Assessment of wearable Ultrasound Device Combined with AI for Portable Assessment of Central Venous Pressure Compared with Central Venous Catheterization as the reference Standard in critically ill patients: A Cross-Sectional Study

Assessment of wearable Ultrasound Device Combined with AI for Portable Assessment of Central Venous Pressure Compared with Central Venous Catheterization as the reference Standard in critically ill patients: A Cross-Sectional Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Assessment of wearable Ultrasound Device Combined with AI for Portable Assessment of Central Venous Pressure Compared with Central Venous Catheterization as the reference Standard in critically ill patients: A Cross-Sectional Study Liping Dong, Meng Li, Yi Li, Xingxuan Zhang, Jingyi Guo, Shaorong Lu, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6316972/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose Currently, central venous catheterization (CVC) remains the reference standard for evaluating central venous pressure (CVP) in critically ill patients, but its invasiveness and associated complications limit its use. This study combined a wearable ultrasound device with artificial intelligence (AI) models to achieve portable noninvasive assessment of CVP in critically ill patients. Methods This study prospectively enrolled critically ill patients who underwent CVC. We divided participants into a CVP≥8 mmHg group and a CVP<8 mmHg group, and recorded the internal jugular vein (IJV) and common carotid artery (CCA) via wearable ultrasound device. A Dual-Decoder Spatiotemporal Attention Network (DSTA-Net) was trained to enable automatic measurement of wearable ultrasound data and was evaluated via Bland-Altman consistency and Spearman correlation analyses. Moreover, it was compared with existing DeepLabV3+, Swin-Unet, TransUNet, and UNet models. The baseline patient parameters were subsequently combined with the wearable ultrasound parameters to construct a dual-mixing multilayer perception (DM-MLP) model to forecast elevated CVP,which was compared with existing VGG, ResNet, and Transformer models. Diagnostic performance was assessed via receiver operating characteristic (ROC) curves. Results Videos from 272 patients (218:54 [80%:20%] in the training and test sets) were used for DSTA-Net training. Both sets were divided into the CVP≥8 mmHg and CVP<8 mmHg groups. In both sets, the CVP and the manually/DSTA-Net-measured IJV Max Area, IJV Min Area, and IJV Max/CCA Area were greater in the CVP≥8mmHg group. Compared with the DeepLabV3+, Swin-Unet, TransUNet, and UNet models, the DSTA-Net model showed higher Dice and IoU values of 82.62±0.18 and 74.97±0.15 for IJV segmentation, and 80.09±0.19 and 71.71±0.17 for CCA segmentation, respectively. Bland-Altman and Spearman correlation analyses confirmed a high degree of consistency and correlation between manually- and DSTA-Net- measured wearable ultrasound parameters. The ROC curve AUCs of the clinical parameters and manually/DSTA-Net measured DM-MLP models were 0.94[0.88,0.94] and 0.88[0.85,0.91], respectively, indicating better predictive performance than the VGG, ResNet, and Transformer models. Conclusion These results suggest that integrating wearable ultrasound devices with DSTA-Net as an image segmentation model and DM-MLP as a clinical prediction model provides a portable, noninvasive method for predicting elevated CVP. Central venous pressure Dual-Decoder Spatiotemporal Attention Network Dual-Mixing MLP Internal jugular vein Common carotid artery Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Background Elevated central venous pressure (CVP) and insufficient CVP reduction are associated with adverse events and mortality in critically ill patients[ 1 – 3 ], indicating the need to identify elevated CVP. Central venous catheter (CVC), a complex, time-consuming, and complicated invasive procedure is the reference standard for CVP assessment[ 4 , 5 ]. Complications such as infection, thrombus and bleeding, pneumothorax, and hemothorax may occur during CVC placement, limiting its clinical application[ 6 ]. Given these challenges, it is imperative to explore noninvasive methods for assessing CVP. Recently, the internal jugular vein (IJV), due to its direct connection to the right atrium through venous access, has been shown to directly reflect right atrial pressure and could be used in CVP evaluation[ 7 – 9 ]. Ultrasound is a noninvasive, portable technique commonly used to assess neck blood vessels. Traditional ultrasound probes are bulky and require constant hand-holding or a fixed machine for continuous monitoring. However, the prolonged pressure of bulky probes on critically ill patients' necks could cause discomfort and even suffocation[ 10 ]. Several studies have reported that wearable ultrasound devices could be portable, for real-time monitoring of internal organs and related conditions, including blood vessels, muscles, heart, gastrointestinal tract, diaphragm, and lung[ 11 ]. It could also facilitate continuous real-time imaging of the left ventricle during physical activity[ 12 , 13 ]. Compared with traditional ultrasound, wearable ultrasound is more compact and portable, making it suitable for organ monitoring in pre-hospital emergency settings and intensive care units (ICUs)[ 10 ]. Nevertheless, manual measurement is still required, which increases the operation time and cost. Combining wearable ultrasound devices with artificial intelligence (AI) enables real-time measurement, significantly reducing diagnostic and treatment time, which is particularly crucial for critically ill patients. Reports on wearable ultrasound devices integrated with AI to evaluate the IJV for predicting CVP are lacking. Herein, we used a wearable ultrasound device combined with AI to achieve real-time dynamic measurement of the IJV and CCA, accurately predicting elevated CVP in critically ill patients, providing a novel approach for noninvasive CVP assessment, which could significantly improve hemodynamic monitoring, provide timely and reliable healthcare decision support, and reduce mortality. Materials and methods Study population This prospective, single-center, stratified, cross-sectional clinical trial was performed in accordance with the Declaration of Helsinki, was approved by the institutional review board of Shanghai Sixth People's Hospital (institutional review board number: 2023-007-(3); approval date: August 30, 2023), and registered in the Chinese Clinical Trial Registry ( https://www.chictr.org.cn/ ; registration number: ChiCTR230007586). None of the study participants had been reported previously, and received no financial compensation. All patients or their first-degree relatives provided written informed consent. We prospectively enrolled 318 critically ill patients who received CVC, between December 1, 2023, and January 1, 2024, at the ICU Shanghai Sixth People's Hospital, China. The inclusion criteria were as follows: aged over 18 years, underwent CVP monitoring, and consecutive enrollment. The exclusion criteria, which were designed to eliminate confounding factors potentially interfering with accurate CVP measurement, included: (a) receiving positive end-expiratory pressure ventilation, (b) a history of neck radiation therapy or surgery, (c) congenital heart disease, (d) moderate to severe tricuspid regurgitation, (e) right heart dilation, (f) inability to lie flat due to unstable vital signs, (g) inability to visualize cervical vessels, and (h) poor image quality. The flowchart of patient enrolment is shown in Fig. 1 . Reference standard of central venous pressure The CVP of consenting participants was assessed by a professional emergency physician and a nursing staff, each with experience of > 100 cases of Central venous catheterization (CVC), which is a reference standard for monitoring CVP in critically ill patients. It was typically inserted into the right atrium via either the IJV or the femoral vein, and it was connected to a pressure transducer and an integrated bedside monitor. The pressure transducer was strategically positioned along the mid-axillary line at the level of the fourth intercostal space, which corresponds to the midpoint of the right atrium[ 14 – 16 ]. To accurately reflect cardiac preload, CVP measurements should be taken at the base of the “c” wave in the pressure waveform, which corresponds to the Q wave on the electrocardiogram (ECG) (Fig. 2 ). Additionally, it was essential to time these measurements correctly: CVP should be assessed at the end of expiration when the patient was breathing calmly or at the beginning of expiration during spontaneous breathing. This practice helped to mitigate the influence of intrathoracic pressure on CVP readings [ 8 , 15 ]. Following septic shock treatment guidelines, we classified critically ill patients into two CVP groups: CVP < 8 mmHg group and CVP ≥ 8 mmHg group, patients with a CVP< 8 mmHg might indicate potential hypovolemia or inadequate preload. In contrast, patients with a CVP ≥ 8 mmHg might suggest fluid overload or elevated right atrial pressures[ 14 ]. Wearable ultrasound acquisition and measurement The wearable ultrasound device, model Cloud-35LL (Stork, Chengdu, China), consists of a patch transducer and a main unit (Fig. 3 ). The detailed description of the wearable ultrasound architecture is provided in Appendix 1. Before conducting wearable ultrasound measurements, the patient for CVC should lie flat in a 0° position, ensuring that there was no pillow or support under the head. The patient’s neck was in a neutral position, gently rotated to the left (not exceeding 30°), to expose the right IJV and CCA. The solid coupling agent was placed between the patch transducer and the patient’s skin. Due to variations in neck length, the probe should be lightly placed at the junction of the middle and lower thirds of the neck. The probe was adjusted slightly to obtain the clearest circular cross-sectional ultrasound images of the IJV and CCA, while the tape secured the probe and solid coupling gel to the patient’s neck. No pressure should be applied above the patch transducer, and the tape should not be applied too tightly to avoid compressing the IJV and CCA, which may cause deformation. During the patient’s calm breathing, the sonographers operated the wearable ultrasound to record real-time ultrasound videos of the IJV and CCA cross-sections over one breathing cycle, saving the video in mp4 format. Simultaneously, waves and data from the CVP on the monitor were recorded (Fig. 2 ). All manual measurements were performed by two professional sonographers (with 3 and 4 years of experience), and reviewed by a chief sonographer with over 10 years of experience. The sonographers manually traced the intima-media contours of the IJV and CCA. The maximum IJV area (IJV Max Area) at the end of expiration, the minimum IJV area (IJV Min Area) at the end of inspiration, CCA area at the end of expiration, the maximum IJV area /CCA area (IJV Max/CCA Area), and change rate of the IJV (IJV Ratio) were all manually measured and images were saved. All measurements were made without knowledge of the clinical data or previous measurement results. Image segmentation model based on wearable ultrasound In this study, we creatively propose a Dual-Decoder Spatiotemporal Attention Network (DSTA-Net) image segmentation model, with a 4:1 split ratio for training and testing sets. Additionally, we compared DSTA-Net with existing models, including DeepLabV3+, Swin-Unet, TransUNet, and UNet. Ultrasound videos of 2/3 of the subjects were randomly selected to generate a continuous sequence of image frames. Key frames were extracted every 20 frames to ensure that the image data was representative. These extracted images were then manually annotated for subsequent analysis and processing. The annotation was performed via Pair software (version 3.0, Medical Ultrasound Image Computing Lab, MUSIC, Shenzhen, China), where manual marking was performed along the IJV and CCA intima on the key frames. As a result, the DSTA-Net model could automatically measure the IJV Max Area, and IJV Min Area, CCA Area, IJV Max/CCA Area, IJV Ratio over one respiratory cycle. The architecture was consisted of a shared encoder (E) and two decoders (D1 and D2), each serving distinct roles. This dual-decoder design effectively utilized the limited labeled data and optimized the use of temporal information, making it particularly well-suited for scenarios with sparse annotations. All the ultrasound images and frames were standardized to a consistent resolution of 256×256 pixels. To further enhance the model's robustness and generalizability, data augmentation techniques were applied to the small ultrasound datasets. These techniques included random rotations within ± 15 degrees, horizontal and vertical flips, and the addition of Gaussian noise (Fig. 4 a). A detailed introduction to the DSTA-Net model can be found in Appendix 2. Clinical prediction model based on wearable ultrasound Patient characteristics (age, height, weight, body surface area [BSA], body mass index [BMI], systolic blood pressure[SBP], diastolic blood pressure [DBP], heart rate [HR], breath, and wearable ultrasound parameters of IJV Max Area [Manual / DSTA-Net], IJV Min Area [Manual / DSTA-Net], IJV Ratio [Manual / DSTA-Net], CCA Area [Manual / DSTA-Net], and IJV Max/CCA Area [Manual / DSTA-Net]) were used as indicators to creatively train a DM-MLP clinical prediction model to predict elevated CVP (CVP ≥ 8 mmHg). The core of this model includes two innovative MLP modules: the attribute-mixing MLP and the case-mixing MLP (Fig. 4 b). A detailed introduction to the DM-MLP model can be found in Appendix 3. Additionally, we trained VGG, ResNet, and Transformer models using the same clinical indicators for comparison with the DM-MLP model. The same training and test sets used for the image segmentation model were used to develop the clinical prediction model. On the basis of whether the wearable ultrasound parameters were measured manually or automatically by DSTA-Net, the models were classified into two types: Manual and DSTA-Net Clinical Prediction Models. Statistical analysis Descriptive analyses and software were conducted using