An ensemble learning approach to digital corona virus preliminary screening from cough sounds

Abstract

This work develops a robust classifier for a COVID-19 pre-screening model from crowdsourced cough sound data. The crowdsourced cough recordings contain a variable number of coughs, with some input sound files more informative than the others. Accurate detection of COVID-19 from the sound datasets requires overcoming two main challenges (i) the variable number of coughs in each recording and (ii) the low number of COVID-positive cases compared to healthy coughs in the data. We use two open datasets of crowdsourced cough recordings and segment each cough recording into non-overlapping coughs. The segmentation enriches the original data without oversampling by splitting the original cough sound files into non-overlapping segments. Splitting the sound files enables us to increase the samples of the minority class (COVID-19) without changing the feature distribution of the COVID-19 samples resulted from applying oversampling techniques. Each cough sound segment is transformed into six image representations for further analyses. We conduct extensive experiments with shallow machine learning, Convolutional Neural Network (CNN), and pre-trained CNN models. The results of our models were compared to other recently published papers that apply machine learning to cough sound data for COVID-19 detection. Our method demonstrated a high performance using an ensemble model on the testing dataset with area under receiver operating characteristics curve = 0.77, precision = 0.80, recall = 0.71, F1 measure = 0.75, and Kappa = 0.53. The results show an improvement in the prediction accuracy of our COVID-19 pre-screening model compared to the other models.

Attention-based hybrid CNN-LSTM and spectral data augmentation for COVID-19 diagnosis from cough sound

“…For instance, someone who suffers from dry cough will have more chance to be classified as Likely-COVID-19 and then, more chance to isolate the patient. Model performance and sensitivity : Our model was able to correctly classify more than 91% of unseen data, and also correctly identified more than 90% of Likely-COVID-19 cases, outperforming some previous studies which worked on different datasets (Brown et al, 2020 ; Mohammed et al, 2021 ; Park et al, 2019 ). However, the error rate for classifying positive cases as negative is 9.07%.…”

“…The testing resulted in 80% and 72% for precision and recall, respectively and 82% ROC-AUC. In another work, Mohamed et al (Mohammed et al, 2021 ) proposed to ensemble a CNN model trained from scratch, VGG16 and Tuned-VGG16 to classify cough sounds as COVID-19 positive or negative cases. The authors collected 20min and 4s for positive class and 4 hours, 30 min and 15 seconds for negative class.…”

“… Sensitiv. AUC (Lella & Pja, 2022 ) Subset of UC Signal/NU 0.9545 NA NA (Imran et al, 2020 ) Privately collected NU/NU 0.9285 0.9457 NA (Schuller et al, 2021 ) Subsets of UC NU/NU UAR (0.739) NA NA (Brown et al, 2020 ) UC Signal/NU NA 0.81 0.88 (Mohammed et al, 2021 ) Collected from Github Signal/NU NA 0.71 0.77 (Pahar et al, 2021 ) Coswara + Sarcos Signal/SMOTE 0.9533 0.91 0.938 (Tena et al, 2022 ) UC+UL+Coswara+ Pertussis+Virufy NU/NU 0.8979 0.9381 0.9604 (Harvill et al, 2021 ) COUGHVID DiCOVA Spectrogram/NU NA NA 0.7683 0.8535 (Xue & Salim, 2021 ) Coswara + UC …”

A comprehensive survey on the biomedical signal processing methods for the detection of COVID-19

“…It is most important to have an easy tool for diagnosing, screening and supervising the virus and its proliferation. An automatic method is used for detecting and monitoring the presence of COVID-19 or its symptoms are developed using AI (artificial intelligence) based approaches [ 5 ]. Many AI techniques using speech and other audio models having many opportunities in this space [ 6 ].…”

Application of Decision Tree‐Based Classification Algorithm on Content Marketing

“…e first part refers to supervised learning-based methods that require an independent training process and predict testing data using an already learned model. e representative supervised models include sparse/collaborative representation [8][9][10], support vector machine [11][12][13], ensemble learning [14][15][16][17], and so on. e second part refers to unsupervised learningbased models that do not demand training samples and determine entire classes by considering the correlations among samples.…”