Identification of high-risk population of pneumoconiosis using deep learning segmentation of lung 3D images and radiomics texture analysis.

OBJECTION: The aim of this study is to develop an early-warning model for identifying high-risk populations of pneumoconiosis by combining lung 3D images and radiomics lung texture features. METHODS: A retrospective study was conducted, including 600 dust-exposed workers and 300 confirmed pneumoconiosis patients. Chest computed tomography (CT) images were divided into a training set and a test set in a 2:1 ratio. Whole-lung segmentation was performed using deep learning models for feature extraction of radiomics. Two feature selection algorithms and five classification models were used. The optimal model was selected using a 10-fold cross-validation strategy, and the calibration curve and decision curve were evaluated. To verify the applicability of the model, the diagnostic efficiency and accuracy between the model and human interpretation were compared. Additionally, the risk probabilities for different risk groups defined by the model were compared at different time intervals. RESULTS: Four radiomics features were ultimately used to construct the predictive model. The logistic regression model was the most stable in both the training set and testing set, with an area under curve (AUC) of 0.964 (95 % confidence interval [CI], 0.950-0.976) and 0.947 (95 %CI, 0.925-0.964). In the training and testing sets, the Brier scores were 0.092 and 0.14, respectively, with threshold probability ranges of 2 %-99 % and 2 %-85 %. These results indicate that the model exhibits good calibration and clinical benefit. The comparison between the model and human interpretation showed that the model was not inferior in terms of diagnostic efficiency and accuracy. Additionally, the high-risk population identified by the model was diagnosed as pneumoconiosis two years later. CONCLUSION: This study provides a meticulous and quantifiable method for detecting and assessing the risk of pneumoconiosis, building upon accurate diagnosis. Employing risk scoring and probability estimation, not only enhances the efficiency of diagnostic physicians but also provides a valuable reference for controlling the occurrence of pneumoconiosis.

Assessment of the Respiratory Health of Working US Coal Miners Since 2014-Radiography, Spirometry, and Symptom Assessments.

OBJECTIVE: The aim of the study is to summarize Coal Workers' Health Surveillance Program findings since 2014, focusing on prevalence of radiographic pneumoconiosis and abnormal spirometry by region. METHODS: Analysis included the most recent Coal Workers' Health Surveillance Program encounter for working miners during October 1, 2014-June 30, 2022. Central Appalachia consists of Kentucky, Virginia, and West Virginia. RESULTS: Pneumoconiosis prevalence was highest among underground miners, with 318 (6.2%) having radiographic evidence of disease. Central Appalachian miners were more likely to have evidence of pneumoconiosis (relative risk = 4.1 [3.4-5.0]) and abnormal spirometry (relative risk = 1.4 [1.2-1.6]) compared with all others. CONCLUSIONS: Central Appalachia is a hotspot for pneumoconiosis, as well as for other indicators of respiratory impairment in coal miners. Outreach for occupational respiratory health surveillance should focus on those areas most heavily affected, ensuring that miners are not hindered by perceived or actual barriers to this secondary intervention.

Characterizing Lung Particulates Using Quantitative Microscopy in Coal Miners With Severe Pneumoconiosis.

CONTEXT.­: Current approaches for characterizing retained lung dust using pathologists' qualitative assessment or scanning electron microscopy with energy-dispersive spectroscopy (SEM/EDS) have limitations. OBJECTIVE.­: To explore polarized light microscopy coupled with image-processing software, termed quantitative microscopy-particulate matter (QM-PM), as a tool to characterize in situ dust in lung tissue of US coal miners with progressive massive fibrosis. DESIGN.­: We developed a standardized protocol using microscopy images to characterize the in situ burden of birefringent crystalline silica/silicate particles (mineral density) and carbonaceous particles (pigment fraction). Mineral density and pigment fraction were compared with pathologists' qualitative assessments and SEM/EDS analyses. Particle features were compared between historical (born before 1930) and contemporary coal miners, who likely had different exposures following changes in mining technology. RESULTS.­: Lung tissue samples from 85 coal miners (62 historical and 23 contemporary) and 10 healthy controls were analyzed using QM-PM. Mineral density and pigment fraction measurements with QM-PM were comparable to consensus pathologists' scoring and SEM/EDS analyses. Contemporary miners had greater mineral density than historical miners (186 456 versus 63 727/mm3; P = .02) and controls (4542/mm3), consistent with higher amounts of silica/silicate dust. Contemporary and historical miners had similar particle sizes (median area, 1.00 versus 1.14 µm2; P = .46) and birefringence under polarized light (median grayscale brightness: 80.9 versus 87.6; P = .29). CONCLUSIONS.­: QM-PM reliably characterizes in situ silica/silicate and carbonaceous particles in a reproducible, automated, accessible, and time/cost/labor-efficient manner, and shows promise as a tool for understanding occupational lung pathology and targeting exposure controls.