R (version 3.5.0), Python 3.9.1 (Python Software Foundation, http://www.python.org ), PyTorch version 1.10.0 (PyTorch.org), and CUDA 11.3 with cuDNN 8.2 (NVIDIA, https://developer.nvidia.com/cudnn ) for all computational tasks. Analyses were performed on hardware equipped with 8 NVIDIA RTX 4090 GPUs, enabling efficient parallel computation and deep learning model training. The Shapiro-Wilk and Kolmogorov-Smirnov tests were used to test for normality. Normally and non-normally distributed data are presented as the means ± standard deviations and medians and interquartile ranges (IQRs), respectively. The Mann-Whitney U test was used to evaluate the statistical significance of differences for continuous data. The accuracies of the DSTA-Net, DeepLabV3+, Swin-Unet, TransUNet, and UNet image segement models are represented by the Dice coefficient (Dice), intersection over Union (IoU), perimeter ratio (R-perimeter), and area ratio (R-area). The consistency and correlation between the DSTA-Net model and manual measurements were assessed using Bland-Altman (B-A) and Spearman correlation analyses. Using the CVC as the reference standard, the effectiveness of the DM-MLP, VGG, ResNet, and Transformer models for assessing elevated CVP was evaluated via the area under the curve (AUC) of the Receiver Operating Characteristic (ROC) curve, accuracy, sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). The significance level was set at 5%. Results Baseline patient characteristics This study initially included 318 critically ill patients who underwent CVC; however, some patients were excluded due to positive pressure ventilation (n = 10), moderate to severe tricuspid regurgitation (n = 9), right heart dilation (n = 8), inability to measure the cervical vessels due to bandaging (n = 6), and inability to lie flat (n = 13). Overall, 272 critically ill patients were included for DSTA-Net training, of whom 218 (80%) and 54 (20%) were allocated to the training and test sets, respectively. The training set was divided into the CVP ≥ 8 mmHg (n = 87; mean age 58.86 years ± 18.14 [SD]; 38 women) and CVP<8 mmHg (n = 131; mean age 52.21years ± 21.27 [SD]; 77 women) groups, as was the test set (CVP ≥ 8 mmHg group: n = 22; mean age 58.05 years ± 16.03 [SD]; 12 women; CVP<8 mmHg group: n = 32; mean age 52.47 ± 18.10[SD]; 17 women). In the training set, the systolic blood pressure (SBP) and diastolic blood pressure (DBP) were significantly greater in the CVP ≥ 8 mmHg group than that in the CVP<8 mmHg group. Additionally, in both the training and test sets, the CVP, manual and DSTA-Net IJV Max Area, IJV Min Area, and IJV Max/CCA Area were significantly greater in the CVP ≥ 8 mmHg compared to the CVP<8 mmHg group. Furthermore, the IJV Ratio, which was based on both manual and DSTA-Net measurements, was significantly elevated in the CVP ≥ 8 mmHg group of the training set. The baseline patient characteristics and wearable ultrasound parameters for the training and test sets of critically ill patients are presented in Table 1 and Appendix Table 1 . Table 1 Baseline patient characteristics and wearable ultrasound parameters for the training and test sets of critically ill patient. BSA: Body surface area, BMI: Body mass index, SBP: Systolic blood pressure, DBP: Diastolic blood pressure, HR: Heart rate, IJV Max Area: Internal jugular vein maximal area, IJV Min Area: Internal jugular vein minimal area, IJV Ratio: (IJV maximum are- IJV minimum area)/ IJV maximum are, CCA Area: Common carotid artery area at the end of expiration, IJV Max Area/CCA Area: Internal jugular vein maximal area/common carotid artery area at the end of expiration. Training set(218 patients) Test set(54 patients) P Age (y) 54.88 ± 20.30 54.74 ± 17.35 0.964 Height(cm) 169(160–173) 168(160–170) 0.319 Weight(kg) 65(60–75) 65(55–75) 0.727 BSA(m 2 ) 1.73 ± 0.20 1.70 ± 0.18 0.351 BMI (kg/m 2 ) 23.40(21.22–26.12) 23.91(20.76–25.95) 0.979 SBP(mmHg) 123(110–141) 119(108–140) 0.265 DBP(mmHg) 70(63–80) 68(59–78) 0.220 HR(bqm) 89(79–100) 93(81–100) 0.185 Breath(bpm/min) 20(18–22) 20(17–22) 0.782 Manual-IJV Max Area(mm 2 ) 131.35(85.20-186.71) 120.8(75.18-210.21) 0.507 Manual-IJV Min Area(mm 2 ) 77.73(46.70-124.58) 56.52(36.88-125.43) 0.144 Manual-IJV Ratio 30.27(16.51–54.04) 34.45(16.54–65.13) 0.320 Manual-CCA Area(mm 2 ) 57.10(44.57-73.00) 61.15(47.33–76.45) 0.323 Manual-IJV Max Area/CCA Are 2.29(1.48–3.36) 2.04(1.20–3.19) 0.251 DSTA-Net IJV Max Area(mm 2 ) 146.53(95.74–212.20) 137.3(83.39-235.38) 0.535 DSTA-Net IJV Min Area(mm 2 ) 81.77(48.93-131.75) 59.71(38.74-132.38) 0.149 DSTA-Net IJV Ratio 34.74(21.15–57.17) 40.65(21.50-67.75) 0.327 DSTA-Net CCA Area(mm 2 ) 62.40(49.26–79.67) 69.07(51.70-85.18) 0.209 DSTA-Net IJV Max Area/CCA Area 2.32(1.54–3.58) 2.16(1.26–3.30) 0.249 CVP, mmHg 6(4–9) 6(5–8) 0.841 Manual DM-MLP CVP, mmHg 6.19(4.03–9.08) 6.63(4.49–8.22) 0.824 DSTA-Net DM-MLP CVP, mmHg 6.22(3.85–9.26) 6.72(4.21–8.81) 0.648 Accuracy of the wearable ultrasound image segmentation model The DSTA-Net model achieved Dice and IoU values of 82.62 ± 0.18 and 74.97 ± 0.15 for IJV segmentation, and 80.09 ± 0.19 and 71.71 ± 0.17 for CCA segmentation, outperforming DeepLabV3+ (IJV Dice: 77.58 ± 0.23, CCA Dice: 73.85 ± 0.24, IJV IoU: 70.97 ± 0.14, CCA IoU: 65.23 ± 0.18), Swin-Unet (IJV Dice: 74.92 ± 0.23, CCA Dice: 72.62 ± 0.21, IJV IoU: 69.65 ± 0.13, CCA IoU: 66.25 ± 0.14), TransUNet (IJV Dice: 75.45 ± 0.15, CCA Dice: 72.56 ± 0.18, IJV IoU: 69.15 ± 0.13, CCA IoU: 65.18 ± 0.16), and Unet (IJV Dice: 80.22 ± 0.14, CCA Dice: 70.52 ± 0.17, IJV IoU: 70.75 ± 0.19, CCA IoU: 60.79 ± 0.12). Furthermore, in terms of metrics such as the area ratio (R-area IJV:11.81 ± 0.10, CCA:12.56 ± 0.11) and perimeter ratio (R-perimeter IJV:11.58 ± 0.10, CCA:10.55 ± 0.09), the DSTA-Net model also demonstrated better segmentation performance, with smaller R-area and R-perimeter (Table. 2). Consistency of the wearable ultrasound image segmentation model For the training and testing sets, the mean differences in the IJV Min Area between the manual and DSTA-Net measurements were − 4.82 and − 4.24, with respective LOA ranges of [-12.06, 2.42] and [-11.45, 2.97], respectively. For the IJV Ratio, the mean differences were − 3.75 and − 3.59, with LOA ranges of [-10.54, 3.04] and [-9.93, 2.76], respectively. The mean differences for IJV Max/CCA were − 0.088 and − 0.034, with LOA ranges of [-0.48, 0.30] and [-0.33, 0.26], respectively. The small differences and narrow LOA ranges indicate good agreement between the two methods in the measurements of the minimum area, minimum ratio, and maximum/CCA values. However, for the IJV Max Area, the mean differences were − 17.34 and − 16.76, with LOA ranges of [-41.84, 7.16] and [-40.73, 7.21] in the training and testing sets, respectively. For the CCA Area, the mean differences were − 5.27 and − 6.47, with LOA ranges of [-11.53, 0.99] and [-12.59, -0.35], respectively. The relatively larger mean differences and negatively skewed LOA ranges indicate that the DSTA-Net measurements were slightly lower than the manual measurements. The correlation analysis indicated a high degree of correlation between the manual and DSTA-Net measurements of the IJV Max Area (r = 0.99, p<0.01), IJV Min Area (r = 0.99, p<0.01), CCA Area (r = 0.99, p<0.01), IJV Max/CCA Area (r = 0.99, p<0.01), and IJV Ratio (r = 0.99, p<0.01) in critically ill patients(Table 3 and Fig. 5 ). Similarly, when participants were divided into the CVP < 8 mmHg and CVP ≥ 8 mmHg groups, both the manual and DSTA-Net model measurements of the areas of the IJV and CCA demonstrated reliable consistency and correlation. The detailed results are presented in Appendix Table 2 and Appendix Fig. 1. Table 2 The accuracy of the DSTA-Net, DeepLabV3+, Swin-Unet, TransUNet, and UNet models in segmenting the internal jugular vein and common carotid artery. IJV: Internal jugular vein, CCA: Common carotid artery, Dice: Dice coefficient, IoU: Intersection over union, R-Perimeter: Perimeter ratio, R-Area: Area ratio. IJV CCA Dice (%) IoU (%) R-Area (%) R-Perimeter (%) Dice (%) IoU (%) R-Area (%) R-Perimeter (%) DSTA-Net 82.62 ± 0.18 74.97 ± 0.15 11.81 ± 0.10 11.58 ± 0.10 80.09 ± 0.19 71.71 ± 0.17 12.56 ± 0.11 10.55 ± 0.09 DeepLabV3+ 77.58 ± 0.23 70.97 ± 0.14 20.24 ± 0.12 17.12 ± 0.10 73.85 ± 0.24 65.23 ± 0.18 18.41 ± 0.10 19.12 ± 0.08 Swin-Unet 74.92 ± 0.23 69.65 ± 0.13 21.93 ± 0.11 16.24 ± 0.09 72.62 ± 0.21 66.25 ± 0.14 19.39 ± 0.12 18.69 ± 0.10 TransUNet 75.45 ± 0.15 69.15 ± 0.13 20.71 ± 0.10 21.39 ± 0.12 72.56 ± 0.18 65.18 ± 0.16 21.68 ± 0.14 17.54 ± 0.09 UNet 80.22 ± 0.14 70.75 ± 0.19 16.55 ± 0.14 14.69 ± 0.12 70.52 ± 0.17 60.79 ± 0.12 18.85 ± 0.15 17.64 ± 0.15 Table 3 The B-A consistency analysis between manual measurements and DSTA-Net model measurements in the training and testing set. IJV Max Area: Internal jugular vein maximal area, IJV Min Area: Internal jugular vein minimal area, IJV Ratio: (IJV maximum are- IJV minimum area)/ IJV maximum are, CCA Area: Common carotid artery area at the end of expiration, IJV Max Area/CCA Area: Internal jugular vein maximal area/common carotid artery area at the end of expiration. Mean Difference: the average difference between the measurements of two methods, reflecting the magnitude of systematic error. LOA: Limits of Agreement is expressed as the range of the mean difference ± 1.96 times the standard deviation, used to evaluate the consistency range between the measurements of the two methods. Training set (218 patients) Test set (54 patients) Mean Difference LOA Mean Difference LOA IJV Max Area (mm 2 ) -17.340 [-41.842, 7.161] -16.759 [-40.736, 7.217] IJV Min Area (mm 2 ) -4.823 [-12.0622, 2.416] -4.239 [-11.447, 2.968] IJV Ratio -3.748 [-10.536, 3.040] -3.586 [-9.929, 2.757] CCA Area (mm 2 ) -5.27 [-11.531, 0.987] -6.46 [-12.587, -0.347] IJV Max /CCA -0.088 [-0.479, 0.302] -0.034 [-0.333, 0.265] Evaluation of DM-MLP model for predicting elevated CVP For the DM-MLP model ROC curve, the AUC based on manual measurement is 0.94 [0.88, 0.94], and the AUC based on DSTA-Net measurement is 0.88 [0.85, 0.91], which is significantly greater than those of VGG (manual AUC: 0.88 [0.82, 0.88], DSTA-Net AUC: 0.83 [0.78, 0.84]), RESNET (manual AUC: 0.90 [0.84, 0.90], DSTA-Net AUC: 0.85 [0.80, 0.86]), and Transformer (manual AUC: 0.89 [0.83, 0.90], DSTA-Net AUC: 0.84 [0.80, 0.86]). In terms of accuracy, the DM-MLP model also outperforms all other models, with an accuracy of 89.06 ± 0.62, while VGG achieved 82.69 ± 2.04, ResNet reached 84.42 ± 0.32, and Transformer obtained 83.66 ± 1.58. Additionally, the DM-MLP model shows greater sensitivity (94.51 ± 0.22), specificity (82.44 ± 0.40), PPV( 85.65 ± 0.73) and NPV(93.59 ± 0.78) than the other models. In addition, the AUC, accuracy, sensitivity, specificity, PPV, and NPV of all the clinical prediction models based on manual measurements are greater than those of the DSTA-Net measurement (Table 4 and Fig. 6 ). Table 4 The AUC, Accuracy, Sensitivity, Specificity, PPV and NPV of ROC curve for VGG, Resnet, Transformer and DM-MLP models(Manual/DSTA-Net clinical prediction models). Data in brackets are 95% CIs. Manual clinical prediction models are elevated CVP prediction models based on manual measurement of IJV and CCA, DSTA-Net clinical prediction models are CVP prediction models based on DSTA-Net measurement of IJV and CCA. AUC: Area under the receiver operating characteristic curve, NPV: Negative predictive value, PPV: Positive predictive value. P-value AUC Accuracy Sensitivity Specificity PPV NPV DM-MLP Manual 0.05 0.94 [0.88, 0.94] 89.06 ± 0.62 94.51 ± 0.22 82.44 ± 0.40 85.65 ± 0.73 93.59 ± 0.78 DSTA-Net 0.05 0.88 [0.85, 0.91] 83.06 ± 0.16 91.23 ± 0.45 75.07 ± 0.49 78.97 ± 0.16 88.81 ± 0.23 VGG Manual 0.05 0.88 [0.82, 0.88] 82.69 ± 2.04 88.48 ± 1.77 77.21 ± 1.81 80.28 ± 1.68 87.51 ± 2.08 DSTA-Net 0.05 0.83[0.78, 0.84] 79.04 ± 1.70 84.90 ± 1.75 69.60 ± 1.16 74.82 ± 1.74 82.88 ± 1.17 Resnet Manual 0.05 0.90 [0.84, 0.90] 84.42 ± 0.32 90.02 ± 0.73 78.29 ± 0.69 81.94 ± 0.95 89.44 ± 1.09 DSTA-Net 0.05 0.85 [0.80, 0.86] 79.54 ± 0.43 86.00 ± 0.35 71.45 ± 0.42 76.21 ± 0.19 85.16 ± 0.59 Transformer Manual 0.05 0.89 [0.83, 0.90] 83.66 ± 1.58 89.41 ± 1.57 78.46 ± 1.21 81.90 ± 0.91 87.84 ± 2.06 DSTA-Net 0.05 0.84 [0.80, 0.86] 79.58 ± 1.67 86.38 ± 1.44 70.47 ± 1.13 76.54 ± 0.90 83.81 ± 0.82 Discussion Our results indicate that the combined use of a wearable ultrasound device with the DSTA-Net model allows for more accurate dynamic measurements of the IJV and CCA. Furthermore, the developed DM-MLP model, which combines baseline patient characteristics with wearable ultrasound parameters accurately predictes CVP elevation. These findings represent a significant advancement in predicting CVP. Excessive fluid overload poses significant risks. In one European survey of patients with sepsis, high blood volume was associated with increased mortality[ 17 ]. Additionally, a large randomized study of patients with acute lung injury revealed that high blood volume contributed to increased mortality rates[ 18 ], highlighting the critical importance of hemodynamic management. CVP is an important physiological indicator for assessing blood volume. Monitoring CVP helps clinicians better understand patients’ fluid status, enabling appropriate treatment strategies. Although the CVC is the reference standard for CVP assessment, its invasiveness and complex procedural steps pose certain limitations in clinical applications. Therefore, exploring noninvasive methods for evaluating CVP is necessary. Hill et al. reported that analyzing the relationship between cardiac output and CVP could also be used to evaluate fluid volume status. Nevertheless, this approach requires highly accurate hemodynamic monitoring equipment and skilled operators[ 19 ]. Additionally, physical examinations such as hypotension, reduced urine output, and peripheral edema have some value in assessing CVP[ 20 , 21 ]. However, these indicators are highly subjective and have limited precision and reliability. Previous studies have reported ultrasound measurements of the IVC to assess CVP [ 22 ]. However, the quality of IVC ultrasound images is influenced by patient factors (e.g. obesity, increased bowel gas, and lung gas), making accurate assessment challenging or impossible. Consequently, relevant