Colorectal cancer (CRC) is a leading cause of cancer deaths worldwide. Although polypectomy at early stage reduces CRC incidence, 90% of the polyps are small and diminutive, where removal of them poses risks to patients that may outweigh the benefits. Correctly detecting and predicting polyp type during colonoscopy allows endoscopists to resect and discard the tissue without submitting it for histology, saving time, and costs. Nevertheless, human visual observation of early stage polyps varies. Therefore, this paper aims at developing a fully automatic algorithm to detect and classify hyperplastic and adenomatous colorectal polyps. Adenomatous polyps should be removed, whereas distal diminutive hyperplastic polyps are considered clinically insignificant and may be left in situ . A novel transfer learning application is proposed utilizing features learned from big nonmedical datasets with 1.4-2.5 million images using deep convolutional neural network. The endoscopic images we collected for experiment were taken under random lighting conditions, zooming and optical magnification, including 1104 endoscopic nonpolyp images taken under both white-light and narrowband imaging (NBI) endoscopy and 826 NBI endoscopic polyp images, of which 263 images were hyperplasia and 563 were adenoma as confirmed by histology. The proposed method identified polyp images from nonpolyp images in the beginning followed by predicting the polyp histology. When compared with visual inspection by endoscopists, the results of this study show that the proposed method has similar precision (87.3% versus 86.4%) but a higher recall rate (87.6% versus 77.0%) and a higher accuracy (85.9% versus 74.3%). In conclusion, automatic algorithms can assist endoscopists in identifying polyps that are adenomatous but have been incorrectly judged as hyperplasia and, therefore, enable timely resection of these polyps at an early stage before they develop into invasive cancer.
13 May 2019 in IEEE Journal of Biomedical and Health Informatics
Estimating hospital mortality of patients is important in assisting clinicians to make decisions and hospital providers to allocate resources. This paper proposed a multi-task recurrent neural network with attention mechanisms to predict patients' hospital mortality, using reconstruction of patients' physiological time series as an auxiliary task. Experiments were conducted on a large public electronic health record database, i.e., MIMIC-III. Fifteen physiological measurements during the first 24 h of critical care were used to predict death before hospital discharge. Compared with the conventional simplified acute physiology score (SAPS-II), the proposed multi-task learning model achieved better sensitivity (0.503 ± 0.020 versus 0.365 ± 0.021), when predictions were made based on the same 24-h observation period. The multi-task learning model is recommended to be updated daily with at least a 6-h observation period, in order for it to perform similarly or better than the SAPS-II. In the future, the need for intervention can be considered as another task to further optimize the performance of the multi-task learning model.
13 July 2016 in IEEE Transactions on Biomedical Engineering
Objective: Wireless capsule endoscope (WCE) is a revolutionary approach to diagnose small bowel pathologies. Currently available WCEs are mostly passive devices with image capturing function only, while on-going efforts have been placed on robotizing WCEs or to enhance them with therapeutic functions. In this paper, the authors present a novel inflatable WCE for haemostasis in the gastrointestinal (GI) tracts by balloon tamponade effect. Methods: The proposed wireless capsule consists of a balloon that can be inflated using the endothermic reaction of acid and base. When the balloon reached a precalculated pressure level, it is able to stop at a bleeding site in the bowel, and achieve haemostasis by tamponade effect. The prototype is 14 mm in diameter, with three sections of 13, 35, and 12 mm in length, respectively. The three sections are linked together with flexible joints and enclosed in a silicone balloon. The prototypes were tested in ex vivo porcine intestine models. Results: In the ten ex vivo trials conducted, the inflatable wireless capsule achieved average balloon pressure of 46.0 mmHg and withstood average maximum longitudinal pulling force at 1.46 N. An in vivo study was carried out as a proof-of-concept for treating bleeding in a porcine model. The proposed inflatable WCE succeeded in the animal test by controlling haemostasis within 5 min. No rebleeding was observed in the next 20 min. Conclusion: The results suggested that the inflatable capsule with a real-time bleeding detection algorithm can be implemented. Moreover, the proposed inflatable WCE prototype can achieve haemorrhage control in the lower GI. Significance: To our best knowledge, this is the first study that demonstrated the potential to treat GI haemorrhage by an inflatable WCE. The proposed capsule enables the development of a closed-loop system based on a body sensor network to provide early treatment of GI bleeding for p-medicine.