[Research advances in imaging techniques related to the diagnosis of pneumoconiosis].

Occupational pneumoconiosis (hereinafter referred to as pneumoconiosis) is the most harmful occupational disease in China. According to the current standard GBZ 70-2015 Diagnosis of Occupational Pneumoconiosis, pneumoconiosis is mainly diagnosed and staged by high kilovolt or digital radiography. Chest radiography in pneumoconiosis is the most widely studied and mature imaging technique in the diagnosis of pneumoconiosis. However, this technique has some limitations in the screening of some early pneumoconiosis and occupational health examination, and there is a certain risk of missed diagnosis and misdiagnosis. With the continuous development of imaging examination technology, computed tomography, magnetic resonance imaging, positron emission tomography-computed tomography and artificial intelligence technology as auxiliary imaging examination methods have shown different diagnostic values in the research of auxiliary diagnosis of pneumoconiosis. This paper summarizes the advantages and problems in the application of various kinds of imaging techniques, which provides a direction for the future research of imaging techniques related to the diagnosis of pneumoconiosis.

Fluoro-edenite non-neoplastic diseases in Biancavilla (Sicily, Italy): pleural plaques and/or pneumoconiosis?

BACKGROUND: A mesothelioma cluster in Biancavilla (Sicily, Italy), drew attention to fluoro-edenite, a fibre classified by International Agency for Research on Cancer as carcinogenic to humans. Significant excesses in mortality and morbidity were observed for respiratory diseases and a significant excess of pneumoconiosis hospitalizations was reported. OBJECTIVE: Aim of this study is to assess the characters of the lung damage in Biancavilla residents hospitalized with pneumoconiosis or asbestosis diagnoses. METHODOLOGY: Medical records, available radiographs and computed tomography scans were collected. The obtained imaging was reviewed by a panel of three specialists and focused on pleural and parenchymal abnormalities. Cases with an ILO-BIT or ICOERD score equal or greater than 2 were considered positive for a pneumoconiosis-like damage, cases with a score lower than 2 or insufficient quality of imaging were considered inconclusive. If no pneumoconiotic aspects were present the cases were classified as negative. RESULTS: Out of 38 cases, diagnostic imaging for 25 cases were found. Ten cases out of 25 showed asbestosis-like features, nine subjects were considered negative. In six patients' results were inconclusive. CONCLUSIONS: Asbestosis-like features were substantiated in Biancavilla residents without known occupational exposure to asbestos. Further studies to estimate population respiratory health are required. Experimental studies on the fibrogenic potential of fluoro-edenite are needed.

Tetrandrine slows the radiographic progression of progressive massive fibrosis in pneumoconiosis: a retrospective cohort study.

OBJECTIVE: This study aims to explore the clinical effect of Tetrandrine (Tet) on progressive massive fibrosis (PMF) of pneumoconiosis. METHODS: This retrospective study collected 344 pneumoconiosis patients with PMF, and 127 were eligible for the final analysis, including 57 patients in the Tet group and 70 patients in the control group. The progress of imaging and lung function were compared between the two groups. RESULTS: After 13 months (median) of treatment, the size of PMF was smaller in the Tet group than that in the control group (1526 vs. 2306, p=0.001), and the size was stable in the Tet group (1568 vs. 1526, p= 0.381), while progressed significantly in the control group (2055 vs. 2306, p=0.000). The small nodule profusion and emphysema were also milder than that in the control group (6.0 vs. 7.5, p=0.046 and 8.0 vs. 12, p=0.016 respectively). Pulmonary ventilation function parameters FVC and FEV1 improved in the Tet group (3222 vs. 3301, p=0.021; 2202 vs. 2259, p=0.025 respectively) and decreased in the control group (3272 vs. 3185, p= 0.00; 2094 vs. 1981, p=0.00 respectively). FEV1/FVC was also significantly higher in the Tet group than that in the control group (68.45vs. 60.74, p=0.001). However, similar result was failed to observed for DLco%, which showed a significant decrease in both groups. CONCLUSION: Tet has shown great potential in the treatment of PMF by slowing the progression of pulmonary fibrosis and the decline of lung function.