findings have shown significant discrepancies[ 23 , 24 ]. The IJV, a superficial organ, is less affected by these factors, making it easier to obtain clear images. The right IJV is directly connected to the superior vena cava and the right atrium through a valveless venous system [ 25 ]. Therefore, the filling degree of the IJV could also reflect the change in the CVP. Herein, we found that the IJV Max Area, IJV Min Area, and the IJV Max/CCA Area were significantly higher in CVP ≥ 8mmhg group than CVP<8mmhg group, which is consistent with the findings of Muhammet et al, who demonstrated a significant correlation between the CVP and IJV diameter, indicating that ultrasound noninvasively measures the IJV diameter as an alternative to CVP measurement[ 23 ]. Our study further supports the potential of ultrasound parameters of the IJV Ratio in assessing the CVP. This study innovatively proposed DSTA-Net as the image segmentation model and DM-MLP as the clinical prediction model for our wearable ultrasound device. Compared with the DeepLabV3+, Swin-Unet, TransUNet, and UNet models, the DSTA-Net model achieved more precise segmentation performance, especially in terms of the Dice for IJV segmentation. In terms of the R-area and R-perimeter, the DSTA-Net model also exhibited more balanced performance, better controlling the shape and area of the predicted region, and avoiding overfitting or unrealistic predictions. The excellent agreement and high correlation between the manual and DSTA-Net measurement parameters of the IJV and CCA among patients with varying CVP levels also demonstrated the effectiveness of the DSTA-Net model. Compared with manual measurement, wearable ultrasonic automatic measurement significantly reduces the operation time, improves the efficiency of diagnosis and treatment, and could achieve continuous monitoring in the future, which is challenging to manual measurement[ 26 , 27 ]. Furthermore, DM-MLP model integrated manually- and DSTA-NET-labeled IJV and CCA images as well as clinical data to predict CVP elevation with an AUCs of 0.94 and 0.88, respectively. Compared with the VGG, ResNet, and Transformer models, the DM-MLP model has a greater, accuracy, sensitivity, specificity, PPV, and NPV, indicating superior predictive performance. Although the effectiveness of the DSTA-Net DM-MLP model was lower than that of the Manual DM-MLP model, both were highly effective at predicting elevated CVP. However, the DM-MLP model showed high PPV and NPV in predicting elevated CVP, which may be attributed to the relatively small sample size in our study. In future research, we will increase the accuracy of the DSTA-Net DM-MLP model by increasing the sample size and optimizing the experimental process, to narrow the gap with manually labeled data, and to reduce the PPV and NPV. Additionally, expanding the dataset to include diverse patient groups and clinical conditions would help to improve the model's generalizability and clinical applicability. Limitations This study has several limitations. Firstly, although the DM-MLP model performed well, the sample size for clinical prediction was relatively small, necessitating a larger and more diverse patient population. Further, we only examined the effectiveness of predicting elevated CVP; a detailed assessment of the feasibility and comfort of long-term wear for patients is lacking, largely due to the challenges of working with unconscious ICU patients. Future studies with an expanded sample size should include evaluations of human applicability, comfort, and operability. Secondly, the DM-MLP clinical prediction model a non-transparent model that provides limited insight, making it difficult to understand the mechanics underling predictions. Future research should explore interpretable machine learning models to enhance patient trust and understanding. Thirdly, this study only measured the maximum and minimum cross-sectional area of the IJV during a respiratory cycle, without real-time monitoring of variations throughout the cycle. In the future, we will conduct further in-depth research focusing on these aspects. Conclusions This study integrates wearable ultrasound devices with AI models to achieve portable, noninvasive assessment of elevated CVP, with the potential to revolutionize fluid management and improve patient outcomes, making a significant contribution to the field of noninvasive medical monitoring. Further research should allow the integration of wearable ultrasound devices, while AI could extend from intensive care to other clinical settings, such as outpatient monitoring or home care, providing patients with more portable and continuous assessment tools. Abbreviations CVP Central venous pressure CVC Central venous catheter AI Artificial intelligence IJV Internal jugular vein CCA Common carotid artery ICUs Intensive care units DSTA-Net Dual-decoder spatiotemporal attention network DM-MLP Dual-mixing MLP ECG Electrocardiogram IJV Max Area The maximum IJV area at the end of expiration IJV Min Area The minimum IJV area at the end of inspiration IJV Max/CCA Area The maximum IJV area /CCA area IJV Ratio The change rate of the IJV BSA Body surface area BMI Body mass index SBP Systolic blood Pressure DBP Diastolic blood Pressure HR Heart rate Dice Dice coefficient IoU Intersection over Union B-A analysis Bland‒Altman analysis AUC Area under the curve ROC Receiver operating characteristic NPV Negative predictive value PPV Positive predictive value Declarations Acknowledgments We sincerely thank Zixuan Zheng, PhD (MindRank AI Co., Ltd., Wuxi, China.) for her contribution to image segmentation and annotation. Author contributions Guarantors of integrity of entire study, L.P.D., M.L., Y.L., X.X.Z., T.Y., J.Y.G., S.R.L., X.P.S., W.K.J., X.S.W., F.Y.W., L.P.Z., Y.Y.Z.; project management, funding support, and acquisition of research resources: Y.Y.Z., L.P.Z.; research design, experimental implementation, and data analysis, drafting of the manuscript: L.P.D.; participation in data collection, proposing key research hypotheses: M.L., T.Y., S.R.L., X.P.S., W.K.J., X.S.W., F.Y.W.; statistical analysis: L.P.D., J.Y.G.; review of the manuscript and provision of revision suggestions: Y.L., X.X.Z.,Y.Y.Z. Funding Supported by the Exploratory Clinical Project at the Institutional Level of Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University (No: yynts202302) and the Shanghai Municipal Health Commission Clinical Research Program (No: 20244Y0119). Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Ethics approval and consent to participate This study was approved the institutional review board of Shanghai Sixth People's Hospital (institutional review board number: 2023-007-(3); approval date: August 30, 2023), and performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments. All patients or their first-degree relatives provided written informed consent. Consent for publication Not applicable. Competing interests The authors have no competing interests to disclose. References Mullens W, Abrahams Z, Francis GS, Sokos G, Taylor DO, Starling RC, Young JB, Tang WHW: Importance of venous congestion for worsening of renal function in advanced decompensated heart failure. J Am Coll Cardiol 2009, 53(7):589–596. Lala A, McNulty SE, Mentz RJ, Dunlay SM, Vader JM, AbouEzzeddine OF, DeVore AD, Khazanie P, Redfield MM, Goldsmith SR et al : Relief and Recurrence of Congestion During and After Hospitalization for Acute Heart Failure: Insights From Diuretic Optimization Strategy Evaluation in Acute Decompensated Heart Failure (DOSE-AHF) and Cardiorenal Rescue Study in Acute Decompensated Heart Failure (CARESS-HF). Circ Heart Fail 2015, 8(4):741–748. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, Machado FR, McIntyre L, Ostermann M, Prescott HC et al : Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock 2021. Crit Care Med 2021, 49(11):e1063-e1143. Huang CY, Guiza F, Wouters P, Mebis L, Carra G, Gunst J, Meersseman P, Casaer M, Van den Berghe G, De Vlieger G et al : Development and validation of the creatinine clearance predictor machine learning models in critically ill adults. Crit Care 2023, 27(1):272. Cole R: Does central venous pressure predict fluid responsiveness? Chest 2008, 134(6):1351–1352. McGee DC, Gould MK: Preventing complications of central venous catheterization. N Engl J Med 2003, 348(12):1123–1133. Wang L, Harrison J, Dranow E, Aliyev N, Khor L: Accuracy of Ultrasound Jugular Venous Pressure Height in Predicting Central Venous Congestion. Ann Intern Med 2022, 175(3):344–351. Simon MA, Kliner DE, Girod JP, Moguillansky D, Villanueva FS, Pacella JJ: Detection of elevated right atrial pressure using a simple bedside ultrasound measure. Am Heart J 2010, 159(3):421–427. Simon MA, Schnatz RG, Romeo JD, Pacella JJ: Bedside Ultrasound Assessment of Jugular Venous Compliance as a Potential Point-of-Care Method to Predict Acute Decompensated Heart Failure 30-Day Readmission. J Am Heart Assoc 2018, 7(15):e008184. Hu H, Hu C, Guo W, Zhu B, Wang S: Wearable ultrasound devices: An emerging era for biomedicine and clinical translation. Ultrasonics 2024, 142:107401. Kenny JS, Munding CE, Eibl AM, Eibl JK: Wearable ultrasound and provocative hemodynamics: a view of the future. Crit Care 2022, 26(1):329. A wearable ultrasound patch for continuous heart imaging. Nature 2023. Hu H, Huang H, Li M, Gao X, Yin L, Qi R, Wu RS, Chen X, Ma Y, Shi K et al : A wearable cardiac ultrasound imager. Nature 2023, 613(7945):667–675. Hill B, Smith C: Central venous pressure monitoring in critical care settings. Br J Nurs 2021, 30(4):230–236. Hamzaoui O, Teboul JL: Central venous pressure (CVP). Intensive Care Med 2022, 48(10):1498–1500. Roger C, Muller L, Riou B, Molinari N, Louart B, Kerbrat H, Teboul JL, Lefrant JY: Comparison of different techniques of central venous pressure measurement in mechanically ventilated critically ill patients. Br J Anaesth 2017, 118(2):223–231. Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, Gerlach H, Moreno R, Carlet J, Le Gall JR, Payen D et al : Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 2006, 34(2):344–353. National Heart L, Blood Institute Acute Respiratory Distress Syndrome Clinical Trials N, Wiedemann HP, Wheeler AP, Bernard GR, Thompson BT, Hayden D, deBoisblanc B, Connors AF, Jr., Hite RD et al : Comparison of two fluid-management strategies in acute lung injury. N Engl J Med 2006, 354(24):2564–2575. Thomas Hill B: Role of central venous pressure monitoring in critical care settings. Nurs Stand 2018, 32(23):41–48. Davison R, Cannon R: Estimation of central venous pressure by examination of jugular veins. Am Heart J 1974, 87(3):279–282. McGee SR: Physical examination of venous pressure: a critical review. Am Heart J 1998, 136(1):10–18. Ruge M, Marhefka GD: IVC measurement for the noninvasive evaluation of central venous pressure. J Echocardiogr 2022, 20(3):133–143. Bayraktar M, Kacmaz M: Correlation of internal jugular vein, common carotid artery, femoral artery and femoral vein diameters with central venous pressure. Medicine (Baltimore) 2022, 101(43):e31207. Sekiguchi H, Seaburg LA, Suzuki J, Astorne WJ, Patel AS, Keller AS, Gajic O, Kashani KB: Central venous pressure and ultrasonographic measurement correlation and their associations with intradialytic adverse events in hospitalized patients: A prospective observational study. J Crit Care 2018, 44:168–174. Kosnik N, Kowalski T, Lorenz L, Valacer M, Sakthi-Velavan S: Anatomical review of internal jugular vein cannulation. Folia Morphol (Warsz) 2024, 83(1):1–19. Zhang M, Zhao L, Wang X, Lo WLA, Wen J, Li L, Huang Q: Automatic extraction and measurement of ultrasonic muscle morphological parameters based on multi-stage fusion and segmentation. Ultrasonics 2024, 137:107187. Zhou GQ, Chan P, Zheng YP: Automatic measurement of pennation angle and fascicle length of gastrocnemius muscles using real-time ultrasound imaging. Ultrasonics 2015, 57:72–83. Additional Declarations No competing interests reported. Supplementary Files Appendix1.pdf Appendix2.pdf Appendix3.pdf AppendixFigure1.pdf Graphicalabstractimage.pdf AppendixTable.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6316972","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":445232970,"identity":"d1501cf5-e7c3-42ef-bfcf-ad8002a82613","order_by":0,"name":"Liping Dong","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Liping","middleName":"","lastName":"Dong","suffix":""},{"id":445232971,"identity":"5773d968-ddd3-4c96-b16a-3e39471e8e00","order_by":1,"name":"Meng Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Meng","middleName":"","lastName":"Li","suffix":""},{"id":445232972,"identity":"d7c46931-78a7-4bbe-8483-7a3d2b60aae8","order_by":2,"name":"Yi Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Li","suffix":""},{"id":445232973,"identity":"8d154977-6822-4beb-959d-cabbcf62ea0e","order_by":3,"name":"Xingxuan Zhang","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xingxuan","middleName":"","lastName":"Zhang","suffix":""},{"id":445232974,"identity":"7612b10a-7d5e-4885-8f41-d3a1c1bace5e","order_by":4,"name":"Jingyi Guo","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jingyi","middleName":"","lastName":"Guo","suffix":""},{"id":445232975,"identity":"89835f0e-c973-433d-b7e3-8e1cb4056b2f","order_by":5,"name":"Shaorong Lu","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shaorong","middleName":"","lastName":"Lu","suffix":""},{"id":445232976,"identity":"e674269a-eb28-4142-b1b3-cfe144ab17bf","order_by":6,"name":"Xiaoping Shao","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaoping","middleName":"","lastName":"Shao","suffix":""},{"id":445232977,"identity":"37145667-0443-4676-9a60-7035e0810ebc","order_by":7,"name":"Feiyan wang","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Feiyan","middleName":"","lastName":"wang","suffix":""},{"id":445232978,"identity":"0d2f3962-9f5a-4fe7-815d-20aed2b919d8","order_by":8,"name":"weikang Jiang","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"weikang","middleName":"","lastName":"Jiang","suffix":""},{"id":445232979,"identity":"59e73df5-6ee4-4abc-9c19-e8772f6e0639","order_by":9,"name":"Xianshuai Wu","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xianshuai","middleName":"","lastName":"Wu","suffix":""},{"id":445232980,"identity":"ad0ec550-718e-48e2-baf3-bb7402039c9e","order_by":10,"name":"Liping Zhang","email":"","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":false,"prefix":"","firstName":"Liping","middleName":"","lastName":"Zhang","suffix":""},{"id":445232981,"identity":"8a634dd0-54ec-4a85-9817-459ad55037e2","order_by":11,"name":"Yuanyi Zheng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBElEQVRIiWNgGAWjYDACCTiLsfEBAw+YZUC0lmYDBh4DkrQwsElAVePXIj+7+djDr22H5c0lktsqf8j8SWxgb94mwVBzB6cWxjnH0o1l2w4b7pyR2Habh8cgsYHnWJkEw7FnOLUwS+SYSUu2HWbccAOohQGkBSgiwdhwGKcWNon8byAt9iAthT9AWuTf4NfCI5HDJvmx7XAiSAsD2GESPPi1SEikmUkznEtP3nDmYbM0D4+xcRtPWrFFwjHcWuRnJD+T/FFmbbvhePrDjz975GT72Q9vvPGhBrcWcBDwsgFJgQRg+PUAfQcSSsCrAajwxx8gyX8ASPwgoHQUjIJRMApGJAAAMNJUThR3bYkAAAAASUVORK5CYII=","orcid":"","institution":"Shanghai Jiao Tong University of Medicine Affliated Sixth People’ s Hospital","correspondingAuthor":true,"prefix":"","firstName":"Yuanyi","middleName":"","lastName":"Zheng","suffix":""}],"badges":[],"createdAt":"2025-03-27 05:08:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6316972/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6316972/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81738341,"identity":"d345a149-af8d-4df9-8ae9-eb7b3129eb25","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":165341,"visible":true,"origin":"","legend":"\u003cp\u003eFlow of acute and critically ill patients receiving CVC, CVC=central venous catheterization, (DSTA-Net) model: Dual-Decoder Spatiotemporal Attention Network model, DM-MLP model: Dual-Mixing MLP model.