Artificial intelligence for differentiating COVID-19 from other viral pneumonias on CT: comparative analysis of different models based on quantitative and radiomic approaches.

BACKGROUND: To develop a pipeline for automatic extraction of quantitative metrics and radiomic features from lung computed tomography (CT) and develop artificial intelligence (AI) models supporting differential diagnosis between coronavirus disease 2019 (COVID-19) and other viral pneumonia (non-COVID-19). METHODS: Chest CT of 1,031 patients (811 for model building; 220 as independent validation set (IVS) with positive swab for severe acute respiratory syndrome coronavirus-2 (647 COVID-19) or other respiratory viruses (384 non-COVID-19) were segmented automatically. A Gaussian model, based on the HU histogram distribution describing well-aerated and ill portions, was optimised to calculate quantitative metrics (QM, n = 20) in both lungs (2L) and four geometrical subdivisions (GS) (upper front, lower front, upper dorsal, lower dorsal; n = 80). Radiomic features (RF) of first (RF1, n = 18) and second (RF2, n = 120) order were extracted from 2L using PyRadiomics tool. Extracted metrics were used to develop four multilayer-perceptron classifiers, built with different combinations of QM and RF: Model1 (RF1-2L); Model2 (QM-2L, QM-GS); Model3 (RF1-2L, RF2-2L); Model4 (RF1-2L, QM-2L, GS-2L, RF2-2L). RESULTS: The classifiers showed accuracy from 0.71 to 0.80 and area under the receiving operating characteristic curve (AUC) from 0.77 to 0.87 in differentiating COVID-19 versus non-COVID-19 pneumonia. Best results were associated with Model3 (AUC 0.867 ± 0.008) and Model4 (AUC 0.870 ± 0.011. For the IVS, the AUC values were 0.834 ± 0.008 for Model3 and 0.828 ± 0.011 for Model4. CONCLUSIONS: Four AI-based models for classifying patients as COVID-19 or non-COVID-19 viral pneumonia showed good diagnostic performances that could support clinical decisions.

Diagnostic Performance in Differentiating COVID-19 from Other Viral Pneumonias on CT Imaging: Multi-Reader Analysis Compared with an Artificial Intelligence-Based Model.

Growing evidence suggests that artificial intelligence tools could help radiologists in differentiating COVID-19 pneumonia from other types of viral (non-COVID-19) pneumonia. To test this hypothesis, an R-AI classifier capable of discriminating between COVID-19 and non-COVID-19 pneumonia was developed using CT chest scans of 1031 patients with positive swab for SARS-CoV-2 (n = 647) and other respiratory viruses (n = 384). The model was trained with 811 CT scans, while 220 CT scans (n = 151 COVID-19; n = 69 non-COVID-19) were used for independent validation. Four readers were enrolled to blindly evaluate the validation dataset using the CO-RADS score. A pandemic-like high suspicion scenario (CO-RADS 3 considered as COVID-19) and a low suspicion scenario (CO-RADS 3 considered as non-COVID-19) were simulated. Inter-reader agreement and performance metrics were calculated for human readers and R-AI classifier. The readers showed good agreement in assigning CO-RADS score (Gwet's AC2 = 0.71, p < 0.001). Considering human performance, accuracy = 78% and accuracy = 74% were obtained in the high and low suspicion scenarios, respectively, while the AI classifier achieved accuracy = 79% in distinguishing COVID-19 from non-COVID-19 pneumonia on the independent validation dataset. The R-AI classifier performance was equivalent or superior to human readers in all comparisons. Therefore, a R-AI classifier may support human readers in the difficult task of distinguishing COVID-19 from other types of viral pneumonia on CT imaging.

Antibody Response to COVID-19 Booster Vaccination in Healthcare Workers.

