[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
4.0
3.8
Journals
Bioengineering
Special Issues
Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing
share announcement

Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing

A special issue of Bioengineering (ISSN 2306-5354). This special issue belongs to the section "Biosignal Processing".

Deadline for manuscript submissions: 30 April 2025 | Viewed by 5225

Special Issue Editor

Prof. Dr. Zvi Friedman

E-Mail Website
Guest Editor
Faculty of Biomedical Engineering, Technion-Israel Institute of Technology, Haifa, Israel
Interests: biomedical engineering; ultrasound imaging; biomedical imaging; biomedical signal processing

Special Issue Information

Dear Colleagues,

Biomedical imaging is of great importance in medical diagnosis. The fast, accurate, and rapid detection of particular illnesses is critical. As an example, the detection of cancerous tumors in their early stages may result in proper treatment of the disease. The tremendous advancements in artificial intelligence over recent decades have dramatically changed the field of biomedical image processing and diagnosis based on medical imaging. Nowadays, artificial intelligence is extensively used in biomedical imaging; for example, AI is used for image classification, image segmentation, image retrieval, and image fusion for various types of medical images such as X-rays, MRIs, and CT scans.

In this Special Issue of Bioengineering, Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing, we invite submissions of original research papers and comprehensive surveys that explore the application of artificial intelligence in medical image processing. The major topics of interest for this Special Issue include (but are not limited to):

  • Medical image segmentation;
  • Medical image classification;
  • Knowledge extraction from medical images;
  • Novel architectures for the application of deep learning in medical image processing;
  • Pattern recognition in biomedical signals;
  • Application of AI in image-guided surgery;
  • Application of AI in medical diagnosis (e.g., cancerous tumor detection and skin lesion classification);
  • Biomedical image retrieval;
  • Biomedical image fusion;
  • Biomedical image watermarking.

Prof. Dr. Zvi Friedman
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Bioengineering is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2700 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • artificial intelligence
  • biomedical imaging
  • biomedical signal processing
  • machine learning
  • deep learning
  • image segmentation
  • image classification
  • image retrieval
  • knowledge extraction

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue polices can be found here.