Artificial Intelligence in Quantitative Chest Imaging Analysis for Occupational Lung Disease.

Occupational lung disease manifests complex radiologic findings which have long been a challenge for computer-assisted diagnosis (CAD). This journey started in the 1970s when texture analysis was developed and applied to diffuse lung disease. Pneumoconiosis appears on radiography as a combination of small opacities, large opacities, and pleural shadows. The International Labor Organization International Classification of Radiograph of Pneumoconioses has been the main tool used to describe pneumoconioses and is an ideal system that can be adapted for CAD using artificial intelligence (AI). AI includes machine learning which utilizes deep learning or an artificial neural network. This in turn includes a convolutional neural network. The tasks of CAD are systematically described as classification, detection, and segmentation of the target lesions. Alex-net, VGG16, and U-Net are among the most common algorithms used in the development of systems for the diagnosis of diffuse lung disease, including occupational lung disease. We describe the long journey in the pursuit of CAD of pneumoconioses including our recent proposal of a new expert system.

[Application of a light-weighted convolutional neural network for automatic recognition of coal workers' pneumoconiosis in the early stage].

Objective: To construct and verify a light-weighted convolutional neural network (CNN), and explore its application value for screening the early stage (subcategory 0/1 and stage Ⅰ of pneumoconiosis) of coal workers' pneumoconiosis (CWP) from digital chest radiography (DR) . Methods: A total of 1225 DR images of coal workers who were examined at an Occupational Disease Prevention and Control Institute in Anhui Province from October 2018 to March 2021 were retrospectively collected. All DR images were collectively diagnosed by 3 radiologists with diagnostic qualifications and gave diagnostic results. There were 692 DR images with small opacity profusion 0/- or 0/0 and 533 DR images with small opacity profusion 0/1 to stage Ⅲ of pneumoconiosis. The original chest radiographs were preprocessed differently to generate four datasets, namely 16-bit grayscale original image set (Origin16), 8-bit grayscale original image set (Origin 8), 16-bit grayscale histogram equalized image set (HE16) and 8-bit grayscale histogram equalized image set (HE8). The light-weighted CNN, ShuffleNet, was applied to train the generated prediction model on the four datasets separately. The performance of the four models for pneumoconiosis prediction was evaluated on a test set containing 130 DR images using measures such as the receiver operating characteristic (ROC) curve, accuracy, sensitivity, specificity, and Youden index. The Kappa consistency test was used to compare the agreement between the model predictions and the physician diagnosed pneumoconiosis results. Results: Origin16 model achieved the highest ROC area under the curve (AUC=0.958), accuracy (92.3%), specificity (92.9%), and Youden index (0.8452) for predicting pneumoconiosis, with a sensitivity of 91.7%. And the highest consistency between identification and physician diagnosis was observed for Origin16 model (Kappa value was 0.845, 95%CI: 0.753-0.937, P<0.001). HE16 model had the highest sensitivity (98.3%) . Conclusion: The light-weighted CNN ShuffleNet model can efficiently identify the early stages of CWP, and its application in the early screening of CWP can effectively improve physicians' work efficiency.

Serum Bilirubin Levels and Disease Severity in Patients with Pneumoconiosis.

Aim: To investigate the association between serum bilirubin and disease severity in patients with pneumoconiosis. Methods: The study comprised 45 patients with pneumoconiosis retrospectively; all pneumoconiosis patients were classified into I, II, and III stage according to the radiological severity. Results: Serum direct bilirubin levels were significantly lower in III stage pneumoconiosis patients than those in I/II stage (p = 0.012) but not serum indirect bilirubin. Serum direct bilirubin was negatively correlated with radiological severity in patients with pneumoconiosis (r = -0.320; p = 0.032); by multiple linear-regression analysis, we observed that serum direct bilirubin levels had independent association with radiological severity in patients with pneumoconiosis (beta = -0.459; p = 0.005). Conclusions: Serum direct bilirubin levels are negatively associated with disease severity in patients with pneumoconiosis.

Imaging diagnosis of pneumoconiosis with predominant nodular pattern: HRCT and pathologic findings.