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/95701942585176538d5d6998.png"},{"id":81738647,"identity":"8028a873-bfa9-419b-a4cf-189c94cc33ff","added_by":"auto","created_at":"2025-04-30 23:54:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":479011,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea: \u003c/strong\u003eCVP assessment and wearable ultrasound device measurement in critically ill patients, CVP=central venous pressure. \u003cstrong\u003eb:\u003c/strong\u003e CVP wave, Respiratory wave and IJV and CCA 2D- and segmented- images for wearable ultrasound, a: Represents right atrial contraction, c: Represents the rise in right atrial pressure caused by the closure of the tricuspid valve, x: Represents the pressure decrease due to right atrial relaxation and the retraction of the tricuspid valve, v: Represents right atrial filling, y: Represents the blood flow from the right atrium to the right ventricle. \u003cstrong\u003ec:\u003c/strong\u003e The real scenario of wearable ultrasound for monitoring critically ill patients.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/8ffb930c98b094464b2017c3.png"},{"id":81738344,"identity":"61610237-179c-4b57-9f89-4951fa56134b","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":370565,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea: \u003c/strong\u003eWearable ultrasound device external structure, \u003cstrong\u003eb:\u003c/strong\u003ewearable ultrasound device internal structure.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/e586c55d5cbac988c78c3ab5.png"},{"id":81738491,"identity":"ac53df46-ff20-4fbe-be11-7b6b311ed50c","added_by":"auto","created_at":"2025-04-30 23:46:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":226731,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea: \u003c/strong\u003eDual-Decoder Spatiotemporal Attention Network model of IJV and CCA in critically ill patients, \u003cstrong\u003eb: \u003c/strong\u003eDual-Mixing MLP Model. CCA=common carotid artery, IJV=internal jugular vein.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/923918482af910407e128c09.png"},{"id":81738347,"identity":"d2f85026-72e2-40e1-b2e3-13913f579ea3","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":798828,"visible":true,"origin":"","legend":"\u003cp\u003eIn the training set (above) and test set (below), Bland-Altman plots showing IJV Max Area, IJV Min Area, CCA Area, IJV Max/CCA Area, and IJV Ratio for manual and DSTA-Net model measurements comparison. In the training (above) set and test set (below), Pearson correlation analysis was performed for manual and DSTA-Net model measurements of IJV Max Area, IJV Min Area, CCA Area, IJV Max/CCA Area, and IJV Ratio. In the upper-right cells, different shapes and colors represent the correlation: red indicates a negative correlation, blue indicates a positive correlation, and the color depth signifies the correlation strength—the darker the color, the stronger the correlation. The more elongated the ellipse, the stronger the correlation. The lower-left cells display the specific correlation coefficients. IJV Max Area= maximum internal jugular vein area. IJV Max Area= maximum internal jugular vein area, IJV Min Area= minimum internal jugular vein area, CCA Area= area of the CCA, IJV Max/CCA Area= maximum IJV area / area of the CCA, IJV Ratio= change rate of the IJV over one respiratory cycle.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/b27b76d1bc08d45aeccd0eaa.png"},{"id":81738492,"identity":"66d40e97-33e8-40f7-abc1-afa101f2099b","added_by":"auto","created_at":"2025-04-30 23:46:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":312742,"visible":true,"origin":"","legend":"\u003cp\u003eROC curve of Manual and DSTA-Net DM-MLP, VGC, Resnet, Transformer models predicting elevated CVP.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/b214660d3d5e2359120e9f1a.png"},{"id":83019896,"identity":"3cc7f117-3de7-43eb-bfb9-db637ffcff9c","added_by":"auto","created_at":"2025-05-19 07:09:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3641405,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/4c96effc-9fe7-4060-9354-9a2f0fdb73fb.pdf"},{"id":81738335,"identity":"84d78e52-6081-46cb-846d-51baaf9262c2","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":76061,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/ec9763bd55424e7c0c0af90e.pdf"},{"id":81738489,"identity":"8fc5be54-4a59-4842-b248-b72a75518db9","added_by":"auto","created_at":"2025-04-30 23:46:49","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":68613,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/5631a613542169e96ebfcdc6.pdf"},{"id":81738336,"identity":"4d406de9-3173-4fab-8463-b55be18041f3","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":91341,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/2b30f32758486cd085d2478b.pdf"},{"id":81738343,"identity":"ac19b479-7ea3-4160-9797-0fb348b17880","added_by":"auto","created_at":"2025-04-30 23:38:49","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":218315,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixFigure1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/53d63be34a773fb7eba666a0.pdf"},{"id":81738502,"identity":"32303c5c-3fd4-4112-84d2-e92c93146717","added_by":"auto","created_at":"2025-04-30 23:46:50","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":397594,"visible":true,"origin":"","legend":"","description":"","filename":"Graphicalabstractimage.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/13f517175bf01020f4e73fbc.pdf"},{"id":81738649,"identity":"a0f5a3f3-fe7a-4e4d-879e-64831992de12","added_by":"auto","created_at":"2025-04-30 23:54:49","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":15049,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixTable.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6316972/v1/7dc7b3057e942c40225dc472.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Assessment of wearable Ultrasound Device Combined with AI for Portable Assessment of Central Venous Pressure Compared with Central Venous Catheterization as the reference Standard in critically ill patients: A Cross-Sectional Study","fulltext":[{"header":"Background","content":"\u003cp\u003eElevated central venous pressure (CVP) and insufficient CVP reduction are associated with adverse events and mortality in critically ill patients[\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e], indicating the need to identify elevated CVP. Central venous catheter (CVC), a complex, time-consuming, and complicated invasive procedure is the reference standard for CVP assessment[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Complications such as infection, thrombus and bleeding, pneumothorax, and hemothorax may occur during CVC placement, limiting its clinical application[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Given these challenges, it is imperative to explore noninvasive methods for assessing CVP.\u003c/p\u003e \u003cp\u003eRecently, the internal jugular vein (IJV), due to its direct connection to the right atrium through venous access, has been shown to directly reflect right atrial pressure and could be used in CVP evaluation[\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Ultrasound is a noninvasive, portable technique commonly used to assess neck blood vessels. Traditional ultrasound probes are bulky and require constant hand-holding or a fixed machine for continuous monitoring. However, the prolonged pressure of bulky probes on critically ill patients' necks could cause discomfort and even suffocation[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Several studies have reported that wearable ultrasound devices could be portable, for real-time monitoring of internal organs and related conditions, including blood vessels, muscles, heart, gastrointestinal tract, diaphragm, and lung[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. It could also facilitate continuous real-time imaging of the left ventricle during physical activity[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Compared with traditional ultrasound, wearable ultrasound is more compact and portable, making it suitable for organ monitoring in pre-hospital emergency settings and intensive care units (ICUs)[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Nevertheless, manual measurement is still required, which increases the operation time and cost. Combining wearable ultrasound devices with artificial intelligence (AI) enables real-time measurement, significantly reducing diagnostic and treatment time, which is particularly crucial for critically ill patients. Reports on wearable ultrasound devices integrated with AI to evaluate the IJV for predicting CVP are lacking.\u003c/p\u003e \u003cp\u003eHerein, we used a wearable ultrasound device combined with AI to achieve real-time dynamic measurement of the IJV and CCA, accurately predicting elevated CVP in critically ill patients, providing a novel approach for noninvasive CVP assessment, which could significantly improve hemodynamic monitoring, provide timely and reliable healthcare decision support, and reduce mortality.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy population\u003c/h2\u003e \u003cp\u003eThis prospective, single-center, stratified, cross-sectional clinical trial was performed in accordance with the Declaration of Helsinki, was approved by the institutional review board of Shanghai Sixth People's Hospital (institutional review board number: 2023-007-(3); approval date: August 30, 2023), and registered in the Chinese Clinical Trial Registry (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.chictr.org.cn/\u003c/span\u003e\u003cspan address=\"https://www.chictr.org.cn/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e; registration number: ChiCTR230007586). None of the study participants had been reported previously, and received no financial compensation. All patients or their first-degree relatives provided written informed consent.\u003c/p\u003e \u003cp\u003eWe prospectively enrolled 318 critically ill patients who received CVC, between December 1, 2023, and January 1, 2024, at the ICU Shanghai Sixth People's Hospital, China. The inclusion criteria were as follows: aged over 18 years, underwent CVP monitoring, and consecutive enrollment. The exclusion criteria, which were designed to eliminate confounding factors potentially interfering with accurate CVP measurement, included: (a) receiving positive end-expiratory pressure ventilation, (b) a history of neck radiation therapy or surgery, (c) congenital heart disease, (d) moderate to severe tricuspid regurgitation, (e) right heart dilation, (f) inability to lie flat due to unstable vital signs, (g) inability to visualize cervical vessels, and (h) poor image quality. The flowchart of patient enrolment is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eReference standard of central venous pressure\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe CVP of consenting participants was assessed by a professional emergency physician and a nursing staff, each with experience of \u0026gt;\u0026thinsp;100 cases of Central venous catheterization (CVC), which is a reference standard for monitoring CVP in critically ill patients. It was typically inserted into the right atrium via either the IJV or the femoral vein, and it was connected to a pressure transducer and an integrated bedside monitor. The pressure transducer was strategically positioned along the mid-axillary line at the level of the fourth intercostal space, which corresponds to the midpoint of the right atrium[\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. To accurately reflect cardiac preload, CVP measurements should be taken at the base of the \u0026ldquo;c\u0026rdquo; wave in the pressure waveform, which corresponds to the Q wave on the electrocardiogram (ECG) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Additionally, it was essential to time these measurements correctly: CVP should be assessed at the end of expiration when the patient was breathing calmly or at the beginning of expiration during spontaneous breathing. This practice helped to mitigate the influence of intrathoracic pressure on CVP readings [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Following septic shock treatment guidelines, we classified critically ill patients into two CVP groups: CVP\u0026thinsp;\u0026lt;\u0026thinsp;8 mmHg group and CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg group, patients with a CVP\u0026lt; 8 mmHg might indicate potential hypovolemia or inadequate preload. In contrast, patients with a CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg might suggest fluid overload or elevated right atrial pressures[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eWearable ultrasound acquisition and measurement\u003c/h3\u003e\n\u003cp\u003eThe wearable ultrasound device, model Cloud-35LL (Stork, Chengdu, China), consists of a patch transducer and a main unit (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The detailed description of the wearable ultrasound architecture is provided in Appendix 1.