Published Papers (4 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

17 pages, 6810 KiB  
Article
Breast Tumor Detection and Diagnosis Using an Improved Faster R-CNN in DCE-MRI
by Haitian Gui, Han Jiao, Li Li, Xinhua Jiang, Tao Su and Zhiyong Pang
Bioengineering 2024, 11(12), 1217; https://doi.org/10.3390/bioengineering11121217 - 1 Dec 2024
Viewed by 598
Abstract
AI-based breast cancer detection can improve the sensitivity and specificity of detection, especially for small lesions, which has clinical value in realizing early detection and treatment so as to reduce mortality. The two-stage detection network performs well; however, it adopts an imprecise ROI [...] Read more.
AI-based breast cancer detection can improve the sensitivity and specificity of detection, especially for small lesions, which has clinical value in realizing early detection and treatment so as to reduce mortality. The two-stage detection network performs well; however, it adopts an imprecise ROI during classification, which can easily include surrounding tumor tissues. Additionally, fuzzy noise is a significant contributor to false positives. We adopted Faster RCNN as the architecture, introduced ROI aligning to minimize quantization errors and feature pyramid network (FPN) to extract different resolution features, added a bounding box quadratic regression feature map extraction network and three convolutional layers to reduce interference from tumor surrounding information, and extracted more accurate and deeper feature maps. Our approach outperformed Faster R-CNN, Mask R-CNN, and YOLOv9 in breast cancer detection across 485 internal cases. We achieved superior performance in mAP, sensitivity, and false positive rate ((0.752, 0.950, 0.133) vs. (0.711, 0.950, 0.200) vs. (0.718, 0.880, 0.120) vs. (0.658, 0.680, 405)), which represents a 38.5% reduction in false positives compared to manual detection. Additionally, in a public dataset of 220 cases, our model also demonstrated the best performance. It showed improved sensitivity and specificity, effectively assisting doctors in diagnosing cancer. Full article
(This article belongs to the Special Issue Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>Flowchart of the study procedure.</p> Full article ">Figure 2
<p>The architecture of our proposed model BC R-CNN.</p> Full article ">Figure 3
<p>PDN structure.</p> Full article ">Figure 4
<p>Four-quadrant location.</p> Full article ">Figure 5
<p>Background noise reduction: (<b>a</b>) MRI before noise reduction; (<b>b</b>) MRI after noise reduction; (<b>c</b>) segmented breast of original MRI; (<b>d</b>) segmented breast of noise reduction MRI.</p> Full article ">Figure 6
<p>U-Net++ breast edge segmented: (<b>a</b>) sagittal breast MRI and single breast MRI at axial plane; (<b>b</b>) masks; (<b>c</b>) segmented breasts.</p> Full article ">Figure 7
<p>AUC performance comparison of different models: (<b>a</b>) internal dataset, (<b>b</b>) public dataset.</p> Full article ">Figure 8
<p>Breast tumor location and diagnosis comparison of Faster R-CNN and our proposed model: (<b>a1</b>,<b>b1</b>,<b>c1</b>) detected by Faster R-CNN; (<b>a2</b>,<b>b2</b>,<b>c2</b>) detected by our proposed model. (<b>a1</b>,<b>b1</b>) false positive; (<b>c1</b>) diagnosed with a lower score; (<b>c2</b>) diagnosed at a higher score.</p> Full article ">
17 pages, 3475 KiB  
Article
Machine Learning Models for Tracking Blood Loss and Resuscitation in a Hemorrhagic Shock Swine Injury Model
by Jose M. Gonzalez, Ryan Ortiz, Lawrence Holland, Austin Ruiz, Evan Ross and Eric J. Snider
Bioengineering 2024, 11(11), 1075; https://doi.org/10.3390/bioengineering11111075 - 27 Oct 2024
Viewed by 731
Abstract
Hemorrhage leading to life-threatening shock is a common and critical problem in both civilian and military medicine. Due to complex physiological compensatory mechanisms, traditional vital signs may fail to detect patients’ impending hemorrhagic shock in a timely manner when life-saving interventions are still [...] Read more.
Hemorrhage leading to life-threatening shock is a common and critical problem in both civilian and military medicine. Due to complex physiological compensatory mechanisms, traditional vital signs may fail to detect patients’ impending hemorrhagic shock in a timely manner when life-saving interventions are still viable. To address this shortcoming of traditional vital signs in detecting hemorrhagic shock, we have attempted to identify metrics that can predict blood loss. We have previously combined feature extraction and machine learning methodologies applied to arterial waveform analysis to develop advanced metrics that have enabled the early and accurate detection of impending shock in a canine model of hemorrhage, including metrics that estimate blood loss such as the Blood Loss Volume Metric, the Percent Estimated Blood Loss metric, and the Hemorrhage Area metric. Importantly, these metrics were able to identify impending shock well before traditional vital signs, such as blood pressure, were altered enough to identify shock. Here, we apply these advanced metrics developed using data from a canine model to data collected from a swine model of controlled hemorrhage as an interim step towards showing their relevance to human medicine. Based on the performance of these advanced metrics, we conclude that the framework for developing these metrics in the previous canine model remains applicable when applied to a swine model and results in accurate performance in these advanced metrics. The success of these advanced metrics in swine, which share physiological similarities to humans, shows promise in developing advanced blood loss metrics for humans, which would result in increased positive casualty outcomes due to hemorrhage in civilian and military medicine. Full article
