Pneumonia is the leading cause of death from infection in children worldwide. Auscultation has unique advantages in the preliminary screening and follow-up examination of pneumonia due to its non-invasive and radiation-free nature. However, the accuracy of auscultation largely depends on the clinician's experience. Therefore, this study utilizes largescale surface respiratory sound signals for pneumonia screening and visualization to assist doctors in clinical diagnosis. A Butterworth filter is applied to the respiratory sound signals to reduce interference from heart sounds and environmental noise. Short-time Fourier transform (STFT) and Mel filter banks are used to obtain spectrograms and Mel-spectrograms of the respiratory sound signals. These spectrograms and Mel-spectrograms are fed into a convolutional neural network (CNN) to extract deep features and perform features fusion. A support vector machine (SVM) is used as a classifier for binary detection of the respiratory sound signals. Finally, the detection results are visualized in combination with multi-channel positional information. The computational results demonstrate that the predicted pneumonia lesion areas are consistent with the actual lesion areas, achieving the detection rate for pneumonia is 100%, the mean TTAR is 0.32, and the mean TPAR is 0.29. This study employs large-scale respiratory sound signals for the visualization of pneumonia lesion areas, enhancing the intuitiveness of diagnosis. It has significant advantage and potential in environments with limited medical resources and in follow-up examinations for pneumonia.
Hypertension, as an important risk factor for cardiovascular diseases, threatens the lives of adults worldwide every year. Therefore, continuous blood pressure (BP) monitoring is necessary for the prevention and early diagnosis of hypertension. To achieve cuffless BP monitoring, an end-to-end BP waveforms prediction model was implemented using photoplethysmography (PPG) and deep learning in this paper. The proposed model successfully predicted three blood pressure values: systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) from PPG waveforms using U-Net, and their mean absolute errors (MAE) and standard deviations (STD) were in the range of 4.54 ±7.42 mmHg, 2.47 ±4.62 mmHg, 1.67 ±3.65 mmHg, respectively, compared to the reference BP values. This provides a possibility to realize non-invasive continuous blood pressure monitoring in daily life through portable wearable devices.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.