Objective: To evaluate the mean increase of anti-S IgG antibody titer between the basal, pre-booster level to the titer assessed 14 days after the booster dose of BNT162b2. Patients and Methods: The RENAISSANCE study is an observational, longitudinal, prospective, population-based study, conducted on healthcare workers of Niguarda Hospital in Milan, Italy who received a BNT162b2 booster dose at least 180 days after their second dose or after positivity for SARS-CoV-2 and accepted to take part in the study. The RENAISSANCE study was conducted from January 1, 2021 through December 28, 2021. Findings: 1,738 subjects were enrolled among healthcare workers registered for the booster administration at our hospital. Overall, 0.4% of subjects were seronegative at the pre-booster evaluation, and 1 subject had a titer equal to 50 AU/ml: none of the evaluated subjects was seronegative after the booster dose. Thus, the efficacy of the booster in our population was universal. Mean increase of pre- to post-booster titer was more significant in subjects who never had SARS-CoV-2 (44 times CI 95% 42-46) compared to those who had it, before (33 times, CI 95% 13-70) or after the first vaccination cycle (12 times, CI 95% 11-14). Differently from sex, age and pre-booster titers affected the post-booster antibody response. Nevertheless, the post-booster titer was very similar in all subgroups, and independent of a prior exposure to SARS-CoV-2, pre-booster titer, sex or age. Conclusion: Our study shows a potent universal antibody response of the booster dose of BNT162b2, regardless of pre-booster vaccine seronegativity.

Polymorphism of the HLA system and weak antibody response to BNT162b2 mRNA vaccine.

The polymorphism of the HLA system has been extensively studied in COVID-19 infection, however there are no data about the role of HLA on vaccine response. We report here the HLA-A, -B, -C, and DRB1 allelic frequencies of n = 111 individuals after BNT162b2 mRNA vaccine, selected on the basis of lower antibody levels (<5% percentile) after the second dose among a total of n = 2569 vaccinees, and compare them with the frequencies of a reference population. We found that differences in the frequencies of the alleles HLA-A*03:01, A*33:03, B*58:01 and at least one haplotype (HLA-A*24:02~C*07:01~B*18:01~DRB1*11:04) are associated with a weaker antibody response after vaccination, together with the age of vaccinees. Our results might suggest a role played by some HLA alleles or haplotypes in antibody production after the BNT162b2 mRNA vaccine, giving insights into the tracking of potentially susceptible individuals across populations. Further studies are needed to better define our exploratory findings and dissect the role of HLA polymorphism on response to anti-COVID-19 vaccines.

SARS-COV-2 screening in allogeneic hematopoietic stem cell donors: Implications for the evaluation process and eligibility.

INTRODUCTION: Soon after the onset of the SARS-CoV-2 pandemic, viral screening by nasopharyngeal swab became mandatory for allogeneic hematopoietic stem cell (HSC) donor eligibility. METHODS: We described our monocenter experience with allogeneic HSC donors from February 1 to the October 31, 2020 to verify whether the introduction of SARS-CoV-2 screening altered the donor eligibility and/or entailed a prolongation of the evaluation process. RESULTS: A total of 21 allogeneic HSC donors were screened during the above-mentioned period upon request by the local transplant physicians or by the Italian Bone Marrow Donor Registry; among the HSC donors (n = 17) who completed the eligibility process and further received the nasopharyngeal swab, all but one were negative for the presence of SARS-CoV-2. The positive donor remained asymptomatic for the whole duration of the infection, which lasted six weeks. However, he was temporarily excluded from donation. The median duration of the evaluation process was not significantly different, compared to the same period of 2019 (p-value = 0.11). CONCLUSION: The mandatory SARS-CoV-2 screening in allogeneic HSC donors allowed for the detection of 6% positivity in this monocenter series over a 9-month period. Despite the inconvenience of this unexpected non-eligibility, the exclusion of a SARS-CoV-2 positive donor represented an important safety measure for the donor, with respect to a new and still partially unknown virus. The screening did not alter the length of the donor evaluation and thus, did not cause a delay in the eligibility process.

Sex differences in electrolyte imbalances caused by SARS-CoV-2: A cross-sectional study.

