U-shaped GAN for Semi-Supervised Learning and Unsupervised Domain Adaptation in High Resolution Chest Radiograph Segmentation

Deep learning has achieved considerable success in medical image segmentation. However, applying deep learning in clinical environments often involves two problems: (1) scarcity of annotated data as data annotation is time-consuming and (2) varying attributes of different datasets due to domain shift. To address these problems, we propose an improved generative adversarial network (GAN) segmentation model, called U-shaped GAN, for limited-annotated chest radiograph datasets. The semi-supervised learning approach and unsupervised domain adaptation (UDA) approach are modeled into a unified framework for effective segmentation. We improve GAN by replacing the traditional discriminator with a U-shaped net, which predicts each pixel a label. The proposed U-shaped net is designed with high resolution radiographs (1,024 × 1,024) for effective segmentation while taking computational burden into account. The pointwise convolution is applied to U-shaped GAN for dimensionality reduction, which decreases the number of feature maps while retaining their salient features. Moreover, we design the U-shaped net with a pretrained ResNet-50 as an encoder to reduce the computational burden of training the encoder from scratch. A semi-supervised learning approach is proposed learning from limited annotated data while exploiting additional unannotated data with a pixel-level loss. U-shaped GAN is extended to UDA by taking the source and target domain data as the annotated data and the unannotated data in the semi-supervised learning approach, respectively. Compared to the previous models dealing with the aforementioned problems separately, U-shaped GAN is compatible with varying data distributions of multiple medical centers, with efficient training and optimizing performance. U-shaped GAN can be generalized to chest radiograph segmentation for clinical deployment. We evaluate U-shaped GAN with two chest radiograph datasets. U-shaped GAN is shown to significantly outperform the state-of-the-art models.

Radiological Cardiothoracic Ratio as a Potential Predictor of Right Ventricular Enlargement in Patients with Suspected Pulmonary Embolism Due to COVID-19

The aim of the study was to determine the usefulness of the radiological cardiothoracic ratio (CTR) as a predictor of right ventricular enlargement in patients with suspected pulmonary embolism during COVID-19. The study group consisted of 61 patients with confirmed COVID-19, suspected of pulmonary embolism based on physical examination and laboratory tests (age: 67.18 ± 12.47 years). Computed tomography angiography (CTA) of pulmonary arteries and chest radiograph in AP projection with cardiothoracic ratio assessment were performed in all patients. Right ventricular enlargement was diagnosed by the ratio of right ventricular to left ventricular (RV/LV) dimensions in pulmonary CTA with two cut-off points: ≥0.9 and ≥1.0. Heart silhouette enlargement was found when CTR on the chest radiograph in the projection AP > 0.55. The mean values of RV/LV and CTR in the studied group were 0.96 ± 0.23 and 0.57 ± 0.05, respectively. Pulmonary embolism was diagnosed in 45.9%. Right ventricular enlargement was documented in 44.3% or 29.5% depending on the adopted criterion RV/LV ≥ 0.9 or RV/LV ≥ 1.0. Heart silhouette enlargement was found in 60.6%. Patients with confirmed pulmonary embolism (PE+) had a significantly higher RV/LV ratio and CTR than patients with excluded pulmonary embolism (PE−) (RV/LV: PE+ 1.08 ± 0.24, PE− 0.82 ± 0.12; CTR: PE+ 0.60 ± 0.05, PE− 0.54 ± 0.04; p < 0.05). The correlation analysis showed a statistically significant positive correlation between the RV/LV ratio and CTR (r = 0.59, p < 0.05). Based on the ROC curves, CTR values were determined as the optimal cut-off points for the prediction of right ventricular enlargement (RV/LV ≥ 0.9 or RV/LV ≥ 1.0), being 0.54 and 0.55, respectively. The sensitivity, specificity, and accuracy of the CTR criterion >0.54 as a predictor of RV/LV ratio ≥0.9 were 0.412, 0.963, and 0.656, respectively, while those of the CTR criterion >0.55 as a predictor of RV/LV ratio ≥1.0 were 0.488, 0.833, and 0.590, respectively. In summary, in patients with suspected pulmonary embolism during COVID-19, the radiographic cardiothoracic ratio can be considered as a prognostic factor for right ventricular enlargement, especially as a negative predictor of right ventricular enlargement in the case of lower CTR values.

