Feasibility study of deep-learning-based bone suppression incorporated with single-energy material decomposition technique in chest X-rays.

Automatic localization of pneumonia on chest X-rays (CXRs) is highly desirable both as an interpretive aid to the radiologist and for timely diagnosis of the disease. However, pneumonia's amorphous appearance on CXRs and complexity of normal anatomy in the chest present key challenges that hinder accurate localization. Existing studies in this area are either not optimized to preserve spatial information of abnormality or depend on expensive expert-annotated bounding boxes. We present a novel generative adversarial network (GAN)-based machine learning approach for this problem, which is weakly supervised (does not require any location annotations), was trained to retain spatial information, and can produce pixel-wise abnormality maps highlighting regions of abnormality (as opposed to bounding boxes around abnormality). Our method is based on the Wasserstein GAN framework and, to the best of our knowledge, the first application of GANs to this problem. Specifically, from an abnormal CXR as input, we generated the corresponding pseudo normal CXR image as output. The pseudo normal CXR is the "hypothetical" normal, if the same abnormal CXR were not to have any abnormalities. We surmise that the difference between the pseudo normal and the abnormal CXR highlights the pixels suspected to have pneumonia and hence is our output abnormality map. We trained our algorithm on an "unpaired" data set of abnormal and normal CXRs and did not require any location annotations such as bounding boxes/segmentations of abnormal regions. Furthermore, we incorporated additional prior knowledge/constraints into the model and showed that they help improve localization performance. We validated the model on a data set consisting of 14184 CXRs from the Radiological Society of North America pneumonia detectionchallenge. We evaluated our methods by comparing the generated abnormality maps with radiologist annotated bounding boxes using receiver operating characteristic (ROC) analysis, image similarity metrics such as normalized cross-correlation/mutual information, and abnormality detection rate.We also present visual examples of the abnormality maps, covering various scenarios of abnormality occurrence. Results demonstrate the ability to highlight regions of abnormality with the best method achieving an ROC area under the curve (AUC) of 0.77 and a detection rate of 85%.The GAN tended to perform better as prior knowledge/constraints were incorporated into themodel. We presented a novel GAN based approach for localizing pneumonia on CXRs that (1) does not require expensive hand annotated location ground truth; and (2) was trained to produce abnormality maps at the pixel level as opposed to bounding boxes. We demonstrated the efficacy of our methods via quantitative and qualitative results.

Introduction: Chest x-ray (CXR) abnormalities after cardiopulmonary resuscitation are common. Mechanisms by which these abnormalities develop are not fully elucidated, but aspiration of secretions and regurgitated gastric contents during prehospital airway management may be an important modifiable cause. Hypothesis: We hypothesized that endotracheal intubation (ETI) is associated with decreased incidence of CXR abnormalities after out-of-hospital cardiac arrest (OHCA), as compared to bag-valve-mask (BVM) or supraglottic airway (SGA) use. Methods: We conducted a retrospective review including resuscitated OHCA patients treated at a single academic medical center from 2010-2015. We included patients that had an initial CXR obtained within 66 hours of arrival. We excluded patients with tracheostomy, patients without positive pressure ventilation on emergency department (ED) arrival, or missing initial airway management data. We classified patients by airway management at the time of ED arrival into three groups: BVM, SGA, and ETI. Board certified thoracic imaging radiologists determined if there was any CXR abnormality, and if the observed abnormality was likely due to aspiration. The incidence of any abnormality and aspiration were compared between groups. A multivariable logistic regression model was used to adjust for baseline clinical characteristics. Results: Of the 766 subjects included in the study, 22 (3%) had BVM, 68 (9%) had SGA, and 676 (88%) had ETI. Most 58% were male, 34% had initial rhythm VF/VT, and 61% had a witnessed arrest. Any abnormality on CXR was identified in 543 (71%) cases, and likely aspiration was observed in 205 (27%) cases. Incidence of CXR abnormality was not significantly different between groups: BVM group 18/22 (82%), reference; SGA group 52/68 (76%), OR 0.75, 95% CI 0.13-4.31; ETI group 473/676 (70%), OR 0.81, 95% CI 0.16-4.01. Incidence of aspiration on CXR was also not different between groups: BVM group 6/22 (27%), reference; SGA group 19/68 (28%), OR 1.04, 95% CI 0.18-6.22; ETI group 180/676 (27%), OR 1.26, 95% CI 0.25-6.32. Conclusion: Prehospital airway management strategy for resuscitated OHCA patients was not associated with a significant difference in the incidence of any abnormality or aspiration on CXR.

