ObjectivesWe tested artificial intelligence (AI) to support the diagnosis of COVID-19 using chest X-ray (CXR). Diagnostic performance was computed for a system trained on CXRs of Italian subjects from two hospitals in Lombardy, Italy. MethodsWe used for training and internal testing an ensemble of ten convolutional neural networks (CNNs) with mainly bedside CXRs of 250 COVID-19 and 250 non-COVID-19 subjects from two hospitals.We then tested such system on bedside CXRs of an independent group of 110 patients 36 non-COVID-19) from one of the two hospitals. A retrospective reading was performed by two radiologists in the absence of any clinical information, with the aim to differentiate COVID-19 from non-COVID-19 patients. Real-time polymerase chain reaction served as reference standard.Results At 10-fold cross-validation, our AI model classified COVID-19 and non COVID-19 patients with 0.78 sensitivity (95% confidence interval [CI] 0.74-0.81), 0.82 specificity (95% CI 0.78-0.85) and 0.89 area under the curve (AUC) (95% CI 0.86-0.91). For the independent dataset, AI showed 0.80 sensitivity (95% CI 0.72-0.86) (59/74), 0.81 specificity (29/36) (95% CI 0.73-0.87), and 0.81 AUC (95% CI 0.73-0.87). Radiologists' reading obtained 0.63 sensitivity (95% CI 0.52-0.74) and 0.78 specificity (95% CI 0.61-0.90) in one centre and 0.64 sensitivity (95% CI 0.52-0.74) and 0.86 specificity (95% CI 0.71-0.95) in the other.Conclusions This preliminary experience based on ten CNNs trained on a limited training dataset shows an interesting potential of AI for COVID-19 diagnosis. Such tool is in training with new CXRs to further increase its performance.

We reviewed technical parameters, acquisition protocols and adverse reactions (ARs) for contrast-enhanced spectral mammography (CESM). A systematic search in databases, including MEDLINE/EMBASE, was performed to extract publication year, country of origin, study design; patients; mammography unit/vendor, radiation dose, low-/high-energy tube voltage; contrast molecule, concentration and dose; injection modality, ARs and acquisition delay; order of views; examination time. Of 120 retrieved articles, 84 were included from 22 countries (September 2003–January 2019), totalling 14012 patients. Design was prospective in 44/84 studies (52%); in 70/84 articles (83%), a General Electric unit with factory-set kVp was used. Per-view average glandular dose, reported in 12/84 studies (14%), ranged 0.43–2.65 mGy. Contrast type/concentration was reported in 79/84 studies (94%), with Iohexol 350 mgI/mL mostly used (25/79, 32%), dose and flow rate in 72/84 (86%), with 1.5 mL/kg dose at 3 mL/s in 62/72 studies (86%). Injection was described in 69/84 articles (82%), automated in 59/69 (85%), manual in 10/69 (15%) and flush in 35/84 (42%), with 10–30 mL dose in 19/35 (54%). An examination time < 10 min was reported in 65/84 studies (77%), 120 s acquisition delay in 65/84 (77%) and order of views in 42/84 (50%) studies, beginning with the craniocaudal view of the non-suspected breast in 7/42 (17%). Thirty ARs were reported by 14/84 (17%) studies (26 mild, 3 moderate, 1 severe non-fatal) with a pooled rate of 0.82% (fixed-effect model). Only half of CESM studies were prospective; factory-set kVp, contrast 1.5 mL/kg at 3 mL/s and 120 s acquisition delay were mostly used; only 1 severe AR was reported. CESM protocol standardisation is advisable.

Background We aimed to train and test a deep learning classifier to support the diagnosis of coronavirus disease 2019 (COVID-19) using chest x-ray (CXR) on a cohort of subjects from two hospitals in Lombardy, Italy. Methods We used for training and validation an ensemble of ten convolutional neural networks (CNNs) with mainly bedside CXRs of 250 COVID-19 and 250 non-COVID-19 subjects from two hospitals (Centres 1 and 2). We then tested such system on bedside CXRs of an independent group of 110 patients (74 COVID-19, 36 non-COVID-19) from one of the two hospitals. A retrospective reading was performed by two radiologists in the absence of any clinical information, with the aim to differentiate COVID-19 from non-COVID-19 patients. Real-time polymerase chain reaction served as the reference standard. Results At 10-fold cross-validation, our deep learning model classified COVID-19 and non-COVID-19 patients with 0.78 sensitivity (95% confidence interval [CI] 0.74–0.81), 0.82 specificity (95% CI 0.78–0.85), and 0.89 area under the curve (AUC) (95% CI 0.86–0.91). For the independent dataset, deep learning showed 0.80 sensitivity (95% CI 0.72–0.86) (59/74), 0.81 specificity (29/36) (95% CI 0.73–0.87), and 0.81 AUC (95% CI 0.73–0.87). Radiologists’ reading obtained 0.63 sensitivity (95% CI 0.52–0.74) and 0.78 specificity (95% CI 0.61–0.90) in Centre 1 and 0.64 sensitivity (95% CI 0.52–0.74) and 0.86 specificity (95% CI 0.71–0.95) in Centre 2. Conclusions This preliminary experience based on ten CNNs trained on a limited training dataset shows an interesting potential of deep learning for COVID-19 diagnosis. Such tool is in training with new CXRs to further increase its performance.

No abstract

Epicardial adipose tissue (EAT) is a metabolically activated beige adipose tissue, non-homogeneously surrounding the myocardium. Physiologically, EAT regulates toxic fatty acids, protects the coronary arteries against mechanical strain, regulates proinflammatory cytokines, stimulates the production of nitric oxide, reduces oxidative stress, and works as a thermogenic source against hypothermia. Conversely, EAT has pathologic paracrine interactions with the surrounded vessels, and might favour the onset of atrial fibrillation. In addition, initial atherosclerotic lesions can promote inflammation and trigger the EAT production of cytokines increasing vascular inflammation, which, in turn, may help the development of collateral vessels but also of self-stimulating, dysregulated inflammatory process, increasing coronary artery disease severity. Variations in EAT were also linked to metabolic syndrome. Echocardiography first estimated EAT measuring its thickness on the free wall of the right ventricle but does not allow accurate volumetric EAT estimates. Cardiac CT (CCT) and cardiac MR (CMR) allow for three-dimensional EAT estimates, the former showing higher spatial resolution and reproducibility but being limited by radiation exposure and long segmentation times, the latter being radiation-free but limited by lower spatial resolution and reproducibility, higher cost, and difficulties for obese patients. EAT radiodensity at CCT could to be related to underlying metabolic processes. The correlation between EAT and response to certain pharmacological therapies has also been investigated, showing promising results. In the future, semi-automatic or fully automatic techniques, machine/deep-learning methods, if validated, will facilitate research for various EAT measures and may find a place in CCT/CMR reporting.