(This article belongs to the Special Issue Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>The ML development pathway for one of the four ML models. The ML training data was split, resulting in an ML model (blue) after undergoing the training pipeline. The ML model (blue) is then input into the testing stage, resulting in ML model predictions for three swine completely left out of the training stage. The process displayed is repeated a total of four times with different arrangements of the four groups, with each group completely left out for testing once.</p> Full article ">Figure 2
<p>(<b>a</b>) ML model developed using non-detrended data tested on features extracted from non-detrended data. (<b>b</b>) ML model developed using non-detrended data tested on features extracted from detrended data.</p> Full article ">Figure 3
<p>(<b>a</b>) ML model developed using detrended data tested on features extracted from detrended data. (<b>b</b>) ML model developed using detrended data tested on features extracted from non-detrended data.</p> Full article ">Figure 4
<p>(<b>a</b>) ML model developed using unfiltered data for the prediction of PEBL. (<b>b</b>) ML model developed using unfiltered data for the prediction of HemArea.</p> Full article ">Figure 5
<p>(<b>a</b>) ML model trained on human data, tested on swine for CRM predictions. (<b>b</b>) DL model trained on human data, tested on swine for CRM predictions.</p> Full article ">Figure 6
<p>Comparative performance of each predictive model for (<b>a</b>) Area under the ROC curve, (<b>b</b>) Time to consistent hemorrhage prediction (note: logarithmic scale on <span class="html-italic">y</span>-axis), and (<b>c</b>) Ratio of metric value after resuscitation to during hemorrhage to create a signal-to-noise metric (note: logarithmic scale on <span class="html-italic">y</span>-axis). (<b>d</b>) Comparison of coefficient of variation (%) for the signal-to-noise resuscitation to hemorrhage ratios across each trained model.</p> Full article ">
14 pages, 7020 KiB  
Article
Automated Restarting Fast Proximal Gradient Descent Method for Single-View Cone-Beam X-ray Luminescence Computed Tomography Based on Depth Compensation
by Peng Gao, Huangsheng Pu, Tianshuai Liu, Yilin Cao, Wangyang Li, Shien Huang, Ruijing Li, Hongbing Lu and Junyan Rong
Bioengineering 2024, 11(2), 123; https://doi.org/10.3390/bioengineering11020123 - 26 Jan 2024
Viewed by 1223
Abstract
Single-view cone-beam X-ray luminescence computed tomography (CB-XLCT) has recently gained attention as a highly promising imaging technique that allows for the efficient and rapid three-dimensional visualization of nanophosphor (NP) distributions in small animals. However, the reconstruction performance is hindered by the ill-posed nature [...] Read more.
Single-view cone-beam X-ray luminescence computed tomography (CB-XLCT) has recently gained attention as a highly promising imaging technique that allows for the efficient and rapid three-dimensional visualization of nanophosphor (NP) distributions in small animals. However, the reconstruction performance is hindered by the ill-posed nature of the inverse problem and the effects of depth variation as only a single view is acquired. To tackle this issue, we present a methodology that integrates an automated restarting strategy with depth compensation to achieve reconstruction. The present study employs a fast proximal gradient descent (FPGD) method, incorporating L0 norm regularization, to achieve efficient reconstruction with accelerated convergence. The proposed approach offers the benefit of retrieving neighboring multitarget distributions without the need for CT priors. Additionally, the automated restarting strategy ensures reliable reconstructions without the need for manual intervention. Numerical simulations and physical phantom experiments were conducted using a custom CB-XLCT system to demonstrate the accuracy of the proposed method in resolving adjacent NPs. The results showed that this method had the lowest relative error compared to other few-view techniques. This study signifies a significant progression in the development of practical single-view CB-XLCT for high-resolution 3−D biomedical imaging. Full article
(This article belongs to the Special Issue Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Flowchart of the re-DC-FPGD method.</p> Full article ">Figure 2
<p>Schematic of the numerical simulations: (<b>a</b>) 3−D digital mouse with main organs and tumors (the investigated region was 2.6 cm in height), where the black circles depict the central slices of the tumors (represented in red); (<b>b</b>) CB-XLCT imaging system with initial phantom position setup, where the red line depicts the tomographic outline of the mouse and the two blue circles depict the tumors with NPs.</p> Full article ">Figure 3