The radiological patterns of known pneumoconiosis have been changing in recent years. The basic pathology in pneumoconiosis is the presence of dust macules, mixed dust fibrosis, nodules, diffuse interstitial fibrosis, and progressive massive fibrosis. These pathologic changes can coexist in dust-exposed workers. High resolution CT reflects pathological findings in pneumoconiosis and is useful for the diagnosis. Pneumoconiosis such as silicosis, coal workers' pneumoconiosis, graphite pneumoconiosis, and welder's pneumoconiosis, has predominant nodular HRCT pattern. Diffuse interstitial pulmonary fibrosis is sometimes found in the lungs of this pneumoconiosis. In the early stages of metal lung, such as aluminosis and hard metal lung, centrilobular nodules are predominant findings, and in the advanced stages, reticular opacities are predominant findings. The clinician must understand the spectrum of expected imaging patterns related to known dust exposures and novel exposures. In this article, HRCT and pathologic findings of pneumoconiosis with predominant nodular opacities are shown.

Perceptual and objective physical quality of chest images: a comparison between digital radiographic chest images processed for cancer screening and pneumoconiosis screening in Japan.

This study (1) evaluated the perceptual and objective physical quality of digital radiographic chest images processed for different purposes (routine hospital use, lung cancer screening, and pneumoconiosis screening), and (2) quantified objectively the quality of chest images visually graded by the Japan National Federation of Industrial Health Organization (ZENEIREN). Four observers rated the images using a visual grading score (VGS) according to ZENEIREN's quality criteria. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were measured. Between groups, differences were assessed using ANOVA (followed by Bonferroni multiple comparisons) or unpaired t-test. The Pearson's correlation coefficients were calculated for the correlation between perceptual quality and objective physical image quality. The image quality perceived by the observers and the SNR measurements were highest for the images generated using parameters recommended for lung cancer screening. The images processed for pneumoconiosis screening were rated poorest by the observers and showed the lowest objective physical quality measurements. The chest images rated high quality by ZENEIREN generally showed a higher objective physical image quality. The SNR correlated well with VGS, but CNR did not. Highly significant differences between the processing parameters indicate that image processing strongly influences the perceptual quality of digital radiographic chest images.

Application of a light-weighted convolutional neural network for automatic recognition of coal workers' pneumoconiosis in the early stage / 中华劳动卫生职业病杂志

Objective: To construct and verify a light-weighted convolutional neural network (CNN), and explore its application value for screening the early stage (subcategory 0/1 and stage Ⅰ of pneumoconiosis) of coal workers' pneumoconiosis (CWP) from digital chest radiography (DR) . Methods: A total of 1225 DR images of coal workers who were examined at an Occupational Disease Prevention and Control Institute in Anhui Province from October 2018 to March 2021 were retrospectively collected. All DR images were collectively diagnosed by 3 radiologists with diagnostic qualifications and gave diagnostic results. There were 692 DR images with small opacity profusion 0/- or 0/0 and 533 DR images with small opacity profusion 0/1 to stage Ⅲ of pneumoconiosis. The original chest radiographs were preprocessed differently to generate four datasets, namely 16-bit grayscale original image set (Origin16), 8-bit grayscale original image set (Origin 8), 16-bit grayscale histogram equalized image set (HE16) and 8-bit grayscale histogram equalized image set (HE8). The light-weighted CNN, ShuffleNet, was applied to train the generated prediction model on the four datasets separately. The performance of the four models for pneumoconiosis prediction was evaluated on a test set containing 130 DR images using measures such as the receiver operating characteristic (ROC) curve, accuracy, sensitivity, specificity, and Youden index. The Kappa consistency test was used to compare the agreement between the model predictions and the physician diagnosed pneumoconiosis results. Results: Origin16 model achieved the highest ROC area under the curve (AUC=0.958), accuracy (92.3%), specificity (92.9%), and Youden index (0.8452) for predicting pneumoconiosis, with a sensitivity of 91.7%. And the highest consistency between identification and physician diagnosis was observed for Origin16 model (Kappa value was 0.845, 95%CI: 0.753-0.937, P<0.001). HE16 model had the highest sensitivity (98.3%) . Conclusion: The light-weighted CNN ShuffleNet model can efficiently identify the early stages of CWP, and its application in the early screening of CWP can effectively improve physicians' work efficiency.

[Study on optimization of image processing parameters of pneumoconiosis by DR].

Objective: To explore the effect of different post-processing parameters of digital radiography (DR) on the quality of chest X-ray for pneumoconiosis diagnosis, and to provide suggestions on parameter setting suitable for this kind of DR machine. Methods: From January 1, 2022 to June 30, 2022, the chest films of 35 workers in the department of radiology of Hangzhou occupational disease prevention and treatment hospital were randomly selected and printed after setting different image post-processing parameters. The quality of chest film was evaluated by the measurement of optical densitometer and the combination of subjective and objective by professional physicians. Results: When the density is set to 2 and the contrast/detail contrast is 4.5, the optical density of each area of DR chest film meets the requirements of chest X-ray quality, and the qualified rate of physician quality evaluation is the highest. Conclusion: Reasonable setting of image post-processing parameters can improve the quality of chest radiograph.