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eBefore conducting wearable ultrasound measurements, the patient for CVC should lie flat in a 0\u0026deg; position, ensuring that there was no pillow or support under the head. The patient\u0026rsquo;s neck was in a neutral position, gently rotated to the left (not exceeding 30\u0026deg;), to expose the right IJV and CCA. The solid coupling agent was placed between the patch transducer and the patient\u0026rsquo;s skin. Due to variations in neck length, the probe should be lightly placed at the junction of the middle and lower thirds of the neck. The probe was adjusted slightly to obtain the clearest circular cross-sectional ultrasound images of the IJV and CCA, while the tape secured the probe and solid coupling gel to the patient\u0026rsquo;s neck. No pressure should be applied above the patch transducer, and the tape should not be applied too tightly to avoid compressing the IJV and CCA, which may cause deformation. During the patient\u0026rsquo;s calm breathing, the sonographers operated the wearable ultrasound to record real-time ultrasound videos of the IJV and CCA cross-sections over one breathing cycle, saving the video in mp4 format. Simultaneously, waves and data from the CVP on the monitor were recorded (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAll manual measurements were performed by two professional sonographers (with 3 and 4 years of experience), and reviewed by a chief sonographer with over 10 years of experience. The sonographers manually traced the intima-media contours of the IJV and CCA. The maximum IJV area (IJV Max Area) at the end of expiration, the minimum IJV area (IJV Min Area) at the end of inspiration, CCA area at the end of expiration, the maximum IJV area /CCA area (IJV Max/CCA Area), and change rate of the IJV (IJV Ratio) were all manually measured and images were saved. All measurements were made without knowledge of the clinical data or previous measurement results.\u003c/p\u003e\n\u003ch3\u003eImage segmentation model based on wearable ultrasound\u003c/h3\u003e\n\u003cp\u003eIn this study, we creatively propose a Dual-Decoder Spatiotemporal Attention Network (DSTA-Net) image segmentation model, with a 4:1 split ratio for training and testing sets. Additionally, we compared DSTA-Net with existing models, including DeepLabV3+, Swin-Unet, TransUNet, and UNet. Ultrasound videos of 2/3 of the subjects were randomly selected to generate a continuous sequence of image frames. Key frames were extracted every 20 frames to ensure that the image data was representative. These extracted images were then manually annotated for subsequent analysis and processing. The annotation was performed via Pair software (version 3.0, Medical Ultrasound Image Computing Lab, MUSIC, Shenzhen, China), where manual marking was performed along the IJV and CCA intima on the key frames. As a result, the DSTA-Net model could automatically measure the IJV Max Area, and IJV Min Area, CCA Area, IJV Max/CCA Area, IJV Ratio over one respiratory cycle. The architecture was consisted of a shared encoder (E) and two decoders (D1 and D2), each serving distinct roles. This dual-decoder design effectively utilized the limited labeled data and optimized the use of temporal information, making it particularly well-suited for scenarios with sparse annotations. All the ultrasound images and frames were standardized to a consistent resolution of 256\u0026times;256 pixels. To further enhance the model's robustness and generalizability, data augmentation techniques were applied to the small ultrasound datasets. These techniques included random rotations within \u0026plusmn;\u0026thinsp;15 degrees, horizontal and vertical flips, and the addition of Gaussian noise (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). A detailed introduction to the DSTA-Net model can be found in Appendix 2.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eClinical prediction model based on wearable ultrasound\u003c/h3\u003e\n\u003cp\u003ePatient characteristics (age, height, weight, body surface area [BSA], body mass index [BMI], systolic blood pressure[SBP], diastolic blood pressure [DBP], heart rate [HR], breath, and wearable ultrasound parameters of IJV Max Area [Manual / DSTA-Net], IJV Min Area [Manual / DSTA-Net], IJV Ratio [Manual / DSTA-Net], CCA Area [Manual / DSTA-Net], and IJV Max/CCA Area [Manual / DSTA-Net]) were used as indicators to creatively train a DM-MLP clinical prediction model to predict elevated CVP (CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg). The core of this model includes two innovative MLP modules: the attribute-mixing MLP and the case-mixing MLP (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). A detailed introduction to the DM-MLP model can be found in Appendix 3. Additionally, we trained VGG, ResNet, and Transformer models using the same clinical indicators for comparison with the DM-MLP model. The same training and test sets used for the image segmentation model were used to develop the clinical prediction model. On the basis of whether the wearable ultrasound parameters were measured manually or automatically by DSTA-Net, the models were classified into two types: Manual and DSTA-Net Clinical Prediction Models.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eDescriptive analyses and software were conducted using R (version 3.5.0), Python 3.9.1 (Python Software Foundation, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.python.org\u003c/span\u003e\u003cspan address=\"http://www.python.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), PyTorch version 1.10.0 (PyTorch.org), and CUDA 11.3 with cuDNN 8.2 (NVIDIA, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://developer.nvidia.com/cudnn\u003c/span\u003e\u003cspan address=\"https://developer.nvidia.com/cudnn\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for all computational tasks. Analyses were performed on hardware equipped with 8 NVIDIA RTX 4090 GPUs, enabling efficient parallel computation and deep learning model training. The Shapiro-Wilk and Kolmogorov-Smirnov tests were used to test for normality. Normally and non-normally distributed data are presented as the means\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviations and medians and interquartile ranges (IQRs), respectively. The Mann-Whitney U test was used to evaluate the statistical significance of differences for continuous data. The accuracies of the DSTA-Net, DeepLabV3+, Swin-Unet, TransUNet, and UNet image segement models are represented by the Dice coefficient (Dice), intersection over Union (IoU), perimeter ratio (R-perimeter), and area ratio (R-area). The consistency and correlation between the DSTA-Net model and manual measurements were assessed using Bland-Altman (B-A) and Spearman correlation analyses. Using the CVC as the reference standard, the effectiveness of the DM-MLP, VGG, ResNet, and Transformer models for assessing elevated CVP was evaluated via the area under the curve (AUC) of the Receiver Operating Characteristic (ROC) curve, accuracy, sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). The significance level was set at 5%.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eBaseline patient characteristics\u003c/h2\u003e \u003cp\u003eThis study initially included 318 critically ill patients who underwent CVC; however, some patients were excluded due to positive pressure ventilation (n\u0026thinsp;=\u0026thinsp;10), moderate to severe tricuspid regurgitation (n\u0026thinsp;=\u0026thinsp;9), right heart dilation (n\u0026thinsp;=\u0026thinsp;8), inability to measure the cervical vessels due to bandaging (n\u0026thinsp;=\u0026thinsp;6), and inability to lie flat (n\u0026thinsp;=\u0026thinsp;13). Overall, 272 critically ill patients were included for DSTA-Net training, of whom 218 (80%) and 54 (20%) were allocated to the training and test sets, respectively. The training set was divided into the CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg (n\u0026thinsp;=\u0026thinsp;87; mean age 58.86 years\u0026thinsp;\u0026plusmn;\u0026thinsp;18.14 [SD]; 38 women) and CVP\u0026lt;8 mmHg (n\u0026thinsp;=\u0026thinsp;131; mean age 52.21years\u0026thinsp;\u0026plusmn;\u0026thinsp;21.27 [SD]; 77 women) groups, as was the test set (CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg group: n\u0026thinsp;=\u0026thinsp;22; mean age 58.05 years\u0026thinsp;\u0026plusmn;\u0026thinsp;16.03 [SD]; 12 women; CVP\u0026lt;8 mmHg group: n\u0026thinsp;=\u0026thinsp;32; mean age 52.47\u0026thinsp;\u0026plusmn;\u0026thinsp;18.10[SD]; 17 women). In the training set, the systolic blood pressure (SBP) and diastolic blood pressure (DBP) were significantly greater in the CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg group than that in the CVP\u0026lt;8 mmHg group. Additionally, in both the training and test sets, the CVP, manual and DSTA-Net IJV Max Area, IJV Min Area, and IJV Max/CCA Area were significantly greater in the CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg compared to the CVP\u0026lt;8 mmHg group. Furthermore, the IJV Ratio, which was based on both manual and DSTA-Net measurements, was significantly elevated in the CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg group of the training set. The baseline patient characteristics and wearable ultrasound parameters for the training and test sets of critically ill patients are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Appendix Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline patient characteristics and wearable ultrasound parameters for the training and test sets of critically ill patient. BSA: Body surface area, BMI: Body mass index, SBP: Systolic blood pressure, DBP: Diastolic blood pressure, HR: Heart rate, IJV Max Area: Internal jugular vein maximal area, IJV Min Area: Internal jugular vein minimal area, IJV Ratio: (IJV maximum are- IJV minimum area)/ IJV maximum are, CCA Area: Common carotid artery area at the end of expiration, IJV Max Area/CCA Area: Internal jugular vein maximal area/common carotid artery area at the end of expiration.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining set(218 patients)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTest set(54 patients)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge (y)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e54.88\u0026thinsp;\u0026plusmn;\u0026thinsp;20.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e54.74\u0026thinsp;\u0026plusmn;\u0026thinsp;17.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.964\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHeight(cm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e169(160\u0026ndash;173)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e168(160\u0026ndash;170)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.319\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWeight(kg)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65(60\u0026ndash;75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65(55\u0026ndash;75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.727\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBSA(m\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.73\u0026thinsp;\u0026plusmn;\u0026thinsp;0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.70\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.351\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBMI (kg/m\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.40(21.22\u0026ndash;26.12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23.91(20.76\u0026ndash;25.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSBP(mmHg)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e123(110\u0026ndash;141)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e119(108\u0026ndash;140)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.265\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDBP(mmHg)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e70(63\u0026ndash;80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e68(59\u0026ndash;78)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.220\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHR(bqm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e89(79\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93(81\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.185\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBreath(bpm/min)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20(18\u0026ndash;22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20(17\u0026ndash;22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.782\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual-IJV Max Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e131.35(85.20-186.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e120.8(75.18-210.21)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.507\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual-IJV Min Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e77.73(46.70-124.58)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e56.52(36.88-125.43)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.144\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual-IJV Ratio\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30.27(16.51\u0026ndash;54.04)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e34.45(16.54\u0026ndash;65.13)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.320\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual-CCA Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e57.10(44.57-73.