COVID-19 Detection from Cough Recording by means of Explainable Deep Learning

The new coronavirus disease (COVID-19), declared a pandemic on 11 March 2020 by the World Health Organization, has caused over 6 million victims worldwide. Because of the rapid spread of the virus, with the aim to perform screening we exploit deep learning model to quickly diagnose altered respiratory conditions. In this paper, we propose a method to recognize and classify cough audio files into three classes to distinguish patients with COVID-19 disease, symptomatic ones and healthy subjects, with the use of a convolutional neural network (CNN). Cough audios were recorded by using a smartphone and its built-in microphone. From cough recordings, we generate spectrogram images and we obtain an accuracy equal to 0.82 with a deep learning network developed by authors. Our method also provides heatmaps, which show the relevant input areas used by the model for the final forecast, and this aspect ensures the explainability of the method. © 2022 IEEE.

Evaluating lung perfusion SPECT/CT imaging in patients with COVID-19 through radiomics and formal methods

Background-Aim: The inflammatory cascade in patients (pts) with COVID-19 may lead to pulmonary embolism (PE), worsening prognosis. Lung perfusion SPECT/CT (Q-scan) in symptomatic pts discharged after COVID-19 can confirm or rule out pulmonary vascular involvement, helping the differential diagnosis with other respiratory diseases. We aim to investigate an innovative methodology, based on radiomic features and formal methods, as a virtual second look able to detect perfusion abnormalities to better define appropriate patient-centered diagnostic and therapeutic strategies. Methods: A total of 23 pts with a recent history of COVID-19, without any previous pulmonary disease (e.g. lung cancer, emphysema, or pathological findings at CT such as lung bullae) were enrolled for Q-scan for persistent dyspnea 1 month after discharge. They were classified as negative (14 pts) and positive (9 pts) for lung perfusion abnormalities by visual and semiquantitative analysis. Q-Lung® software by GE Healthcare was used to obtain percent evaluation of pulmonary lobar perfusion (cts/volume % for each lobe), assuming as a normal value any defect lower than 10% for each lobe. We analysed these data using an innovative methodology based on formal methods techniques centered on mathematical logical reasoning, to build a formal and rigorous representation of a system merging patients clinical conditions and disease-specific characteristics, to confirm or exclude the disease. Results: In a comparative analysis with Q-Scan results, the model showed concordant features in 13/23 pts, identifying perfusion defects in 8/9 pts with a positive Q-Scan, and excluding perfusion defects in 5/14 pts with a negative Q-Scan. Discordant results were observed in the remaining 10/23 pts, in particular in negative pts: however, in this sub-group, the Q-Lung semiquantitative analysis revealed perfusion defects lower than 10% per lobe, which we considered unsignificant but may deserve further evaluation. Conclusions: Although our data are still preliminary and based on a limited population, this methodology based on formal methods showed promising concordance with Q-scan results and needs to be implemented with further analyses including co-registered CT data. When compared to artificial intelligence techniques, this mathematical reasoning may enable (i) to use a reduced dataset of patients and/ or images, without having any impact on the robustness of the model;(ii) to produce an intuitive model easy to understand;(iii) to represent a rigorous and formal tool that may be used by medical specialists in a clinical setting.

A Proposal to Ensure Social Distancing with Deep Learning-based Object Detection

Social distancing is becoming really important in last month as a vehicle to limit the COVID-19 Coronavirus pandemic. Generally speaking, it is effective to control the spread of contagious diseases. In this context there is the need to monitor social distancing violations: for this reason in this paper we propose a social distancing detector able to count the violations by analysing video streams. Preliminary results show that the proposed method can be employed to guarantee social distancing. Moreover we discuss several suggestions aimed to improve the following proposal.