Calculating the target exposure index using a deep convolutional neural network and a rule base

Development and clinical evaluation of an AI-assisted respiratory state classification system for chest X-rays: A BMI-Specific approach

PURPOSE

In this study, we aimed to develop and clinically evaluate an artificial intelligence (AI)-assisted support system for determining inhalation and exhalation states on chest X-ray images, focusing on the specific challenge of respiratory state determination.

METHODS

We developed a body mass index (BMI)-specific approach for respiratory state classification in chest X-rays using separate models for normal and obesity groups. Feature extraction was performed using four pre-trained networks (EfficientNet B0, GoogleNet, Xception, and VGG16) combined with Naive Bayes classification. A database of 3200 chest X-ray images from 1600 patients, labeled for respiratory states using temporal subtraction techniques, was utilized. The system's clinical utility was assessed through an observational study involving eight radiological technologists with varying experience levels.

RESULTS

The approach combining EfficientNet B0 late-layer with Naive Bayes classification and GoogleNet's end-to-end model demonstrated the highest performance. The support system significantly improved the area under the curve from 0.728 to 0.796 in the normal BMI group and from 0.752 to 0.817 in the obesity group (p < 0.05), showing particular effectiveness in classifying exhalation states in obese patients.

CONCLUSION

The developed AI-assisted support system enhances radiological technologists' ability to determine respiratory states across varying levels of experience, particularly in challenging cases involving obese patients. This system contributes to improving image quality assessment and workflow efficiency by potentially reducing unnecessary re-imaging.

Monitoring clinical exposure index and deviation index for dose optimization based on national diagnostic reference level: Focusing on general radiography of extremities

BACKGROUND

The International Electrotechnical Commission established the concept of the exposure index (EI), target exposure index (EIT) and deviation index (DI). Some studies have conducted to utilize the EI as a patient dose monitoring tool in the digital radiography (DR) system.

OBJECTIVE

To establish the appropriate clinical EIT, this study aims to introduce the diagnostic reference level (DRL) for general radiography and confirm the usefulness of clinical EI and DI.

METHODS

The relationship between entrance surface dose (ESD) and clinical EI is obtained by exposure under the national radiography conditions of Korea for 7 extremity examinations. The EI value when the ESD is the DRL is set as the clinical EIT, and the change of DI is then checked.

RESULTS

The clinical EI has proportional relationship with ESD and is affected by the beam quality. When the clinical EIT is not adjusted according to the revision of DRLs, there is a difference of up to 2.03 in the DI value and may cause an evaluation error of up to 1.6 times for patient dose.

CONCLUSIONS

If the clinical EIT is periodically managed according to the environment of medical institution, the appropriate patient dose and image exposure can be managed based on the clinical EI, EIT, and DI.

Deep learning for chest X-ray analysis: A survey

Recent advances in deep learning have led to a promising performance in many medical image analysis tasks. As the most commonly performed radiological exam, chest radiographs are a particularly important modality for which a variety of applications have been researched. The release of multiple, large, publicly available chest X-ray datasets in recent years has encouraged research interest and boosted the number of publications. In this paper, we review all studies using deep learning on chest radiographs published before March 2021, categorizing works by task: image-level prediction (classification and regression), segmentation, localization, image generation and domain adaptation. Detailed descriptions of all publicly available datasets are included and commercial systems in the field are described. A comprehensive discussion of the current state of the art is provided, including caveats on the use of public datasets, the requirements of clinically useful systems and gaps in the current literature.

Proper Management of the Clinical Exposure Index Based on Body Thickness Using Dose Optimization Tools in Digital Chest Radiography: A Phantom Study

In radiography, the exposure index (EI), as per the International Electrotechnical Commission standard, depends on the incident beam quality and exposure dose to the digital radiography system. Today automatic exposure control (AEC) systems are commonly employed to obtain the optimal image quality. An AEC system can maintain a constant incident exposure dose on the image receptor regardless of the patient thickness. In this study, we investigated the relationship between body thickness, entrance surface dose (ESD), EI, and the exposure indicator (S value) with the aim of using EI as the dose optimization tool in digital chest radiography (posterior-anterior and lateral projection). The exposure condition from the Korean national survey for determining diagnostic reference levels and two digital radiography systems (photostimulable phosphor plate and indirect flat panel detector) were used. As a result, ESD increased as the phantom became thicker with constant exposure indicator, which indicates similar settings to an AEC system, but the EI indicated comparatively constant values without following the tendency of ESD. Therefore, body thickness should be considered under the AEC system for introducing EI as the dose optimization tool in digital chest radiography.

USE OF CLINICAL EXPOSURE INDEX AND DEVIATION INDEX BASED ON NATIONAL DIAGNOSTIC REFERENCE LEVEL AS DOSE-OPTIMIZATION TOOLS FOR GENERAL RADIOGRAPHY IN KOREA

The International Electrotechnical Commission introduced the concepts of exposure index (EI), target exposure index (EIT) and deviation index (DI) to manage and optimize patient dose in real time. In this study, we have proposed an appropriate method for setting the EIT based on the Korean national diagnostic reference levels (DRLs). Furthermore, we evaluated the use of clinical EI, EIT and DI as tools for patient dose optimization in clinical environments by observing the changes in DI with those in EIT. According to the Korean national exposure conditions, we conducted experiments on three representative radiographic examinations (chest posterior-anterior, lateral and abdomen anterior-posterior) of clinical environments. As the exposure conditions and DRLs varied, the clinical EI, EIT and DI also varied. These results reveal that the clinical EI, EIT and DI can be used as tools for optimizing the patient dose if EIT is periodically and properly updated.

Hybrid deep learning for detecting lung diseases from X-ray images

Lung disease is common throughout the world. These include chronic obstructive pulmonary disease, pneumonia, asthma, tuberculosis, fibrosis, etc. Timely diagnosis of lung disease is essential. Many image processing and machine learning models have been developed for this purpose. Different forms of existing deep learning techniques including convolutional neural network (CNN), vanilla neural network, visual geometry group based neural network (VGG), and capsule network are applied for lung disease prediction. The basic CNN has poor performance for rotated, tilted, or other abnormal image orientation. Therefore, we propose a new hybrid deep learning framework by combining VGG, data augmentation and spatial transformer network (STN) with CNN. This new hybrid method is termed here as VGG Data STN with CNN (VDSNet). As implementation tools, Jupyter Notebook, Tensorflow, and Keras are used. The new model is applied to NIH chest X-ray image dataset collected from Kaggle repository. Full and sample versions of the dataset are considered. For both full and sample datasets, VDSNet outperforms existing methods in terms of a number of metrics including precision, recall, F0.5 score and validation accuracy. For the case of full dataset, VDSNet exhibits a validation accuracy of 73%, while vanilla gray, vanilla RGB, hybrid CNN and VGG, and modified capsule network have accuracy values of 67.8%, 69%, 69.5% and 63.8%, respectively. When sample dataset rather than full dataset is used, VDSNet requires much lower training time at the expense of a slightly lower validation accuracy. Hence, the proposed VDSNet framework will simplify the detection of lung disease for experts as well as for doctors.