Background: Chest X-rays are rapidly gaining prominence as a prevalent diagnostic tool, as recognized by the World Health Organization (WHO). However, interpreting chest X-rays can be demanding and time-consuming, even for experienced radiologists, leading to potential misinterpretations and delays in treatment. Method: The purpose of this research is the development of a RadAI model. The RadAI model can accurately detect four types of lung abnormalities in chest X-rays and generate a report on each identified abnormality. Moreover, deep learning algorithms, particularly convolutional neural networks (CNNs), have demonstrated remarkable potential in automating medical image analysis, including chest X-rays. This work addresses the challenge of chest X-ray interpretation by fine tuning the following three advanced deep learning models: Feature-selective and Spatial Receptive Fields Network (FSRFNet50), ResNext50, and ResNet50. These models are compared based on accuracy, precision, recall, and F1-score. Results: The outstanding performance of RadAI shows its potential to assist radiologists to interpret the detected chest abnormalities accurately. Conclusions: RadAI is beneficial in enhancing the accuracy and efficiency of chest X-ray interpretation, ultimately supporting the timely and reliable diagnosis of lung abnormalities.

BackgroundMost new cases of active tuberculosis in Norway are presently caused by imported strains and not transmission within the country. Screening for tuberculosis with a Mantoux test of everybody and a chest X-ray of those above 15 years of age is compulsory on arrival for asylum seekers.We aimed to assess the effectiveness of entry screening of a cohort of asylum seekers. Cases detected by screening were compared with cases detected later. Further we have characterized cases with active tuberculosis.MethodsAll asylum seekers who arrived at the National Reception Centre between January 2005 - June 2006 with an abnormal chest X-ray or a Mantoux test ≥ 6 mm were included in the study and followed through the health care system. They were matched with the National Tuberculosis Register by the end of May 2008.Cases reported within two months after arrival were defined as being detected by screening.ResultsOf 4643 eligible asylum seekers, 2237 were included in the study. Altogether 2077 persons had a Mantoux ≥ 6 mm and 314 had an abnormal chest X-ray. Of 28 cases with tuberculosis, 15 were detected by screening, and 13 at 4-27 months after arrival. Abnormal X-rays on arrival were more prevalent among those detected by screening. Female gender and Somalian origin increased the risk for active TB.ConclusionIn spite of an imperfect follow-up of screening results, a reasonable number of TB cases was identified by the programme, with a predominance of pulmonary TB.

The objective of our study was to compare the effect of dual-energy subtraction and bone suppression software alone and in combination with computer-aided detection (CAD) on the performance of human observers in lung nodule detection. One hundred one patients with from one to five lung nodules measuring 5-29 mm and 42 subjects with no nodules were retrospectively selected and randomized. Three independent radiologists marked suspicious-appearing lesions on the original chest radiographs, dual-energy subtraction images, and bone-suppressed images before and after postprocessing with CAD. Marks of the observers and CAD marks were compared with CT as the reference standard. Data were analyzed using nonparametric tests and the jackknife alternative free-response receiver operating characteristic (JAFROC) method. Using dual-energy subtraction alone (p = 0.0198) or CAD alone (p = 0.0095) improved the detection rate compared with using the original conventional chest radiograph. The combination of bone suppression and CAD provided the highest sensitivity (51.6%) and the original nonenhanced conventional chest radiograph alone provided the lowest (46.9%; p = 0.0049). Dual-energy subtraction and bone suppression provided the same false-positive (p = 0.2702) and true-positive (p = 0.8451) rates. Up to 22.9% of lesions were found only by the CAD program and were missed by the readers. JAFROC showed no difference in the performance between modalities (p = 0.2742-0.5442). Dual-energy subtraction and the electronic bone suppression program used in this study provided similar detection rates for pulmonary nodules. Additionally, CAD alone or combined with bone suppression can significantly improve the sensitivity of human observers for pulmonary nodule detection.