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61.15(47.33\u0026ndash;76.45)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.323\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual-IJV Max Area/CCA Are\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.29(1.48\u0026ndash;3.36)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.04(1.20\u0026ndash;3.19)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.251\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net IJV Max Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e146.53(95.74\u0026ndash;212.20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e137.3(83.39-235.38)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.535\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net IJV Min Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81.77(48.93-131.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e59.71(38.74-132.38)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.149\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net IJV Ratio\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e34.74(21.15\u0026ndash;57.17)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40.65(21.50-67.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.327\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net CCA Area(mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e62.40(49.26\u0026ndash;79.67)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69.07(51.70-85.18)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.209\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net IJV Max Area/CCA Area\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.32(1.54\u0026ndash;3.58)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.16(1.26\u0026ndash;3.30)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.249\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCVP, mmHg\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6(4\u0026ndash;9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6(5\u0026ndash;8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.841\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eManual DM-MLP CVP, mmHg\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.19(4.03\u0026ndash;9.08)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.63(4.49\u0026ndash;8.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net DM-MLP CVP, mmHg\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.22(3.85\u0026ndash;9.26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.72(4.21\u0026ndash;8.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.648\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAccuracy of the wearable ultrasound image segmentation model\u003c/h3\u003e\n\u003cp\u003eThe DSTA-Net model achieved Dice and IoU values of 82.62\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18 and 74.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15 for IJV segmentation, and 80.09\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19 and 71.71\u0026thinsp;\u0026plusmn;\u0026thinsp;0.17 for CCA segmentation, outperforming DeepLabV3+ (IJV Dice: 77.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.23, CCA Dice: 73.85\u0026thinsp;\u0026plusmn;\u0026thinsp;0.24, IJV IoU: 70.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14, CCA IoU: 65.23\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18), Swin-Unet (IJV Dice: 74.92\u0026thinsp;\u0026plusmn;\u0026thinsp;0.23, CCA Dice: 72.62\u0026thinsp;\u0026plusmn;\u0026thinsp;0.21, IJV IoU: 69.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13, CCA IoU: 66.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14), TransUNet (IJV Dice: 75.45\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15, CCA Dice: 72.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18, IJV IoU: 69.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13, CCA IoU: 65.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.16), and Unet (IJV Dice: 80.22\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14, CCA Dice: 70.52\u0026thinsp;\u0026plusmn;\u0026thinsp;0.17, IJV IoU: 70.75\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19, CCA IoU: 60.79\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12). Furthermore, in terms of metrics such as the area ratio (R-area IJV:11.81\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10, CCA:12.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11) and perimeter ratio (R-perimeter IJV:11.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10, CCA:10.55\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09), the DSTA-Net model also demonstrated better segmentation performance, with smaller R-area and R-perimeter (Table. 2).\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eConsistency of the wearable ultrasound image segmentation model\u003c/h2\u003e \u003cp\u003eFor the training and testing sets, the mean differences in the IJV Min Area between the manual and DSTA-Net measurements were \u0026minus;\u0026thinsp;4.82 and \u0026minus;\u0026thinsp;4.24, with respective LOA ranges of [-12.06, 2.42] and [-11.45, 2.97], respectively. For the IJV Ratio, the mean differences were \u0026minus;\u0026thinsp;3.75 and \u0026minus;\u0026thinsp;3.59, with LOA ranges of [-10.54, 3.04] and [-9.93, 2.76], respectively. The mean differences for IJV Max/CCA were \u0026minus;\u0026thinsp;0.088 and \u0026minus;\u0026thinsp;0.034, with LOA ranges of [-0.48, 0.30] and [-0.33, 0.26], respectively. The small differences and narrow LOA ranges indicate good agreement between the two methods in the measurements of the minimum area, minimum ratio, and maximum/CCA values. However, for the IJV Max Area, the mean differences were \u0026minus;\u0026thinsp;17.34 and \u0026minus;\u0026thinsp;16.76, with LOA ranges of [-41.84, 7.16] and [-40.73, 7.21] in the training and testing sets, respectively. For the CCA Area, the mean differences were \u0026minus;\u0026thinsp;5.27 and \u0026minus;\u0026thinsp;6.47, with LOA ranges of [-11.53, 0.99] and [-12.59, -0.35], respectively. The relatively larger mean differences and negatively skewed LOA ranges indicate that the DSTA-Net measurements were slightly lower than the manual measurements. The correlation analysis indicated a high degree of correlation between the manual and DSTA-Net measurements of the IJV Max Area (r\u0026thinsp;=\u0026thinsp;0.99, p\u0026lt;0.01), IJV Min Area (r\u0026thinsp;=\u0026thinsp;0.99, p\u0026lt;0.01), CCA Area (r\u0026thinsp;=\u0026thinsp;0.99, p\u0026lt;0.01), IJV Max/CCA Area (r\u0026thinsp;=\u0026thinsp;0.99, p\u0026lt;0.01), and IJV Ratio (r\u0026thinsp;=\u0026thinsp;0.99, p\u0026lt;0.01) in critically ill patients(Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Similarly, when participants were divided into the CVP\u0026thinsp;\u0026lt;\u0026thinsp;8 mmHg and CVP\u0026thinsp;\u0026ge;\u0026thinsp;8 mmHg groups, both the manual and DSTA-Net model measurements of the areas of the IJV and CCA demonstrated reliable consistency and correlation. The detailed results are presented in Appendix Table\u0026nbsp;2 and Appendix Fig.\u0026nbsp;1.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe accuracy of the DSTA-Net, DeepLabV3+, Swin-Unet, TransUNet, and UNet models in segmenting the internal jugular vein and common carotid artery. IJV: Internal jugular vein, CCA: Common carotid artery, Dice: Dice coefficient, IoU: Intersection over union, R-Perimeter: Perimeter ratio, R-Area: Area ratio.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eIJV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003eCCA\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDice (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIoU (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eR-Area (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eR-Perimeter (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDice (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eIoU (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eR-Area (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eR-Perimeter (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e82.62\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e74.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e11.81\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e11.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e80.09\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e71.71\u0026thinsp;\u0026plusmn;\u0026thinsp;0.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e12.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e10.55\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDeepLabV3+\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e77.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e70.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e20.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e17.12\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e73.85\u0026thinsp;\u0026plusmn;\u0026thinsp;0.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e65.23\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e18.41\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e19.12\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSwin-Unet\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e74.92\u0026thinsp;\u0026plusmn;\u0026thinsp;0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e69.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e21.93\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e16.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e72.62\u0026thinsp;\u0026plusmn;\u0026thinsp;0.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e66.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e19.39\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e18.69\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTransUNet\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e75.45\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e69.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e20.71\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e21.39\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e72.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e65.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e21.68\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e17.54\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eUNet\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e80.22\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e70.75\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e16.55\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e14.69\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e70.52\u0026thinsp;\u0026plusmn;\u0026thinsp;0.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e60.79\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e18.85\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e17.64\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe B-A consistency analysis between manual measurements and DSTA-Net model measurements in the training and testing set. IJV Max Area: Internal jugular vein maximal area, IJV Min Area: Internal jugular vein minimal area, IJV Ratio: (IJV maximum are- IJV minimum area)/ IJV maximum are, CCA Area: Common carotid artery area at the end of expiration, IJV Max Area/CCA Area: Internal jugular vein maximal area/common carotid artery area at the end of expiration. Mean Difference: the average difference between the measurements of two methods, reflecting the magnitude of systematic error. LOA: Limits of Agreement is expressed as the range of the mean difference\u0026thinsp;\u0026plusmn;\u0026thinsp;1.96 times the standard deviation, used to evaluate the consistency range between the measurements of the two methods.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eTraining set (218 patients)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eTest set (54 patients)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean Difference\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLOA\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMean Difference\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eLOA\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIJV Max Area (mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-17.340\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e[-41.842, 7.161]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-16.759\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c5\"\u003e \u003cp\u003e[-40.736, 7.217]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIJV Min Area (mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-4.823\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e[-12.0622, 2.416]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-4.239\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c5\"\u003e \u003cp\u003e[-11.447, 2.968]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIJV Ratio\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-3.748\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e[-10.536, 3.040]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-3.586\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c5\"\u003e \u003cp\u003e[-9.929, 2.757]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCCA Area (mm\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u003cb\u003e)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-5.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e[-11.531, 0.987]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-6.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c5\"\u003e \u003cp\u003e[-12.587, -0.347]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIJV Max /CCA\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.088\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e[-0.479, 0.302]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.034\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c5\"\u003e \u003cp\u003e[-0.333, 0.265]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eEvaluation of DM-MLP model for predicting elevated CVP\u003c/h2\u003e \u003cp\u003eFor the DM-MLP model ROC curve, the AUC based on manual measurement is 0.94 [0.88, 0.94], and the AUC based on DSTA-Net measurement is 0.88 [0.85, 0.91], which is significantly greater than those of VGG (manual AUC: 0.88 [0.82, 0.88], DSTA-Net AUC: 0.83 [0.78, 0.84]), RESNET (manual AUC: 0.90 [0.84, 0.90], DSTA-Net AUC: 0.85 [0.80, 0.86]), and Transformer (manual AUC: 0.89 [0.83, 0.90], DSTA-Net AUC: 0.84 [0.80, 0.86]). In terms of accuracy, the DM-MLP model also outperforms all other models, with an accuracy of 89.