BACKGROUND: Since SARS-CoV-2 spread, evidence regarding sex differences in progression and prognosis of COVID-19 have emerged. Besides this, studies on patients' clinical characteristics have described electrolyte imbalances as one of the recurrent features of COVID-19. METHODS: We performed a cross-sectional study on all patients admitted to the emergency department (ED) from 1 March to 31 May 2020 who had undergone a blood gas analysis and a nasopharyngeal swab test for SARS-CoV-2 by rtPCR. We defined positive patients as cases (n = 710) and negatives as controls (n = 619), for a total number of patients of 1.329. The study was approved by the local ethics committee Area 3 Milan. Data were automatically extracted from the hospital laboratory SQL-based repository in anonymised form. We considered as outcomes potassium (K+ ), sodium (Na+ ), chlorine (Cl- ) and calcium (Ca++ ) as continuous and as categorical variables, in their relation with age, sex and SARS-CoV-2 infection status. RESULTS: We observed a higher prevalence of hypokalaemia among patients positive for SARS-CoV-2 (13.7% vs 6% of negative subjects). Positive patients had a higher probability to be admitted to the ED with hypokalaemia (OR 2.75, 95% CI 1.8-4.1, P < .0001) and women were twice as likely to be affected than men (OR 2.43, 95% CI 1.67-3.54, P < .001). Odds ratios for positive patients to manifest with an alteration in serum Na+ was (OR 1.6, 95% CI 1.17-2.35, P < .001) and serum chlorine (OR 1.6, 95% CI 1.03-2.69, P < .001). Notably, OR for positive patients to be hypocalcaemic was 7.2 (95% CI 4.8-10.6, P < .0001) with a low probability for women to be hypocalcaemic (OR 0.63, 95% CI 0.4-0.8, P = .005). CONCLUSIONS: SARS-CoV-2 infection is associated with a higher prevalence of hypokalaemia, hypocalcaemia, hypochloraemia and sodium alterations. Hypokalaemia is more frequent among women and hypocalcaemia among men.

Results of the RENAISSANCE Study: REsponse to BNT162b2 COVID-19 vacciNe-short- And long-term Immune reSponSe evAluatioN in health Care workErs.

OBJECTIVE: To evaluate the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) anti-spike (S) IgG antibody production after vaccination with BNT162b2 and the protection from symptomatic breakthrough infections in health care workers. METHODS: This prospective observational study (RENAISSANCE) had as a primary end point the evaluation of serologic response to BNT162b2 14 days after a second dose. SARS-CoV-2 anti-S IgG antibodies were evaluated with LIAISON SARS-CoV-2 TrimericS IgG assay (DiaSorin S.p.A.), which is able to detect the presence of both binding and neutralizing antibodies for trimeric spike glycoprotein. Participants were recruited from February 1, 2021, to February 22, 2021. Occurrence of vaccine breakthrough infections was assessed by reverse transcription-polymerase chain reaction on symptomatic and contact cases up to June 6, 2021. RESULTS: Of 2569 staff evaluated, only 4 were nonresponders (0.16%; 95% CI, 0.04% to 0.41%). All 4 nonresponders were severely immunosuppressed and receiving treatment with mycophenolate mofetil or mycophenolic acid. At 14 days after the second dose, 67.5% (1733) of staff had anti-S IgG titers of 2000 BAU/mL or higher; 19.2% (494), between 1500 and 2000 BAU/mL; 9.8% (251), between 1000 and 1500 BAU/mL; and 3.4% (87), 1000 BAU/mL or lower. Women had a higher probability of having higher titers than men (64.5% [1044/1618] vs 58.3% [410/703]; P=.005). This was confirmed after adjustment for age group (odds ratio, 1.275; 95% CI, 1.062 to 1.531; P=.009). Four months after the end of the vaccination program, only 13 participants (0.26%) had experienced a breakthrough SARS-CoV-2 infection, including 1 nonresponder. This was the only participant requiring hospitalization for severe COVID-19. CONCLUSION: The vaccination campaign among health care workers at the ASST GOM Niguarda has resulted in a marked serologic response and reduction of incident COVID-19 cases. Yet, the lack of protection should not be overlooked in immunocompromised individuals.

