentation of Frontal and Lateral Chest Radiographs

Locating lung field is a critical and fundamental processing stage in the automated analysis of chest radiographs (CXRs) for pulmonary disorders. During the routine examination of CXRs, using both frontal and lateral CXRs can benefit clinical diagnosis of cardiothoracic and lung diseases. However, the accurate segmentation of lung fields on both frontal and lateral CXRs is still challenging due to the blurry boundary of the lung field on lateral CXRs and the poor generalization ability of the models. Existing deep learning-based methods focused on lung field segmentation on frontal CXRs, and the generalization ability of these methods on the different type of CXRs (e.g., pediatric CXRs) and new lung diseases (e.g., COVID-19) has not been tested. In this paper, a view identification assisted fully convolutional network (VI-FCN) is proposed for the segmentation of lung fields on frontal and lateral CXRs simultaneously. The VI-FCN consists of an FCN branch for lung field segmentation and a view identification branch for identification of the frontal and lateral CXRs and for enhancing the lung field segmentation. To improve the generalization ability of VI-FCN, six public datasets and our frontal and lateral CXRs (over 2000 CXRs) were collected for training. The segmentation of lung fields on the Japanese Society of Radiological Technology (JSRT) dataset yields mean dice similarity coefficient (DSC) of 0.979 ± 0.008, mean Jaccard index ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\Omega $ </tex-math></inline-formula> ) of 0.959 ± 0.016, and mean boundary distance (MBD) of 1.023 ± 0.487 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mm</i> . Besides, the VI-FCN achieves mean DSC of 0.973 ± 0.010, mean <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\Omega $ </tex-math></inline-formula> of 0.947 ± 0.018, and mean MBD of 1.923 ± 0.755 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mm</i> for the segmentation of lung fields on our lateral CXRs. The experiments demonstrate the superior performance of the proposed VI-FCN over most of existing state-of-the-art methods. Moreover, the proposed VI-FCN achieves promising results on untrained pediatric CXRs and COVID-19 datasets.

Result of Online Survey on Gonadal Shielding in Pediatric Hip and Pelvic Radiography: Task Group of Gonadal Shielding in Pediatric Hip and Pelvic Radiography, Radiation Protection Division, Japanese Society of Radiological Technology

Result of Online Survey on Gonadal Shielding in Pediatric Hip and Pelvic Radiography: Task Group of Gonadal Shielding in Pediatric Hip and Pelvic Radiography, Radiation Protection Division, Japanese Society of Radiological Technology

Current state of oncologic 18F-FDG PET/CT in Japan: A nationwide survey.

Combined positron emission tomography/computed tomography (PET/CT) has gradually advanced with the introduction of newly developed techniques. However, the recent status of imaging techniques (e.g., scanning range, availability of correction methods, and decisions on performing delayed scan) in oncologic PET/CT with 18F-fluorodeoxyglucose (18F-FDG) in Japan is unclear. We conducted a nationwide cross-sectional survey to document 18F-FDG PET/CT protocols and clarify the recent status of imaging techniques for oncologic 18F-FDG PET/CT in Japan. We conducted a web survey hosted by the Japanese Society of Radiological Technology between October and December 2017. The questionnaire included nine items on the demographics of the respondents, their scan protocols, and additional imaging to their routine protocols. We received responses from 119 Japanese technologists who performed 18F-FDG PET/CT in practice. Almost all the respondents stated that the scanning range was from the top of the head to the pelvis or mid-thigh region. Newly developed techniques were used by fewer than half of the respondents. Most respondents performed additional imaging in consultation with physicians, such as delayed imaging (83%) or an extended scanning range for early imaging (55%). Our survey helps in clarifying the recent state of oncologic 18F-FDG PET/CT imaging techniques in Japan. Given that 18F-FDG PET/CT practices most frequently performed additional imaging along with their routine scan protocol, the practice constitutes the most varied examination performed in Japanese nuclear medicine.

Lung segmentation on chest X‐ray images in patients with severe abnormal findings using deep learning

AbstractSeveral studies have evaluated the usefulness of deep learning for lung segmentation using chest X‐ray (CXR) images with small‐ or medium‐sized abnormal findings. Here, we built a database including both CXR images with severe abnormalities and experts' lung segmentation results, and aimed to evaluate our network's efficacy in lung segmentation from these images. For lung segmentation, CXR images from the Japanese Society of Radiological Technology (JSRT, N = 247) and Montgomery databases (N = 138) were included, and 65 additional images depicting severe abnormalities from a public database were evaluated and annotated by a radiologist, thereby adding lung segmentation results to these images. U‐net was used to segment the lungs in images from the three databases. Subsequently, the U‐net network architecture was automatically optimized for lung segmentation from CXR images using Bayesian optimization. Dice similarity coefficient (DSC) was calculated to confirm segmentation. Our results demonstrated that using baseline U‐net yielded poorer lung segmentation results in our database than those in the JSRT and Montgomery databases, implying that robust segmentation of lungs may be difficult because of severe abnormalities. The DSC values with baseline U‐net for the JSRT, Montgomery and our databases were 0.979, 0.941, and 0.889, respectively, and with optimized U‐net, 0.976, 0.973, and 0.932, respectively. For robust lung segmentation, the U‐net architecture was optimized via Bayesian optimization, and our results demonstrate that the optimized U‐net was more robust than baseline U‐net in lung segmentation from CXR images with large‐sized abnormalities.

The detection of lung cancer using massive artificial neural network based on soft tissue technique

BackgroundA proposed computer aided detection (CAD) scheme faces major issues during subtle nodule recognition. However, radiologists have not noticed subtle nodules in beginning stage of lung cancer while a proposed CAD scheme recognizes non subtle nodules using x-ray images.MethodSuch an issue has been resolved by creating MANN (Massive Artificial Neural Network) based soft tissue technique from the lung segmented x-ray image. A soft tissue image recognizes nodule candidate for feature extortion and classification. X-ray images are downloaded using Japanese society of radiological technology (JSRT) image set. This image set includes 233 images (140 nodule x-ray images and 93 normal x-ray images). A mean size for a nodule is 17.8 mm and it is validated with computed tomography (CT) image. Thirty percent (42/140) abnormal represents subtle nodules and it is split into five stages (tremendously subtle, very subtle, subtle, observable, relatively observable) by radiologists.ResultA proposed CAD scheme without soft tissue technique attained 66.42% (93/140) sensitivity and 66.76% accuracy having 2.5 false positives per image. Utilizing soft tissue technique, many nodules superimposed by ribs as well as clavicles have identified (sensitivity is 72.85% (102/140) and accuracy is 72.96% at one false positive rate).ConclusionIn particular, a proposed CAD system determine sensitivity and accuracy in support of subtle nodules (sensitivity is 14/42 = 33.33% and accuracy is 33.66%) is statistically higher than CAD (sensitivity is 13/42 = 30.95% and accuracy is 30.97%) scheme without soft tissue technique. A proposed CAD scheme attained tremendously minimum false positive rate and it is a promising technique in support of cancerous recognition due to improved sensitivity and specificity.