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.62, while VGG achieved 82.69\u0026thinsp;\u0026plusmn;\u0026thinsp;2.04, ResNet reached 84.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.32, and Transformer obtained 83.66\u0026thinsp;\u0026plusmn;\u0026thinsp;1.58. Additionally, the DM-MLP model shows greater sensitivity (94.51\u0026thinsp;\u0026plusmn;\u0026thinsp;0.22), specificity (82.44\u0026thinsp;\u0026plusmn;\u0026thinsp;0.40), PPV( 85.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.73) and NPV(93.59\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78) than the other models. In addition, the AUC, accuracy, sensitivity, specificity, PPV, and NPV of all the clinical prediction models based on manual measurements are greater than those of the DSTA-Net measurement (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe AUC, Accuracy, Sensitivity, Specificity, PPV and NPV of ROC curve for VGG, Resnet, Transformer and DM-MLP models(Manual/DSTA-Net clinical prediction models). Data in brackets are 95% CIs. Manual clinical prediction models are elevated CVP prediction models based on manual measurement of IJV and CCA, DSTA-Net clinical prediction models are CVP prediction models based on DSTA-Net measurement of IJV and CCA. AUC: Area under the receiver operating characteristic curve, NPV: Negative predictive value, PPV: Positive predictive value.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eP-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eDM-MLP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eManual\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.94 [0.88, 0.94]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e89.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e94.51\u0026thinsp;\u0026plusmn;\u0026thinsp;0.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e82.44\u0026thinsp;\u0026plusmn;\u0026thinsp;0.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e85.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e93.59\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.88 [0.85, 0.91]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e83.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e91.23\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e75.07\u0026thinsp;\u0026plusmn;\u0026thinsp;0.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e78.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e88.81\u0026thinsp;\u0026plusmn;\u0026thinsp;0.23\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eVGG\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eManual\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.88 [0.82, 0.88]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e82.69\u0026thinsp;\u0026plusmn;\u0026thinsp;2.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e88.48\u0026thinsp;\u0026plusmn;\u0026thinsp;1.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e77.21\u0026thinsp;\u0026plusmn;\u0026thinsp;1.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e80.28\u0026thinsp;\u0026plusmn;\u0026thinsp;1.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e87.51\u0026thinsp;\u0026plusmn;\u0026thinsp;2.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.83[0.78, 0.84]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e79.04\u0026thinsp;\u0026plusmn;\u0026thinsp;1.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e84.90\u0026thinsp;\u0026plusmn;\u0026thinsp;1.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e69.60\u0026thinsp;\u0026plusmn;\u0026thinsp;1.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e74.82\u0026thinsp;\u0026plusmn;\u0026thinsp;1.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e82.88\u0026thinsp;\u0026plusmn;\u0026thinsp;1.17\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eResnet\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eManual\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.90 [0.84, 0.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e84.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e90.02\u0026thinsp;\u0026plusmn;\u0026thinsp;0.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e78.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e81.94\u0026thinsp;\u0026plusmn;\u0026thinsp;0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e89.44\u0026thinsp;\u0026plusmn;\u0026thinsp;1.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.85 [0.80, 0.86]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e79.54\u0026thinsp;\u0026plusmn;\u0026thinsp;0.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e86.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e71.45\u0026thinsp;\u0026plusmn;\u0026thinsp;0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e76.21\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e85.16\u0026thinsp;\u0026plusmn;\u0026thinsp;0.59\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eTransformer\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eManual\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.89 [0.83, 0.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e83.66\u0026thinsp;\u0026plusmn;\u0026thinsp;1.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e89.41\u0026thinsp;\u0026plusmn;\u0026thinsp;1.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e78.46\u0026thinsp;\u0026plusmn;\u0026thinsp;1.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e81.90\u0026thinsp;\u0026plusmn;\u0026thinsp;0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e87.84\u0026thinsp;\u0026plusmn;\u0026thinsp;2.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eDSTA-Net\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.84 [0.80, 0.86]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c5\"\u003e \u003cp\u003e79.58\u0026thinsp;\u0026plusmn;\u0026thinsp;1.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e86.38\u0026thinsp;\u0026plusmn;\u0026thinsp;1.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e70.47\u0026thinsp;\u0026plusmn;\u0026thinsp;1.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e76.54\u0026thinsp;\u0026plusmn;\u0026thinsp;0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c9\"\u003e \u003cp\u003e83.81\u0026thinsp;\u0026plusmn;\u0026thinsp;0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur results indicate that the combined use of a wearable ultrasound device with the DSTA-Net model allows for more accurate dynamic measurements of the IJV and CCA. Furthermore, the developed DM-MLP model, which combines baseline patient characteristics with wearable ultrasound parameters accurately predictes CVP elevation. These findings represent a significant advancement in predicting CVP.\u003c/p\u003e \u003cp\u003eExcessive fluid overload poses significant risks. In one European survey of patients with sepsis, high blood volume was associated with increased mortality[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Additionally, a large randomized study of patients with acute lung injury revealed that high blood volume contributed to increased mortality rates[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], highlighting the critical importance of hemodynamic management. CVP is an important physiological indicator for assessing blood volume. Monitoring CVP helps clinicians better understand patients\u0026rsquo; fluid status, enabling appropriate treatment strategies. Although the CVC is the reference standard for CVP assessment, its invasiveness and complex procedural steps pose certain limitations in clinical applications. Therefore, exploring noninvasive methods for evaluating CVP is necessary. Hill et al. reported that analyzing the relationship between cardiac output and CVP could also be used to evaluate fluid volume status. Nevertheless, this approach requires highly accurate hemodynamic monitoring equipment and skilled operators[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Additionally, physical examinations such as hypotension, reduced urine output, and peripheral edema have some value in assessing CVP[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. However, these indicators are highly subjective and have limited precision and reliability.\u003c/p\u003e \u003cp\u003ePrevious studies have reported ultrasound measurements of the IVC to assess CVP [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. However, the quality of IVC ultrasound images is influenced by patient factors (e.g. obesity, increased bowel gas, and lung gas), making accurate assessment challenging or impossible. Consequently, relevant findings have shown significant discrepancies[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The IJV, a superficial organ, is less affected by these factors, making it easier to obtain clear images. The right IJV is directly connected to the superior vena cava and the right atrium through a valveless venous system [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Therefore, the filling degree of the IJV could also reflect the change in the CVP. Herein, we found that the IJV Max Area, IJV Min Area, and the IJV Max/CCA Area were significantly higher in CVP\u0026thinsp;\u0026ge;\u0026thinsp;8mmhg group than CVP\u0026lt;8mmhg group, which is consistent with the findings of Muhammet et al, who demonstrated a significant correlation between the CVP and IJV diameter, indicating that ultrasound noninvasively measures the IJV diameter as an alternative to CVP measurement[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Our study further supports the potential of ultrasound parameters of the IJV Ratio in assessing the CVP.\u003c/p\u003e \u003cp\u003eThis study innovatively proposed DSTA-Net as the image segmentation model and DM-MLP as the clinical prediction model for our wearable ultrasound device. Compared with the DeepLabV3+, Swin-Unet, TransUNet, and UNet models, the DSTA-Net model achieved more precise segmentation performance, especially in terms of the Dice for IJV segmentation. In terms of the R-area and R-perimeter, the DSTA-Net model also exhibited more balanced performance, better controlling the shape and area of the predicted region, and avoiding overfitting or unrealistic predictions. The excellent agreement and high correlation between the manual and DSTA-Net measurement parameters of the IJV and CCA among patients with varying CVP levels also demonstrated the effectiveness of the DSTA-Net model. Compared with manual measurement, wearable ultrasonic automatic measurement significantly reduces the operation time, improves the efficiency of diagnosis and treatment, and could achieve continuous monitoring in the future, which is challenging to manual measurement[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Furthermore, DM-MLP model integrated manually- and DSTA-NET-labeled IJV and CCA images as well as clinical data to predict CVP elevation with an AUCs of 0.94 and 0.88, respectively. Compared with the VGG, ResNet, and Transformer models, the DM-MLP model has a greater, accuracy, sensitivity, specificity, PPV, and NPV, indicating superior predictive performance. Although the effectiveness of the DSTA-Net DM-MLP model was lower than that of the Manual DM-MLP model, both were highly effective at predicting elevated CVP. However, the DM-MLP model showed high PPV and NPV in predicting elevated CVP, which may be attributed to the relatively small sample size in our study. In future research, we will increase the accuracy of the DSTA-Net DM-MLP model by increasing the sample size and optimizing the experimental process, to narrow the gap with manually labeled data, and to reduce the PPV and NPV. Additionally, expanding the dataset to include diverse patient groups and clinical conditions would help to improve the model's generalizability and clinical applicability.\u003c/p\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThis study has several limitations. Firstly, although the DM-MLP model performed well, the sample size for clinical prediction was relatively small, necessitating a larger and more diverse patient population. Further, we only examined the effectiveness of predicting elevated CVP; a detailed assessment of the feasibility and comfort of long-term wear for patients is lacking, largely due to the challenges of working with unconscious ICU patients. Future studies with an expanded sample size should include evaluations of human applicability, comfort, and operability. Secondly, the DM-MLP clinical prediction model a non-transparent model that provides limited insight, making it difficult to understand the mechanics underling predictions. Future research should explore interpretable machine learning models to enhance patient trust and understanding. Thirdly, this study only measured the maximum and minimum cross-sectional area of the IJV during a respiratory cycle, without real-time monitoring of variations throughout the cycle. In the future, we will conduct further in-depth research focusing on these aspects.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study integrates wearable ultrasound devices with AI models to achieve portable, noninvasive assessment of elevated CVP, with the potential to revolutionize fluid management and improve patient outcomes, making a significant contribution to the field of noninvasive medical monitoring. Further research should allow the integration of wearable ultrasound devices, while AI could extend from intensive care to other clinical settings, such as outpatient monitoring or home care, providing patients with more portable and continuous assessment tools.