Triage of Persons With Tuberculosis Symptoms Using Artificial Intelligence–Based Chest Radiograph Interpretation: A Cost-Effectiveness Analysis

Background In settings without access to rapid expert radiographic interpretation, artificial intelligence (AI)–based chest radiograph (CXR) analysis can triage persons presenting with possible tuberculosis (TB) symptoms, to identify those who require additional microbiological testing. However, there is limited evidence of the cost-effectiveness of this technology as a triage tool. Methods A decision analysis model was developed to evaluate the cost-effectiveness of triage strategies with AI-based CXR analysis for patients presenting with symptoms suggestive of pulmonary TB in Karachi, Pakistan. These strategies were compared to the current standard of care using microbiological testing with smear microscopy or GeneXpert, without prior triage. Positive triage CXRs were considered to improve referral success for microbiologic testing, from 91% to 100% for eligible persons. Software diagnostic accuracy was based on a prospective field study in Karachi. Other inputs were obtained from the Pakistan TB Program. The analysis was conducted from the healthcare provider perspective, and costs were expressed in 2020 US dollars. Results Compared to upfront smear microscopy for all persons with presumptive TB, triage strategies with AI-based CXR analysis were projected to lower costs by 19%, from $23233 per 1000 persons, and avert 3%–4% disability-adjusted life-years (DALYs), from 372 DALYs. Compared to upfront GeneXpert, AI-based triage strategies lowered projected costs by 37%, from $34346 and averted 4% additional DALYs, from 369 DALYs. Reinforced follow-up for persons with positive triage CXRs but negative microbiologic tests was particularly cost-effective. Conclusions In lower-resource settings, the addition of AI-based CXR triage before microbiologic testing for persons with possible TB symptoms can reduce costs, avert additional DALYs, and improve TB detection.

A Marmoset Model for Mycobacterium avium Complex Pulmonary Disease

Rationale Mycobacterium avium complex, is the most common nontuberculous mycobacterial respiratory pathogen in humans. Disease mechanisms are poorly understood due to the absence of a reliable animal model for M. avium complex pulmonary disease. Objectives Assess the susceptibility, immunologic and histopathologic responses of the common marmoset ( Callithrix jacchus ) to M. avium complex pulmonary infection. Methods 7 adult female marmosets underwent endobronchial inoculation with 10 8 colonyforming units of M. intracellulare and were monitored for 30 or 60 days. Prior to infection, chest radiograph and serum cytokines were assessed; serum cytokines were also monitored weekly for 30 days. At sacrifice 30 days (3 animals) or 60 days (4 animals) after infection, chest radiograph, serum and bronchoalveolar lavage cytokines, histopathology, and cultures of the bronchoalveolar lavage, lungs, liver, and kidney were analyzed. Measurements and Main Results Five of seven animals (two at 30 days and three at 60 days of infection) had positive lung cultures for M. intracellulare . Extra-pulmonary cultures were positive in three animals. All animals appeared healthy throughout the study. All five animals with positive lung cultures had radiographic changes consistent with pneumonitis. At 30 days, those with M. intracellulare lung infection showed granulomatous inflammation while at 60 days there was less inflammatory change but bronchiectasis was noted. The cytokine response in the bronchoalveolar lavage fluid was uniformly greater in the animals with positive M. intracellulare cultures than those without a productive infection with greater levels at 30-days compared to 60-days. Similarly, serum cytokines were more elevated in the animals that had positive M. intracellulare cultures compared to those without a productive infection, peaking 14-21 days after inoculation. Conclusion Endobronchial instillation of M. intracellulare resulted in pulmonary mycobacterial infection in marmosets with a differential immune response, radiographic and histopathologic abnormalities, and an indolent course consistent with M. avium complex lung infection in humans.