Computed tomography angiography is limited in the intensive care unit (ICU) due to renal insufficiency, hemodynamic instability, and difficulty transporting unstable patients. A portable ventilation/perfusion (V/Q) scan can be used. However, it is commonly believed that an abnormal chest radiograph can result in a nondiagnostic scan. In this retrospective study, we demonstrate that portable V/Q scans can be helpful in ruling in or out clinically significant pulmonary embolism (PE) despite an abnormal chest x-ray in the ICU. Two physicians conducted chart reviews and original V/Q reports. A staff radiologist, with 40 years of experience, rated chest x-ray abnormalities using predetermined criteria. The study was conducted in the ICU. The first 100 consecutive patients with suspected PE who underwent a portable V/Q scan. Those with a portable V/Q scan. A normal baseline chest radiograph was found in only 6% of patients. Fifty-three percent had moderate, 24% had severe, and 10% had very-severe radiographic abnormalities. Despite the abnormal x-rays, 88% of the V/Q scans were low probability for a PE despite an average abnormal radiograph rating of moderate. A high-probability V/Q for PE was diagnosed in 3% of the population despite chest x-ray ratings of moderate to severe. Six patients had their empiric anticoagulation discontinued after obtaining the results of the V/Q scan, and no anticoagulation was started for PE after a low-probability V/Q scan. Despite the large percentage of moderate-to-severe x-ray abnormalities, PE can still be diagnosed (high-probability scan) in the ICU with a portable V/Q scan. Although low-probability scans do not rule out acute PE, it appeared less likely that any patient with a low-probability V/Q scan had severe hypoxemia or hemodynamic instability due to a significant PE, which was useful to clinicians and allowed them to either stop or not start anticoagulation.

The purpose of this look is to increase an automated gadget for detecting chest X-ray imaging abnormalities through a Convolutional Neural community (CNN). The usage of a CNN permits real-time detection of abnormalities that could otherwise be missed via radiologists, as well as increases the efficiency of clinical experts by using being able to hit upon ability abnormalities without the need for extra tests. In this observation, 285 chest scans had been accrued from a local health facility and analyzed using a custom-built CNN. Through the use of various convolutional layers, various functions have been extracted and applied to stumble on abnormalities. Adjustments to the model have been made by including additional convolutional layers and adjusting the filters. The ensuing model carried out a ninety-seven—2 % accuracy fee in identifying abnormalities in chest X-rays. The outcomes of this look display the hit implementation of a CNN in detecting abnormalities in chest X-rays. In addition, further studies are had to refine the model and boom accuracy. Moreover, similar research is recommended to examine its use in scientific decision-making for sufferers. The results of this have a look at can be seen as a step forward in the improvement of automated clinical analysis structures and their usage in clinical decision-making.

Chest X-ray (CXR) is crucial for diagnosing lung diseases, especially lung nodules. Recent studies indicate that bones, such as ribs and clavicles, obscure 82 to 95% of undiagnosed lung cancers. The development of computer-aided detection (CAD) systems with automated bone suppression is vital to improve detection rates and support early clinical decision-making. Current bone suppression methods face challenges: they often depend on manual subtraction of bone-only images from CXRs, leading to inefficiency and poor generalization; there is significant information loss in data compression within deep convolutional end-to-end architectures; and a balance between model efficiency and accuracy has not been sufficiently achieved in existing research. We introduce a novel end-to-end architecture, the mask-guided model, to address these challenges. Leveraging the Pix2Pix framework, our model enhances computational efficiency by reducing parameter count by 92.5%. It features a rib mask-guided module with a mask encoder and cross-attention mechanism, which provides spatial constraints, reduces information loss during encoder compression, and preserves non-relevant areas. An ablation study evaluates the impact of various factors. The model undergoes initial training on digitally reconstructed radiographs (DRRs) derived from CT projections for bone suppression and is fine-tuned on the JSRT dataset to accelerate convergence. The mask-guided model surpasses previous state-of-the-art methods, showing superior bone suppression performance in terms of structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and processing speed. It achieves an SSIM of 0.99 ± 0.002 and a PSNR of 36.14 ± 1.13 on the JSRT dataset. This study underscores the proposed model's effectiveness compared to existing methods, showcasing its capability to reduce model size and increase accuracy. This makes it well-suited for deployment in affordable, low-power hardware devices across various clinical settings.

There is uncertainty in the NHS about which individuals should be offered pre-employment screening by chest X-ray and whether this procedure is of value in the detection of tuberculosis. To provide evidence for practice, pre-employment chest X-ray and tuberculin skin test status were examined retrospectively for employees of a health district. Cases were those with an abnormal chest X-ray; referents were a sample of the remainder with a normal chest X-ray. The majority of the population had positive tuberculin skin tests and there was no difference between cases (58 positive out of 68) and referents (170 positive out of 212). Most of the X-ray abnormalities were trivial; four findings were thought significant, but would have caused no problems if undetected and none of these findings was related to tuberculosis. We conclude that in new employees in Hampstead health district, X-ray abnormalities are rare and not predicted by testing tuberculin skin reactivity; neither procedure is justified routinely as a means of screening for tuberculosis. The situation in districts with a high incidence of tuberculosis needs to be investigated.