Sex differences in electrolyte imbalances caused by SARS-CoV-2: a cross-sectional study

Abstract Background Since SARS-CoV-2 spread, evidences regarding sex differences in progression and prognosis of COVID-19 have emerged. Besides this, studies on patients? clinical characteristics have described electrolyte imbalances as one of the recurrent features of COVID-19. Methods We performed a cross-sectional study on all patients admitted to the emergency department (ED) from 1st March to 31st May 2020 who had undergone a blood gas analysis and a nasopharyngeal swab test for SARS-CoV-2 by rtPCR. We defined positive patients as cases (n. 710) and negatives as controls (n. 619), for a total number of patients of 1.329. The study was approved by the local ethics committee Area 3 Milan. Data were automatically extracted from the hospital laboratory SQL-based repository in anonymized form. We considered as outcomes potassium (K+), sodium (Na+), chlorine (Cl-) and calcium (Ca++) as continuous and as categorical variables, in their relation with age, sex and SARS-CoV-2 infection status. Results We observed a higher prevalence of hypokalemia among patients positive for SARS-CoV-2 (13.7% vs 6% of negative subjects). Positive patients had a higher probability to be admitted to the ED with hypokalemia (OR 2.75, 95% CI 1.8-4.1 p<0.0001) and women were twice as likely to be affected than men (OR 2.43, 95% CI 1.67-3.54 p<0.001). Odds ratios for positive patients to manifest with an alteration in serum Na+ was (OR 1.6, 95% CI 1.17-2.35 p<0.001) and serum chlorine (OR 1.6, 95% CI 1.03-2.69 p<0.001). Notably, OR for positive patients to be hypocalcemic was 7.2 (95% CI 4.8-10.6 p<0.0001) with a low probability for women to be hypocalcemic (OR 0.63, 95% IC 0.4-0.8 p=0.005). Conclusions SARS-CoV-2 infection is associated with a higher prevalence of hypokalemia, hypocal- cemia, hypochloremia and sodium alterations. Hypokalemia is more frequent among women and hypocal- cemia among men.

Frontline Screening for SARS-CoV-2 Infection at Emergency Department Admission by Third Generation Rapid Antigen Test: Can We Spare RT-qPCR?

To complement RT-qPCR testing for diagnosis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections, many countries have introduced the use of rapid antigen tests. As they generally display lower real-life performances than expected, their correct positioning as frontline screening is still controversial. Despite the lack of data from daily clinical use, third generation microfluidic assays (such as the LumiraDx SARS-CoV-2 Ag test) have recently been suggested to have similar performances to RT-qPCR and have been proposed as alternative diagnostic tools. By analyzing 960 nasopharyngeal swabs from 960 subjects at the emergency department admissions of a tertiary COVID-19 hospital, LumiraDx assay demonstrated a specificity of 97% (95% CI: 96-98), and a sensitivity of 85% (95% CI: 82-89) in comparison with RT-qPCR, which increases to 91% (95% CI: 86-95) for samples with a cycle threshold ≤ 29. Fifty false-negative LumiraDx-results were confirmed by direct quantification of genomic SARS-CoV-2 RNA through droplet-digital PCR (median (IQR) load = 5880 (1657-41,440) copies/mL). Subgenomic N and E RNAs were detected in 52% (n = 26) and 56% (n = 28) of them, respectively, supporting the presence of active viral replication. Overall, the LumiraDx test complies with the minimum performance requirements of the WHO. Yet, the risk of a misrecognition of patients with active COVID-19 persists, and the need for confirmatory RT-qPCR should not be amended.

SARS-CoV-2 RNA in plasma samples of COVID-19 affected individuals: a cross-sectional proof-of-concept study.

BACKGROUND: Recent studies showed that plasma SARS-CoV-2 RNA seems to be associated with worse COVID-19 outcome. However, whether specific population can be at higher risk of viremia are to date unexplored. METHODS: This cross-sectional proof-of-concept study included 41 SARS-CoV-2-positive adult individuals (six affected by haematological malignancies) hospitalized at two major hospital in Milan, for those demographic, clinical and laboratory data were available. SARS-CoV-2 load was quantified by ddPCR in paired plasma and respiratory samples. To assess significant differences between patients with and patients without viremia, Fisher exact test and Wilcoxon test were used for categorical and continuous variables, respectively. RESULTS: Plasma SARS-CoV-2 RNA was found in 8 patients (19.5%), with a median (IQR) value of 694 (209-1023) copies/mL. Viremic patients were characterized by an higher mortality rate (50.0% vs 9.1%; p = 0.018) respect to patients without viremia. Viremic patients were more frequently affected by haematological malignancies (62.5% vs. 3.0%; p < 0.001), and had higher viral load in respiratory samples (9,404,000 [586,060-10,000,000] vs 1560 [312-25,160] copies/mL; p = 0.002). CONCLUSIONS: Even if based on a small sample population, this proof-of-concept study poses the basis for an early identification of patients at higher risk of SARS-CoV-2 viremia, and therefore likely to develop severe COVID-19, and supports the need of a quantitative viral load determination in blood and respiratory samples of haematologic patients with COVID-19 in order to predict prognosis and consequently to help their further management.

