scholarly journals Audio MFCC-gram Transformers for respiratory insufficiency detection in COVID-19

2021 ◽  
Author(s):  
Marcelo Matheus Gauy ◽  
Marcelo Finger
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Respiratory Insufficiency ◽  
Network Architectures ◽  
Acoustic Model ◽  
Neural Network Architectures ◽  
Improved Performance

This work explores speech as a biomarker and investigates the detection of respiratory insufficiency (RI) by analyzing speech samples. Previous work [Casanova et al. 2021] constructed a dataset of respiratory insufficiency COVID-19 patient utterances and analyzed it by means of a convolutional neural network achieving an accuracy of 87.04%, validating the hypothesis that one can detect RI through speech. Here, we study how Transformer neural network architectures can improve the performance on RI detection. This approach enables construction of an acoustic model. By choosing the correct pretraining technique, we generate a self-supervised acoustic model, leading to improved performance (96.53%) of Transformers for RI detection.

scholarly journals Calcification detection using convolutional neural network architectures in Intravascular ultrasound images

2020 ◽  
Vol 17 (3) ◽  
pp. 1313
Author(s):  
Hannah Sofian ◽  
Joel Than Chia Ming ◽  
Suraya Muhammad ◽  
Norliza Mohd Noor
Keyword(s):  
Neural Network ◽  
Cardiovascular Disease ◽  
Coronary Artery ◽  
Convolutional Neural Network ◽  
Intravascular Ultrasound ◽  
Predictive Value ◽  
Area Under Curve ◽  
Ultrasound Image ◽  
Network Architectures ◽  
Neural Network Architectures

<p>Cardiovascular disease is the highest leading to death for Non-Communicable disease. Coronary artery calcification disease is part of cardiovascular disease. The built-in of the plaques and the calcification in the coronary artery inner wall make the blood vessel cross-section area narrow. The standard practice by the radiologists and medical clinical are by visual inspection to detect the calcification in the intravascular ultrasound image. Deep learning is the current image processing methods that have high potential to detect calcification analysis using convolutional neural network architecture and classifiers. To detect the absence of calcification and presence calcification on the intravascular ultrasound image, using k-fold =10, we compared the three types of convolutional neural network architectures and the seven types of classifiers with the provided ground truth from MICCAI 2011. We used two types of images named as Cartesian Coordinates image and polar reconstructed coordinate image. The classifiers such as Support Vector Machine, Discriminant analysis, Ensembles and Error-Correcting Output Codes obtained the perfect result with value one for Area Under Curve and all the performance measure result, accuracy, sensitivity, specificity, positive predictive value and negative predictive value. Area Under Curve for Naïve Bayes classifier is 0.9967 and for Decision Tree classifier is 0.9994, obtained using the polar reconstructed coordinate image for InceptionresNet-V2 architecture.</p>

Sign in / Sign up

Export Citation Format

Share Document