<p>Setup of the phantom experiments: (<b>a</b>,<b>b</b>) representative X-ray projections of the phantom in case 1 and case 2. Regions between the blue and green lines are used for the study; (<b>c</b>,<b>d</b>) CT slices indicated by the red lines shown in (<b>a</b>,<b>b</b>).</p> Full article ">Figure 4
<p>Reconstructed images of the tumors with NPs at different imaging views. The slices are those indicated with black circles in <a href="#bioengineering-11-00123-f002" class="html-fig">Figure 2</a>. The white circles represent the real positions of the tumors. The red, green, and magenta lines represent the boundaries of the animal’s body, bones, and liver, respectively: (<b>a</b>–<b>c</b>) imaging views of 0°, 30°, 60°, respectively. All images were normalized to the maximal value.</p> Full article ">Figure 5
<p>Reconstructed images of the tumors in different cases using the three methods. The white circles represent the real positions of the tumors. The red, green, and magenta lines represent the boundaries of the animal’s body, bones, and liver, respectively (<b>a</b>–<b>c</b>) tomographic slices reconstructed by T-FISTA, DC-FL, and the proposed re-DC-FPGD algorithm with an EED of 0.3 cm. (<b>d</b>–<b>f</b>) tomographic slices reconstructed by different algorithms with an EED of 0.2 cm. (<b>g</b>–<b>i</b>) tomographic slices reconstructed by different algorithms with an EED of 0.1 cm.</p> Full article ">Figure 6
<p>The 3−D results of the reconstructed CB-XLCT images using the proposed method in the simulations of the three cases: (<b>a</b>–<b>c</b>) case 1 to case 3. Tumors are represented in blue.</p> Full article ">Figure 7
<p>Phantom experiment results for case 1: (<b>a</b>–<b>c</b>) CB-XLCT slices reconstructed by T-FISTA, DC-FL, and the proposed re-DC-FPGD algorithm, respectively. (<b>d</b>–<b>f</b>) CB-XLCT/CT fusion results of different methods. (<b>g</b>–<b>i</b>) 3−D rendering results of different methods. The red circles in the CB-XLCT images depict the boundaries of the phantom. The blue objects in the 3−D renderings represent the recovered targets.</p> Full article ">Figure 8
<p>Phantom experiment results for case 2: (<b>a</b>–<b>c</b>) CB-XLCT slices reconstructed by T-FISTA, DC-FL, and the proposed re-DC-FPGD algorithm, respectively. (<b>d</b>–<b>f</b>) CB-XLCT/CT fusion results of different methods. (<b>g</b>–<b>i</b>) 3−D rendering results of different methods. The red circles in the CB-XLCT images depict the boundaries of the phantom. The blue objects in the 3−D renderings represent the recovered targets.</p> Full article ">
17 pages, 4006 KiB  
Article
CL-SPO2Net: Contrastive Learning Spatiotemporal Attention Network for Non-Contact Video-Based SpO2 Estimation
by Jiahe Peng, Weihua Su, Haiyong Chen, Jingsheng Sun and Zandong Tian
Bioengineering 2024, 11(2), 113; https://doi.org/10.3390/bioengineering11020113 - 24 Jan 2024
Cited by 3 | Viewed by 1889
Abstract
Video-based peripheral oxygen saturation (SpO2) estimation, utilizing solely RGB cameras, offers a non-contact approach to measuring blood oxygen levels. Previous studies set a stable and unchanging environment as the premise for non-contact blood oxygen estimation. Additionally, they utilized a small amount of labeled [...] Read more.
Video-based peripheral oxygen saturation (SpO2) estimation, utilizing solely RGB cameras, offers a non-contact approach to measuring blood oxygen levels. Previous studies set a stable and unchanging environment as the premise for non-contact blood oxygen estimation. Additionally, they utilized a small amount of labeled data for system training and learning. However, it is challenging to train optimal model parameters with a small dataset. The accuracy of blood oxygen detection is easily affected by ambient light and subject movement. To address these issues, this paper proposes a contrastive learning spatiotemporal attention network (CL-SPO2Net), an innovative semi-supervised network for video-based SpO2 estimation. Spatiotemporal similarities in remote photoplethysmography (rPPG) signals were found in video segments containing facial or hand regions. Subsequently, integrating deep neural networks with machine learning expertise enabled the estimation of SpO2. The method had good feasibility in the case of small-scale labeled datasets, with the mean absolute error between the camera and the reference pulse oximeter of 0.85% in the stable environment, 1.13% with lighting fluctuations, and 1.20% in the facial rotation situation. Full article
(This article belongs to the Special Issue Artificial Intelligence in Biomedical Imaging and Biomedical Signal Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>Proposed SpO2 analysis architecture using CL-SPO2Net. Small blocks with various colors represent the features extracted at running steps of the network. The light pink area represents the 3DCNN, within which we combine unsupervised contrastive learning with supervised label learning to obtain accurate rPPG signals for remote videos. The light green area represents the spatio-temporal feature extraction step of the signal. The results obtained by the CNN-BiLSTM network, alongside those from an attention module, are inputted into a fusion function to yield the estimation results of blood oxygen saturation.</p> Full article ">Figure 2