[Application of artificial intelligence in digital chest radiography diagnosis of pneumoconiosis].

Pneumoconiosis is the occupational disease with the highest burden in China currently. The diagnosis of pneumoconiosis mainly relies on manual reading of X-ray high-kilovoltage or digital photography chest radiograph, which has some problems such as low efficiency, strong subjectivity, and cannot accurately judge the critical lesions. With the progress of machine-aided diagnosis technology, the efficient, objective and quantitative of artificial intelligence diagnosis technology just solve the shortcomings above. This paper reviews the research progress in digital chest radiography diagnosis of pneumoconiosis using artificial intelligence technology, especially deep learning model, combined with the limitations of conventional manual reading, in order to clarify the application prospect of artificial intelligence technology in the diagnosis of pneumoconiosis by digital chest radiography, and provide a direction for future research in this field.

Tracheobronchial variations in Pneumoconiosis cases: multidetector computed tomography diagnosis.

Background: Tracheobronchial variations (TBVs) are more common than previously believed due to the increasing use of multi-detector computed tomography (MDCT). This study aimed to assess TBVs in cases of pneumoconiosis, one of the oldest occupational diseases that still poses a threat to public health. Methods: This was a descriptive study that involved reviewing chest MDCT images of 34 cases of pneumoconiosis and 34 control cases retrospectively from January 2020 to April 2022. Variations in the trachea, right main bronchus, left main bronchus, lobar and segmental branches of the cases in the patient and control groups were evaluated according to Boyden's nomenclature. Results: The frequency of TBV was 32.4% in pneumoconiosis cases. Although the frequency of TBV was higher in the patient group than in the control group, the difference was not statistically significant (p=0.086). Furthermore, there was no significant difference in terms of TBV classification between the patient and control groups (p=0.407). Additionally, the presence of TBV did not affect the distribution of International Labour Organization categories in pneumoconiosis cases (p=0.360). Conclusions: Although our study provides initial insights into the occurrence of TBVs in pneumoconiosis cases, further research is needed to clarify the relationship between these variations and the disease.

Transformer-based factorized encoder for classification of pneumoconiosis on 3D CT images.

In the past decade, deep learning methods have been implemented in the medical image fields and have achieved good performance. Recently, deep learning algorithms have been successful in the evaluation of diagnosis on lung images. Although chest radiography (CR) is the standard data modality for diagnosing pneumoconiosis, computed tomography (CT) typically provides more details of the lesions in the lung. Thus, a transformer-based factorized encoder (TBFE) was proposed and first applied for the classification of pneumoconiosis depicted on 3D CT images. Specifically, a factorized encoder consists of two transformer encoders. The first transformer encoder enables the interaction of intra-slice by encoding feature maps from the same slice of CT. The second transformer encoder explores the inter-slice interaction by encoding feature maps from different slices. In addition, the lack of grading standards on CT for labeling the pneumoconiosis lesions. Thus, an acknowledged CR-based grading system was applied to mark the corresponding pneumoconiosis CT stage. Then, we pre-trained the 3D convolutional autoencoder on the public LIDC-IDRI dataset and fixed the parameters of the last convolutional layer of the encoder to extract CT feature maps with underlying spatial structural information from our 3D CT dataset. Experimental results demonstrated the superiority of the TBFE over other 3D-CNN networks, achieving an accuracy of 97.06%, a recall of 89.33%, precision of 90%, and an F1-score of 93.33%, using 10-fold cross-validation.

Postexposure progression of pneumoconiosis among former Appalachian coal miners.