Genomic epidemiology of SARS-CoV-2 reveals multiple lineages and early spread of SARS-CoV-2 infections in Lombardy, Italy.

From February to April 2020, Lombardy (Italy) reported the highest numbers of SARS-CoV-2 cases worldwide. By analyzing 346 whole SARS-CoV-2 genomes, we demonstrate the presence of seven viral lineages in Lombardy, frequently sustained by local transmission chains and at least two likely to have originated in Italy. Six single nucleotide polymorphisms (five of them non-synonymous) characterized the SARS-CoV-2 sequences, none of them affecting N-glycosylation sites. The seven lineages, and the presence of local transmission clusters within three of them, revealed that sustained community transmission was underway before the first COVID-19 case had been detected in Lombardy.

Development of machine learning models to predict RT-PCR results for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in patients with influenza-like symptoms using only basic clinical data.

BACKGROUND: Reverse Transcription-Polymerase Chain Reaction (RT-PCR) for Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-COV-2) diagnosis currently requires quite a long time span. A quicker and more efficient diagnostic tool in emergency departments could improve management during this global crisis. Our main goal was assessing the accuracy of artificial intelligence in predicting the results of RT-PCR for SARS-COV-2, using basic information at hand in all emergency departments. METHODS: This is a retrospective study carried out between February 22, 2020 and March 16, 2020 in one of the main hospitals in Milan, Italy. We screened for eligibility all patients admitted with influenza-like symptoms tested for SARS-COV-2. Patients under 12 years old and patients in whom the leukocyte formula was not performed in the ED were excluded. Input data through artificial intelligence were made up of a combination of clinical, radiological and routine laboratory data upon hospital admission. Different Machine Learning algorithms available on WEKA data mining software and on Semeion Research Centre depository were trained using both the Training and Testing and the K-fold cross-validation protocol. RESULTS: Among 199 patients subject to study (median [interquartile range] age 65 [46-78] years; 127 [63.8%] men), 124 [62.3%] resulted positive to SARS-COV-2. The best Machine Learning System reached an accuracy of 91.4% with 94.1% sensitivity and 88.7% specificity. CONCLUSION: Our study suggests that properly trained artificial intelligence algorithms may be able to predict correct results in RT-PCR for SARS-COV-2, using basic clinical data. If confirmed, on a larger-scale study, this approach could have important clinical and organizational implications.

Effectiveness of infection-containment measures on SARS-CoV-2 seroprevalence and circulation from May to July 2020, in Milan, Italy.

OBJECTIVE: Through a hospital-based SARS-CoV-2 molecular and serological screening, we evaluated the effectiveness of two months of lockdown and two of surveillance, in Milan, Lombardy, the first to be overwhelmed by COVID-19 pandemics during March-April 2020. METHODS: All subjects presenting at the major hospital of Milan from May-11 to July-5, 2020, underwent a serological screening by chemiluminescent assays. Those admitted were further tested by RT-PCR. RESULTS: The cumulative anti-N IgG seroprevalence in the 2753 subjects analyzed was of 5.1% (95%CI = 4.3%-6.0%), with a peak of 8.4% (6.1%-11.4%) 60-63 days since the peak of diagnoses (March-20). 31/106 (29.2%) anti-N reactive subjects had anti-S1/S2 titers >80 AU/mL. Being tested from May-18 to June-5, or residing in the provinces with higher SARS-CoV-2 circulation, were positively and independently associated with anti-N IgG reactivity (OR [95%CI]: 2.179[1.455-3.264] and 3.127[1.18-8.29], respectively). In the 18 RT-PCR positive, symptomatic subjects, anti-N seroprevalence was 33.3% (95% CI: 14.8%-56.3%). CONCLUSION: SARS-CoV-2 seroprevalence in Milan is low, and in a downward trend after only 60-63 days since the peak of diagnoses. Italian confinement measures were effective, but the risk of contagion remains concrete. In hospital-settings, the performance of molecular and serological screenings upon admission remains highly advisable.