<p>The absorption degree of Hb and HbO2 to light of different wavelengths. The blue, green, and red dashed lines denote the wavelengths of light corresponding to their respective colors.</p> Full article ">Figure 3
<p>The hypothesis of Similar PSD of facial and hand rPPG signals. Each part is divided into four ROIs, in which the human hand is divided into four regions A–D according to the center line, and the facial part is divided into four regions E–H according to the center line. Set <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mfenced close="}" open="{"> <mrow> <mi>A</mi> <mo>,</mo> <mi>B</mi> <mo>,</mo> <mi>C</mi> <mo>,</mo> <mi>D</mi> </mrow> </mfenced> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>V</mi> <mo>=</mo> <mfenced close="}" open="{"> <mrow> <mi>E</mi> <mo>,</mo> <mi>F</mi> <mo>,</mo> <mi>G</mi> <mo>,</mo> <mi>H</mi> </mrow> </mfenced> </mrow> </semantics></math>.</p> Full article ">Figure 4
<p>Combination of contrastive learning and supervised learning. Firstly, during the experimental procedure, video segments designated as <math display="inline"><semantics> <mrow> <msub> <mi>X</mi> <mi>i</mi> </msub> <mo stretchy="false">(</mo> <mi>a</mi> <mo stretchy="false">)</mo> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>X</mi> <mi>i</mi> </msub> <mo stretchy="false">(</mo> <mi>b</mi> <mo stretchy="false">)</mo> </mrow> </semantics></math> were acquired, capturing the facial and hand regions of a participant who maintained a stationary posture throughout the recording phase. The video clip <math display="inline"><semantics> <mrow> <msub> <mi>X</mi> <mi>i</mi> </msub> <mo stretchy="false">(</mo> <mi>c</mi> <mo stretchy="false">)</mo> </mrow> </semantics></math> constituted a regenerated sequence derived from video <math display="inline"><semantics> <mrow> <msub> <mi>X</mi> <mi>i</mi> </msub> <mo stretchy="false">(</mo> <mi>a</mi> <mo stretchy="false">)</mo> </mrow> </semantics></math>, wherein a single frame was extracted and subjected to chromatic data augmentation. The video segment <math display="inline"><semantics> <mrow> <msub> <mi>X</mi> <mi>i</mi> </msub> <mo stretchy="false">(</mo> <mi>d</mi> <mo stretchy="false">)</mo> </mrow> </semantics></math> represented a subject executing a facial rotation maneuver. Time series <math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mi>a</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mi>b</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mi>c</mi> </msub> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mi>d</mi> </msub> </mrow> </semantics></math> are the tensor form of video clips of time length <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>/</mo> <mn>2</mn> </mrow> </semantics></math>. Signals <math display="inline"><semantics> <mrow> <msub> <mi>y</mi> <mi>a</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>y</mi> <mi>b</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>y</mi> <mi>c</mi> </msub> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <msub> <mi>y</mi> <mi>d</mi> </msub> </mrow> </semantics></math> are rPPG signals produced by an rPPG estimator utilizing a 3DCNN architecture, subsequently transformed via Fourier analysis to generate PSD vectors <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi>a</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi>b</mi> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi>c</mi> </msub> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi>d</mi> </msub> </mrow> </semantics></math>, where <math display="inline"><semantics> <mrow> <msub> <mover accent="true"> <mi>y</mi> <mo stretchy="false">˜</mo> </mover> <mn>1</mn> </msub> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mover accent="true"> <mi>y</mi> <mo stretchy="false">˜</mo> </mover> <mn>2</mn> </msub> </mrow> </semantics></math> are the real PPG signals collected from the same object at the time of first <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>/</mo> <mn>2</mn> </mrow> </semantics></math> and last <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>/</mo> <mn>2</mn> </mrow> </semantics></math>.</p> Full article ">Figure 5
<p>The CNN-BiLSTM architecture.</p> Full article ">Figure 6
<p>Example images from the datasets used in this paper: FaceForensics++ (<b>a</b>), UBFC-rPPG (<b>b</b>) and MVPD (<b>c</b>).</p> Full article ">Figure 7
<p>The test results of the CL-SPO2Net on 2 participants. (<b>a</b>,<b>b</b>) show the face test results; (<b>c</b>,<b>d</b>) show the hand test results. This network architecture performs SpO2 estimation at 0.5 Hz on the computer server with a Quadro RTX3080 GPU.</p> Full article ">Figure 8
<p>Comparison of raw physiological signals and signals generated by rPPG contrast learning strategy based on 3DCNN. (<b>a</b>) shows the raw signals of the red, green, and blue channels; (<b>b</b>) shows the time series of six channels, each represented by different colors.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-4
Back to TopTop