BACKGROUND: The prevalence of pneumoconiosis among working United States underground coal miners has been increasing for the past two decades, with the highest rates of disease observed among miners in the central Appalachian states of Kentucky, Virginia, and West Virginia. Surveillance for this disease in the United States focuses on working coal miners, who continue to be occupationally exposed to dust. This study examines the radiographic evidence for postexposure progression of pneumoconiosis in a population of former coal miners no longer occupationally exposed to coal mine dust who were seen at a community radiology clinic in eastern Kentucky. METHODS: Data were obtained and analyzed from clinical records of former coal miners who had a clinic encounter during January 1, 2017-August 1, 2019, a recorded final year of employment, and ≥2 postemployment digital chest radiographs. Radiographs were classified according to the International Labour Office guidelines by at least two B Readers. A final summary pneumoconiosis severity score (range, 0-13), accounting for both small and large opacities, was assigned to each chest radiograph. Progression was defined as an increase in severity score between a miner's radiographs over time. RESULTS: Data for 130 former coal miners were analyzed. All miners were male and most (n = 114, 88%) had worked primarily in Kentucky. Information on race/ethnicity was not available. The most common job types were roof bolters (n = 51, 39%) and continuous miner operators (n = 46, 35%). Forty-one (31.5%) miners had evidence of radiographic disease progression after leaving the workforce, with a median of 3.6 years between first and latest postretirement radiograph. A total of 80 (62%) miners had evidence of pneumoconiosis on their latest radiograph, and two-thirds (n = 53) of these were classified as progressive massive fibrosis (PMF), the most severe form of the disease. CONCLUSIONS: Postexposure progression can occur in former coal miners, emphasizing the potential benefits of continued radiographic follow-up postemployment. In addition to participating in disease screening throughout their careers to detect pneumoconiosis early and facilitate intervention, radiographic follow-up of former coal miners can identify new or progressive radiographic findings even after workplace exposure to respirable coal mine dust ends. Identification of progressive pneumoconiosis in former miners has potential implications for clinical management and eligibility for disability compensation.

Detection and Visualisation of Pneumoconiosis Using an Ensemble of Multi-Dimensional Deep Features Learned from Chest X-rays.

Pneumoconiosis is a group of occupational lung diseases induced by mineral dust inhalation and subsequent lung tissue reactions. It can eventually cause irreparable lung damage, as well as gradual and permanent physical impairments. It has affected millions of workers in hazardous industries throughout the world, and it is a leading cause of occupational death. It is difficult to diagnose early pneumoconiosis because of the low sensitivity of chest radiographs, the wide variation in interpretation between and among readers, and the scarcity of B-readers, which all add to the difficulty in diagnosing these occupational illnesses. In recent years, deep machine learning algorithms have been extremely successful at classifying and localising abnormality of medical images. In this study, we proposed an ensemble learning approach to improve pneumoconiosis detection in chest X-rays (CXRs) using nine machine learning classifiers and multi-dimensional deep features extracted using CheXNet-121 architecture. There were eight evaluation metrics utilised for each high-level feature set of the associated cross-validation datasets in order to compare the ensemble performance and state-of-the-art techniques from the literature that used the same cross-validation datasets. It is observed that integrated ensemble learning exhibits promising results (92.68% accuracy, 85.66% Matthews correlation coefficient (MCC), and 0.9302 area under the precision-recall (PR) curve), compared to individual CheXNet-121 and other state-of-the-art techniques. Finally, Grad-CAM was used to visualise the learned behaviour of individual dense blocks within CheXNet-121 and their ensembles into three-color channels of CXRs. We compared the Grad-CAM-indicated ROI to the ground-truth ROI using the intersection of the union (IOU) and average-precision (AP) values for each classifier and their ensemble. Through the visualisation of the Grad-CAM within the blue channel, the average IOU passed more than 90% of the pneumoconiosis detection in chest radiographs.

Differentiation and prediction of pneumoconiosis stage by computed tomography texture analysis based on U-Net neural network.

BACKGROUND AND OBJECTIVE: The progressive worsening of pneumoconiosis will ensue a hazardous physical condition in patients. This study details the differential diagnosis of the pneumoconiosis stage, by employing computed tomography (CT) texture analysis, based on U-Net neural network. METHODS: The pneumoconiosis location from 92 patients at various stages was extracted by U-Net neural network. Mazda software was employed to analyze the texture features. Three dimensionality reduction methods set the best texture parameters. We applied four methods of the B11 module to analyze the selected texture parameters and calculate the misclassified rate (MCR). Finally, the receiver operating characteristic curve (ROC) of the texture parameters was analyzed, and the texture parameters with diagnostic efficiency were evaluated by calculating the area under curve (AUC). RESULTS: The original film was processed by Gaussian and Laplace filters for a better display of the segmented area of pneumoconiosis in all stages. The MCR value obtained by the NDA analysis method under the MI dimension reduction method was the lowest, at 10.87%. In the filtered texture feature parameters, the best AUC was 0.821. CONCLUSIONS: CT texture analysis based on the U-Net neural network can be used to identify the staging of pneumoconiosis.