Detection and quantification of SARS-CoV-2 by droplet digital PCR in real-time PCR negative nasopharyngeal swabs from suspected COVID-19 patients.

Since SARS-CoV-2-based disease (COVID-19) spreads as a pandemic, the necessity of a highly sensitive molecular diagnosis that can drastically reduce false negatives reverse transcription PCR (rtPCR) results, raises as a major clinical need. Here we evaluated the performance of a ddPCR-based assay to quantify SARS-CoV-2 titer in 55 suspected COVID-19 cases with negative rtPCR results thanks to in-house ddPCR assay (targeting RdRp and host RNaseP). Samples were collected at ASST-GOM Niguarda between February and May 2020 at hospital admission. Clinical and imaging data were obtained for clinical staging and definition of disease severity. Patients were mainly female (45.5%) with a median age of 73 (57-84) years. ddPCR-based assay detected SARS-CoV-2 genome in nasopharyngeal samples of 19 (34.5%) patients (median viral-load: 128 copies/mL, IQR: 72-345). In 15 of them (78.9%), chest CT showed a classical COVID-19 bilateral interstitial pneumonia; 14 patients (73.7%) showed severe COVID-19 manifestations. ddPCR did not identify any trace of SARS-CoV-2 genome in the respiratory samples of the remaining 36 patients. The serological assay performed in a subgroup of 34 patients at the later stage of illness (from 3 days to 90 days after) confirmed the presence of SARS-CoV-2 antibodies in all patients tested positive for SARS-CoV-2 in ddPCR (100%). Contrariwise, negative tests were observed in 95.0% ddPCR negative patients (P<0.001). Thanks to a ddPCR-based assay, we achieved a rapid and accurate SARS-CoV-2 diagnosis in rtPCR-negative respiratory samples of individuals with COVID-19 suspect, allowing the rapid taking care and correct management of these patients.

Clinical and Analytical Performance of an Automated Serological Test That Identifies S1/S2-Neutralizing IgG in COVID-19 Patients Semiquantitatively.

In the coronavirus (CoV) disease 2019 (COVID-19) pandemic, highly selective serological testing is essential to define exposure to severe acute respiratory syndrome CoV 2 (SARS-CoV-2). Many tests have been developed, yet with variable speeds to first results, and are of unknown quality, particularly when considering the prediction of neutralizing capacity. The LIAISON SARS-CoV-2 S1/S2 IgG assay was designed to measure antibodies against the SARS-CoV-2 native S1/S2 proteins in a standardized automated chemiluminescence assay. The clinical and analytical performances of the test were validated in an observational study using residual samples (>1,500) with a positive or negative COVID-19 diagnosis. The LIAISON SARS-CoV-2 S1/S2 IgG assay proved to be highly selective and specific and offered semiquantitative measures of serum or plasma levels of anti-S1/S2 IgG with neutralizing activity. The assay's diagnostic sensitivities were 91.3% and 95.7% at >5 or ≥15 days from diagnosis, respectively, and 100% when assessed against a neutralizing assay. The assay's specificity ranged between 97% and 98.5%. The average imprecision of the assay was a <5% coefficient of variation. Assay performance at 2 different cutoffs was evaluated to optimize predictive values. The automated LIAISON SARS-CoV-2 S1/S2 IgG assay brings efficient, sensitive, specific, and precise serological testing to the laboratory, with the capacity to test large amounts of samples per day; first results are available within 35 min, with a throughput of 170 tests/hour. The semiquantitative results provided by the test also associate with the presence of neutralizing antibodies and may provide a useful tool for the large-scale screening of convalescent-phase plasma for safe therapeutic use.