Deep Feature Fusion of Local and Global Patterns for Early Detection of Lung Abnormalities in Chest X-Rays

Chest X-ray images are widely used in diagnosing medical conditions, however, due to radiologist fatigue and shortage of resources the possibility of spotting peculiarities increases. This work proposes an explainable AI system to identify, quantify, and explain the pathology found in chest x-ray images. It builds credibility, enhances medical judgment, and reflects efficient utilization of radiology assets. This study used ResNet for the chest X-ray image classification and compared it with other related models like EfficientNetB0, DenseNet, VGGNet, and MobileNet. ResNet had the best AUC value of 0.826 thereby garnering it the best classification accuracy hence validating its selection as the main model. Moreover, ResNet adapted explainability to provide visual and textual interpretations and to classify diseases where such actionable insights could improve the reliability of diagnosis and patient outcomes.

BackgroundScreening for latent tuberculosis (LTB) using chest x-ray (CXR) and interferon-gamma release assay (IGRA) is standard in patients starting on biologic and immunosuppressive therapies. Our hypothesis is that the CXR...

Performance of routine Chest X-rays (CXRs) in asymptomatic individuals to assess hyperbaric exposure risk is controversial. The radiation risk may overshadow the low yield in many settings. However, the yield may be higher in certain settings, such as tuberculosis-endemic countries. We evaluated the utility of routine CXR in diving and submarine medical examinations in South Africa. Records of 2,777 CXRs during 3,568 fitness examinations of 894 divers and submariners spanning 31 years were reviewed to determine the incidence of CXR abnormality. Associated factors were evaluated using odds ratios and a binomial logistic regression model, with a Kaplan-Meier plot to describe the duration of service until first abnormal CXR. An abnormal CXR was reported in 1.1% per person year of service, yielding a cumulative incidence of 6.5% (58/894) of the study participants. Only four individuals had a clinical indication for the CXR in their medical history. A range of potential pathologies were seen, of which 15.5% were declared disqualifying and the rest (84.5%) were treated, or further investigation showed that the person could be declared fit. In South Africa, a routine CXR has a role to play in detecting abnormalities that are incompatible with pressure exposures. The highest number of abnormalities were found during the initial examinations and in individuals with long service records. Only four individuals had a clinical indication for their CXR during the 31-year span of our study. Similar studies should be performed to make recommendations in other countries and settings.

The purpose of the study was to investigate to what extent marine engineers who are exposed to mineral oil mist in the engine rooms of ships have lung function disturbances and chest X-ray abnormalities. A previous study examined a cohort of 492 seamen from which we received 169 full-scale (40 x 40 cm) chest radiographs. Among these films were 68 from marine engineers and 101 from seamen who had never worked as engineers (controls). Spirometric data were obtained from 44 engineers and from 71 of the controls. The chest X-rays were classified independently by two certified B-readers, using the ILO classification. Only 7% of the seamen had parenchymal or pleural changes. The only statistically significant difference between the two groups for the chest X-ray findings was for all comments made by the ILO readers which were termed "any symbol" (9.3% for the engineers and 0.6% for the controls; RR = 14.6 (95% CI 1.1-75.5)) and the FEV% (FEV1/FVC x 100) for the spirometric values (82.8 (SD = 10.8) for the engineers and 86.6 (SD = 8.0) for the controls; P = 0.03). The results of the study indicate a slight lung function impairment and an increase in chest X-ray abnormalities, other than small opacities, among the marine engineers.

Due to the Coronavirus (COVID-19) cases growing rapidly, the effective screening of infected patients is becoming a necessity. One such way is through chest radiography. With the high stakes of false negatives being potential cause of innumerable more cases, expert opinions on x-rays are high in demand. In this scenario, Deep Learning and Machine Learning techniques offer fast and effective ways of detecting abnormalities in chest x-rays and can help in identifying patients affected by COVID-19. In this paper, we did comparative analysis of various Machine Learning and Deep Learning techniques on chest x-rays based on accuracy, precision, recall, f1 score, and Matthews correlation coefficient. It was observed that improved results were obtained using Deep Learning.KeywordsCOVID-19CoronavirusMachine learningDeep learning