Active preference-based optimization for human-in-the-loop feature selection

In various classification problems characterized by a large number of features, feature selection (FS) is essential to guarantee generalization capabilities. The FS problem is often ill-posed due to significant correlations among features, which may lead to several different feature subsets with comparable scores in terms of classification performance. However, not all these subsets are equivalent from a domain-oriented point of view due to known relationships among features and their different acquisition costs in production to deploy the trained classifier. In this paper, we consider the potential benefits of including the domain expert’s preferences in the FS task, thus integrating both objective elements (e.g., classification accuracy) and subjective (often not quantifiable) considerations in the selection process. This goes in the direction of increasing the interpretability and the trustworthiness of the machine learning model, which is an often desired property in many application domains such as in medicine. The proposed method consists of an iterative procedure. At each iteration, the expert is asked to express a “human” preference on pairs of classifiers, each one trained from a different subset of features. The expressed preferences are used algorithmically to update a suitable surrogate function that mimics the latent subjective expert’s objective function, and then to propose a new classifier for testing and comparison. The proposed method has been tested on academic and experimental FS problems, and notably, on a COVID’19 patients record. The preliminary experimental results are promising, in that a parsimonious and accurate solution is obtained after a relatively short number of iterations.

How Artificial Intelligence and New Technologies Can Help the Management of the COVID-19 Pandemic.

The COVID-19 pandemic has worked as a catalyst, pushing governments, private companies, and healthcare facilities to design, develop, and adopt innovative solutions to control it, as is often the case when people are driven by necessity. After 18 months since the first case, it is time to think about the pros and cons of such technologies, including artificial intelligence-which is probably the most complex and misunderstood by non-specialists-in order to get the most out of them, and to suggest future improvements and proper adoption. The aim of this narrative review was to select the relevant papers that directly address the adoption of artificial intelligence and new technologies in the management of pandemics and communicable diseases such as SARS-CoV-2: environmental measures; acquisition and sharing of knowledge in the general population and among clinicians; development and management of drugs and vaccines; remote psychological support of patients; remote monitoring, diagnosis, and follow-up; and maximization and rationalization of human and material resources in the hospital environment.

Prediction of 28-day mortality in critically ill patients with COVID-19: Development and internal validation of a clinical prediction model.

BACKGROUND: COVID-19 pandemic has rapidly required a high demand of hospitalization and an increased number of intensive care units (ICUs) admission. Therefore, it became mandatory to develop prognostic models to evaluate critical COVID-19 patients. MATERIALS AND METHODS: We retrospectively evaluate a cohort of consecutive COVID-19 critically ill patients admitted to ICU with a confirmed diagnosis of SARS-CoV-2 pneumonia. A multivariable Cox regression model including demographic, clinical and laboratory findings was developed to assess the predictive value of these variables. Internal validation was performed using the bootstrap resampling technique. The model's discriminatory ability was assessed with Harrell's C-statistic and the goodness-of-fit was evaluated with calibration plot. RESULTS: 242 patients were included [median age, 64 years (56-71 IQR), 196 (81%) males]. Hypertension was the most common comorbidity (46.7%), followed by diabetes (15.3%) and heart disease (14.5%). Eighty-five patients (35.1%) died within 28 days after ICU admission and the median time from ICU admission to death was 11 days (IQR 6-18). In multivariable model after internal validation, age, obesity, procaltitonin, SOFA score and PaO2/FiO2 resulted as independent predictors of 28-day mortality. The C-statistic of the model showed a very good discriminatory capacity (0.82). CONCLUSIONS: We present the results of a multivariable prediction model for mortality of critically ill COVID-19 patients admitted to ICU. After adjustment for other factors, age, obesity, procalcitonin, SOFA and PaO2/FiO2 were independently associated with 28-day mortality in critically ill COVID-19 patients. The calibration plot revealed good agreements between the observed and expected probability of death.

