Background More than 20% of hospitalized patients with coronavirus disease 2019 (COVID-19) develop acute respiratory distress syndrome (ARDS) requiring intensive care unit (ICU) admission. The long-term respiratory sequelae in ICU survivors remain unclear. Research question what are the major long-term pulmonary sequelae in critical COVID-19 survivors? Study Design and Methods Consecutive patients with COVID-19 requiring ICU admission were recruited and evaluated 3 months after hospitalization discharge. The follow-up comprised symptom and quality of life, anxiety and depression questionnaires, pulmonary function tests, exercise test (6-minute walking test (6MWT)) and chest computed tomography (CT). Results 125 ICU patients with ARDS secondary to COVID-19 were recruited between March and June 2020. At the 3-month follow-up, 62 patients were available for pulmonary evaluation. The most frequent symptoms were dyspnea (46.7%), and cough (34.4%). Eighty-two percent of patients showed a lung diffusing capacity of less than 80%. The median (IQR) distance in the 6MWT was 400 (362;440) meters. CT scans were abnormal in 70.2% of patients, showing reticular lesions in 49.1% and fibrotic patterns in 21.1%. Patients with more severe alterations on chest CT had worse pulmonary function and presented more degrees of desaturation in the 6MWT. Factors associated with the severity of lung damage on chest CT were age and length of invasive mechanical ventilation during the ICU stay. Interpretation Pulmonary structural abnormalities and functional impairment are highly prevalent in surviving ICU patients with ARDS secondary to COVID-19 3 months after hospital discharge. Pulmonary evaluation should be considered for all critical COVID-19 survivors 3 months post discharge.

Modern biology frequently relies on machine learning to provide predictions and improve decision processes. There have been recent calls for more scrutiny on machine learning performance and possible limitations. Here we present a set of community-wide recommendations aiming to help establish standards of machine learning validation in biology. Adopting a structured methods description for machine learning based on DOME (data, optimization, model, evaluation) will allow both reviewers and readers to better understand and assess the performance and limitations of a method or outcome. The recommendations are complemented by a machine learning summary table which can be easily included in the supplementary material of published papers. Redundancy between data splitsMaximum pairwise identity within and between training and test set is 25% enforced with UniqueProt tool. Availability of dataYes, URL: http://protein.bio.unipd.it/espritz/ Optimization Algorithm BRNN (Bi-directional recurrent neural network) with ensemble averaging. Meta-predictionsNo. Data encodingSliding window of length 23 residues on input sequence with "one hot" encoding (i.e. 20 inputs per residue).

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Question We evaluated whether the time between first respiratory support and intubation of patients receiving invasive mechanical ventilation (IMV) due to COVID-19 was associated with mortality or pulmonary sequelae. Materials and methods Prospective cohort of critical COVID-19 patients on IMV. Patients were classified as early intubation if they were intubated within the first 48 h from the first respiratory support or delayed intubation if they were intubated later. Surviving patients were evaluated after hospital discharge. Results We included 205 patients (140 with early IMV and 65 with delayed IMV). The median [p25;p75] age was 63 [56.0; 70.0] years, and 74.1% were male. The survival analysis showed a significant increase in the risk of mortality in the delayed group with an adjusted hazard ratio (HR) of 2.45 (95% CI 1.29–4.65). The continuous predictor time to IMV showed a nonlinear association with the risk of in-hospital mortality. A multivariate mortality model showed that delay of IMV was a factor associated with mortality (HR of 2.40; 95% CI 1.42–4.1). During follow-up, patients in the delayed group showed a worse DLCO (mean difference of − 10.77 (95% CI − 18.40 to − 3.15), with a greater number of affected lobes (+ 1.51 [95% CI 0.89–2.13]) and a greater TSS (+ 4.35 [95% CI 2.41–6.27]) in the chest CT scan. Conclusions Among critically ill patients with COVID-19 who required IMV, the delay in intubation from the first respiratory support was associated with an increase in hospital mortality and worse pulmonary sequelae during follow-up.

Deep neural networks are representation learning techniques. During training, a deep net is capable of generating a descriptive language of unprecedented size and detail in machine learning. Extracting the descriptive language coded within a trained CNN model (in the case of image data), and reusing it for other purposes is a field of interest, as it provides access to the visual descriptors previously learnt by the CNN after processing millions of images, without requiring an expensive training phase. Contributions to this field (commonly known as feature representation transfer or transfer learning) have been purely empirical so far, extracting all CNN features from a single layer close to the output and testing their performance by feeding them to a classifier. This approach has provided consistent results, although its relevance is limited to classification tasks. In a completely different approach, in this paper we statistically measure the discriminative power of every single feature found within a deep CNN, when used for characterizing every class of 11 datasets. We seek to provide new insights into the behavior of CNN features, particularly the ones from convolutional layers, as this can be relevant for their application to knowledge representation and reasoning. Our results confirm that low and middle level features may behave differently to high level features, but only under certain conditions. We find that all CNN features can be used for knowledge representation purposes both by their presence or by their absence, doubling the information a single CNN feature may provide. We also study how much noise these features may include, and propose a thresholding approach to discard most of it. All these insights have a direct application to the generation of CNN embedding spaces.