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCVP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCentral venous pressure\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCVC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCentral venous catheter\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArtificial intelligence\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIJV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eInternal jugular vein\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCommon carotid artery\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eICUs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIntensive care units\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDSTA-Net\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDual-decoder spatiotemporal attention network\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDM-MLP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDual-mixing MLP\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eECG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eElectrocardiogram\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIJV Max Area\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eThe maximum IJV area at the end of expiration\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIJV Min Area\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eThe minimum IJV area at the end of inspiration\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIJV Max/CCA Area\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eThe maximum IJV area /CCA area\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIJV Ratio\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eThe change rate of the IJV\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBSA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBody surface area\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBMI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBody mass index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSBP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSystolic blood Pressure\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDBP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDiastolic blood Pressure\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHeart rate\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDice\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDice coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIoU\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIntersection over Union\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eB-A analysis\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBland‒Altman analysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArea under the curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNegative predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePositive predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe sincerely thank Zixuan Zheng, PhD (MindRank AI Co., Ltd., Wuxi, China.) for her contribution to image segmentation and annotation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGuarantors of integrity of entire study, L.P.D., M.L., Y.L., X.X.Z., T.Y., J.Y.G., S.R.L., X.P.S., W.K.J., X.S.W., F.Y.W., L.P.Z., Y.Y.Z.; project management, funding support, and acquisition of research resources: Y.Y.Z., L.P.Z.; research design, experimental implementation, and data analysis, drafting of the manuscript: L.P.D.; participation in data collection, proposing key research hypotheses: M.L., T.Y., S.R.L., X.P.S., W.K.J., X.S.W., F.Y.W.; statistical analysis: L.P.D., J.Y.G.; review of the manuscript and provision of revision suggestions: Y.L., X.X.Z.,Y.Y.Z.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupported by the Exploratory Clinical Project at the Institutional Level of Shanghai Sixth People\u0026apos;s Hospital Affiliated to Shanghai Jiao Tong University (No: yynts202302) and the Shanghai Municipal Health Commission Clinical Research Program (No: 20244Y0119).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved the institutional review board of Shanghai Sixth People\u0026apos;s Hospital (institutional review board number: 2023-007-(3); approval date: August 30, 2023), and performed in accordance with the ethical standards as laid down in the 1964\u003c/p\u003e\n\u003cp\u003eDeclaration of Helsinki and its later amendments. All patients or their first-degree relatives provided written informed consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no competing interests to disclose.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMullens W, Abrahams Z, Francis GS, Sokos G, Taylor DO, Starling RC, Young JB, Tang WHW: Importance of venous congestion for worsening of renal function in advanced decompensated heart failure. J Am Coll Cardiol 2009, 53(7):589\u0026ndash;596.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLala A, McNulty SE, Mentz RJ, Dunlay SM, Vader JM, AbouEzzeddine OF, DeVore AD, Khazanie P, Redfield MM, Goldsmith SR \u003cem\u003eet al\u003c/em\u003e: Relief and Recurrence of Congestion During and After Hospitalization for Acute Heart Failure: Insights From Diuretic Optimization Strategy Evaluation in Acute Decompensated Heart Failure (DOSE-AHF) and Cardiorenal Rescue Study in Acute Decompensated Heart Failure (CARESS-HF). Circ Heart Fail 2015, 8(4):741\u0026ndash;748.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEvans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, Machado FR, McIntyre L, Ostermann M, Prescott HC \u003cem\u003eet al\u003c/em\u003e: Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock 2021. Crit Care Med 2021, 49(11):e1063-e1143.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang CY, Guiza F, Wouters P, Mebis L, Carra G, Gunst J, Meersseman P, Casaer M, Van den Berghe G, De Vlieger G \u003cem\u003eet al\u003c/em\u003e: Development and validation of the creatinine clearance predictor machine learning models in critically ill adults. Crit Care 2023, 27(1):272.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCole R: Does central venous pressure predict fluid responsiveness? Chest 2008, 134(6):1351\u0026ndash;1352.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGee DC, Gould MK: Preventing complications of central venous catheterization. N Engl J Med 2003, 348(12):1123\u0026ndash;1133.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang L, Harrison J, Dranow E, Aliyev N, Khor L: Accuracy of Ultrasound Jugular Venous Pressure Height in Predicting Central Venous Congestion. Ann Intern Med 2022, 175(3):344\u0026ndash;351.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimon MA, Kliner DE, Girod JP, Moguillansky D, Villanueva FS, Pacella JJ: Detection of elevated right atrial pressure using a simple bedside ultrasound measure. Am Heart J 2010, 159(3):421\u0026ndash;427.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimon MA, Schnatz RG, Romeo JD, Pacella JJ: Bedside Ultrasound Assessment of Jugular Venous Compliance as a Potential Point-of-Care Method to Predict Acute Decompensated Heart Failure 30-Day Readmission. J Am Heart Assoc 2018, 7(15):e008184.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu H, Hu C, Guo W, Zhu B, Wang S: Wearable ultrasound devices: An emerging era for biomedicine and clinical translation. Ultrasonics 2024, 142:107401.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKenny JS, Munding CE, Eibl AM, Eibl JK: Wearable ultrasound and provocative hemodynamics: a view of the future. Crit Care 2022, 26(1):329.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eA wearable ultrasound patch for continuous heart imaging. Nature 2023.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu H, Huang H, Li M, Gao X, Yin L, Qi R, Wu RS, Chen X, Ma Y, Shi K \u003cem\u003eet al\u003c/em\u003e: A wearable cardiac ultrasound imager. Nature 2023, 613(7945):667\u0026ndash;675.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHill B, Smith C: Central venous pressure monitoring in critical care settings. Br J Nurs 2021, 30(4):230\u0026ndash;236.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamzaoui O, Teboul JL: Central venous pressure (CVP). Intensive Care Med 2022, 48(10):1498\u0026ndash;1500.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoger C, Muller L, Riou B, Molinari N, Louart B, Kerbrat H, Teboul JL, Lefrant JY: Comparison of different techniques of central venous pressure measurement in mechanically ventilated critically ill patients. Br J Anaesth 2017, 118(2):223\u0026ndash;231.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, Gerlach H, Moreno R, Carlet J, Le Gall JR, Payen D \u003cem\u003eet al\u003c/em\u003e: Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 2006, 34(2):344\u0026ndash;353.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNational Heart L, Blood Institute Acute Respiratory Distress Syndrome Clinical Trials N, Wiedemann HP, Wheeler AP, Bernard GR, Thompson BT, Hayden D, deBoisblanc B, Connors AF, Jr., Hite RD \u003cem\u003eet al\u003c/em\u003e: Comparison of two fluid-management strategies in acute lung injury. N Engl J Med 2006, 354(24):2564\u0026ndash;2575.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThomas Hill B: Role of central venous pressure monitoring in critical care settings. Nurs Stand 2018, 32(23):41\u0026ndash;48.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDavison R, Cannon R: Estimation of central venous pressure by examination of jugular veins. Am Heart J 1974, 87(3):279\u0026ndash;282.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGee SR: Physical examination of venous pressure: a critical review. Am Heart J 1998, 136(1):10\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuge M, Marhefka GD: IVC measurement for the noninvasive evaluation of central venous pressure. J Echocardiogr 2022, 20(3):133\u0026ndash;143.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBayraktar M, Kacmaz M: Correlation of internal jugular vein, common carotid artery, femoral artery and femoral vein diameters with central venous pressure. Medicine (Baltimore) 2022, 101(43):e31207.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSekiguchi H, Seaburg LA, Suzuki J, Astorne WJ, Patel AS, Keller AS, Gajic O, Kashani KB: Central venous pressure and ultrasonographic measurement correlation and their associations with intradialytic adverse events in hospitalized patients: A prospective observational study. J Crit Care 2018, 44:168\u0026ndash;174.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKosnik N, Kowalski T, Lorenz L, Valacer M, Sakthi-Velavan S: Anatomical review of internal jugular vein cannulation. Folia Morphol (Warsz) 2024, 83(1):1\u0026ndash;19.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang M, Zhao L, Wang X, Lo WLA, Wen J, Li L, Huang Q: Automatic extraction and measurement of ultrasonic muscle morphological parameters based on multi-stage fusion and segmentation. Ultrasonics 2024, 137:107187.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou GQ, Chan P, Zheng YP: Automatic measurement of pennation angle and fascicle length of gastrocnemius muscles using real-time ultrasound imaging. Ultrasonics 2015, 57:72\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Central venous pressure, Dual-Decoder Spatiotemporal Attention Network, Dual-Mixing MLP, Internal jugular vein, Common carotid artery","lastPublishedDoi":"10.21203/rs.3.rs-6316972/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6316972/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose \u003c/strong\u003eCurrently, central venous catheterization (CVC) remains the reference standard for evaluating central venous pressure (CVP) in critically ill patients, but its invasiveness and associated complications limit its use. This study combined a wearable ultrasound device with artificial intelligence (AI) models to achieve portable noninvasive assessment of CVP in critically ill patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e This study prospectively enrolled critically ill patients who underwent CVC. We divided participants into a CVP≥8 mmHg group and a CVP\u0026lt;8 mmHg group, and recorded the internal jugular vein (IJV) and common carotid artery (CCA) via wearable ultrasound device. A Dual-Decoder Spatiotemporal Attention Network (DSTA-Net) was trained to enable automatic measurement of wearable ultrasound data and was evaluated via Bland-Altman consistency and Spearman correlation analyses. Moreover, it was compared with existing DeepLabV3+, Swin-Unet, TransUNet, and UNet models. The baseline patient parameters were subsequently combined with the wearable ultrasound parameters to construct a dual-mixing multilayer perception (DM-MLP) model to forecast elevated CVP,which was compared with existing VGG, ResNet, and Transformer models. Diagnostic performance was assessed via receiver operating characteristic (ROC) curves.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults \u003c/strong\u003eVideos from 272 patients (218:54 [80%:20%] in the training and test sets) were used for DSTA-Net training. Both sets were divided into the CVP≥8 mmHg and CVP\u0026lt;8 mmHg groups. In both sets, the CVP and the manually/DSTA-Net-measured IJV Max Area, IJV Min Area, and IJV Max/CCA Area were greater in the CVP≥8mmHg group. Compared with the DeepLabV3+, Swin-Unet, TransUNet, and UNet models, the DSTA-Net model showed higher Dice and IoU values of 82.62±0.18 and 74.97±0.15 for IJV segmentation, and 80.09±0.19 and 71.71±0.17 for CCA segmentation, respectively. Bland-Altman and Spearman correlation analyses confirmed a high degree of consistency and correlation between manually- and DSTA-Net- measured wearable ultrasound parameters. The ROC curve AUCs of the clinical parameters and manually/DSTA-Net measured DM-MLP models were 0.94[0.88,0.94] and 0.88[0.85,0.91], respectively, indicating better predictive performance than the VGG, ResNet, and Transformer models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion \u003c/strong\u003eThese results suggest that integrating wearable ultrasound devices with DSTA-Net as an image segmentation model and DM-MLP as a clinical prediction model provides a portable, noninvasive method for predicting elevated CVP.\u003c/p\u003e","manuscriptTitle":"Assessment of wearable Ultrasound Device Combined with AI for Portable Assessment of Central Venous Pressure Compared with Central Venous Catheterization as the reference Standard in critically ill patients: A Cross-Sectional Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-30 23:38:44","doi":"10.21203/rs.3.rs-6316972/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2a7d2447-e9c9-40ac-a9ff-0a5e21ffe6f2","owner":[],"postedDate":"April 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-05-19T07:08:43+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-30 23:38:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6316972","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6316972","identity":"rs-6316972","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}