A Machine Learning Approach for Mortality Prediction in COVID-19 Pneumonia: Development and Evaluation of the Piacenza Score.

BACKGROUND: Several models have been developed to predict mortality in patients with COVID-19 pneumonia, but only a few have demonstrated enough discriminatory capacity. Machine learning algorithms represent a novel approach for the data-driven prediction of clinical outcomes with advantages over statistical modeling. OBJECTIVE: We aimed to develop a machine learning-based score-the Piacenza score-for 30-day mortality prediction in patients with COVID-19 pneumonia. METHODS: The study comprised 852 patients with COVID-19 pneumonia, admitted to the Guglielmo da Saliceto Hospital in Italy from February to November 2020. Patients' medical history, demographics, and clinical data were collected using an electronic health record. The overall patient data set was randomly split into derivation and test cohorts. The score was obtained through the naïve Bayes classifier and externally validated on 86 patients admitted to Centro Cardiologico Monzino (Italy) in February 2020. Using a forward-search algorithm, 6 features were identified: age, mean corpuscular hemoglobin concentration, PaO2/FiO2 ratio, temperature, previous stroke, and gender. The Brier index was used to evaluate the ability of the machine learning model to stratify and predict the observed outcomes. A user-friendly website was designed and developed to enable fast and easy use of the tool by physicians. Regarding the customization properties of the Piacenza score, we added a tailored version of the algorithm to the website, which enables an optimized computation of the mortality risk score for a patient when some of the variables used by the Piacenza score are not available. In this case, the naïve Bayes classifier is retrained over the same derivation cohort but using a different set of patient characteristics. We also compared the Piacenza score with the 4C score and with a naïve Bayes algorithm with 14 features chosen a priori. RESULTS: The Piacenza score exhibited an area under the receiver operating characteristic curve (AUC) of 0.78 (95% CI 0.74-0.84, Brier score=0.19) in the internal validation cohort and 0.79 (95% CI 0.68-0.89, Brier score=0.16) in the external validation cohort, showing a comparable accuracy with respect to the 4C score and to the naïve Bayes model with a priori chosen features; this achieved an AUC of 0.78 (95% CI 0.73-0.83, Brier score=0.26) and 0.80 (95% CI 0.75-0.86, Brier score=0.17), respectively. CONCLUSIONS: Our findings demonstrated that a customizable machine learning-based score with a purely data-driven selection of features is feasible and effective for the prediction of mortality among patients with COVID-19 pneumonia.

[Late occurrence of pulmonary embolsim in SARS-CoV-2 pneumonia: a case series]. / L'embolia polmonare come complicanza tardiva della polmonite da SARS-CoV-2: una serie di casi clinici.

Several studies suggested that the acute phase of SARS-CoV-2 infection may be associated with a hypercoagulable state and increased risk for venous thromboembolism but the incidence of thrombotic complications in the late phase of the disease is currently unknown. The present article describes three cases of patients with SARS-CoV-2 pneumonia and late occurrence of pulmonary embolism. Case 1: a 57-year-old man diagnosed with pulmonary embolism and type B aortic dissection after 12 days from SARS-CoV-2 pneumonia. Laboratory panel at the time of pulmonary embolism showed no signs of ongoing inflammation but only an elevated D-dimer. Case 2: a 76-year-old man with a diagnosis of SARS-CoV-2 pneumonia followed by pulmonary embolism 20 days later, high-resolution computed tomography on that time showed a partial resolution of crazy paving consolidation. Case 3: a 77-year-old man with SARS-CoV-2 pneumonia who developed a venous thromboembolic event despite thromboprophylaxis with low molecular weight heparin. Also in this patients no markers of inflammation were present at the time of complication.The present cases raise the possibility that in SARS-CoV-2 infection the hypercoagulable state may persist over the active inflammation phase and cytokine storm. These findings suggest a role for medium-long term therapeutic anticoagulation started at the time of SARS-CoV-2 pneumonia diagnosis.