Using the complex sink-source model (CSSM) and the Hertz potential method (HPM), the electromagnetic field expressions of tightly focused ultrashort azimuthally polarized pulses can be obtained. By numerically solving the relativistic Newton-Lorentz equation, the acceleration and confinement of electrons by the sub-cycle and few-cycle azimuthally polarized ultrashort pulses in vacuum are studied. Considering the radiation reaction force, it is found that electrons with an initial kinetic energy of less than 1MeV can be accelerated to hundreds of MeV and can be confined in the range of less than 1 micron for hundreds of femtoseconds in the direction perpendicular to the pulse propagation (transverse direction) by the pulses. With the increase of the beam waist and the intensity of the pulse, the electrons can obtain the exit kinetic energy exceeding 1GeV. When electrons are accelerated by the few-cycle pulses, the confined time of the electrons in the transverse direction is three times longer than that of the sub-cycle pulse. When the initial velocity of the electron points to a point in front of the focus, the electron can obtain the maximum exit kinetic energy. The change of the angular frequency corresponding to the spectral peak of the electromagnetic radiation from the electron acceleration with the electric field amplitude parameter E0 of the pulse is studied. The phenomena of redshift and blueshift of the spectrum peak frequency of the electron radiation with the E0 are found. These studies provide the methods to confine the movement of electrons in certain directions and accelerate electrons in the same time.
Accurate cuffless blood pressure (BP) estimation remains challenging, particularly under dynamic conditions with significant intra-individual BP variations. This study introduces UTransBPNet, a novel, calibration-free model for cuffless BP estimation. It integrates a squeeze-and-excitation-enhanced Unet architecture for short-range feature extraction with a transformer and cross attention module to capture long-range dependencies from high-resolution, multi-channel physiological signals, further refined through an optimized fine-tuning scheme. Comprehensive validations were conducted across multiple dynamic datasets—Dataset_Drink, Dataset_Exercise, and Dataset_MIMIC—in both scenario-specific and cross-scenario settings. Results demonstrate that UTransBPNet outperformed existing models in tracking BP variations under dynamic conditions, achieving individual Pearson’s correlation coefficients of 0.61 ± 0.17 and 0.62 ± 0.13 for systolic BP (SBP) and diastolic BP (DBP) in Dataset_Drink, 0.82 ± 0.11 and 0.72 ± 0.18 in Dataset_Exercise, and low mean absolute differences (MADs) of 4.38 and 2.25 mmHg in Dataset_MIMIC. The analysis also highlights the impact of dataset characteristics on model performance, such as distribution shift, distribution imbalance and individual BP variability, highlighting the need for well-curated data to ensure generalizability. This study advances the development of robust, cuffless BP estimation models for real-world applications.
1 October 2021 in Computers in Biology and Medicine
Pulse arrival time (PAT) has been broadly investigated for its potential for cuffless blood pressure (BP) estimation and ease of measurement by wearable devices. It is also of great significance to explore whether PAT conveys complementary information to BP for vascular health assessment. In this paper, the differences between the 24-h ambulatory BP and wearable-based PAT were compared among 12 young normotensives and 15 elderly hypertensives in terms of the mean and coefficients of variation (CoVs). The correlations of the nocturnal normalized PAT (n-PAT) and BP with two arterial stiffness-related parameters (i.e., the intrinsic elastic modulus E-0 and the vascular modulation factor alpha) estimated by a proposed model-based method were also compared. The results showed that the inter-subject variances of the nocturnal average n-PAT were significantly different between the hypertensives and the normotensives (P < 0.001), and the intra-subject CoVs of the nocturnal n-PAT were also significantly different between the two groups (P < 0.05). However, these findings were not shown in the nocturnal BP. The correlation coefficient between the nocturnal average n-PAT and ln(E-0) is larger than that with the nocturnal BP, i.e., 0.91 vs. 0.56. Furthermore, the result also revealed that all the hypertensives receiving antihypertensive medications did not achieve the optimal control of the nocturnal BP while presented diverse arterial stiffness indicated by the nocturnal average n-PAT and ln(E-0). It is concluded that wearable-based PAT contains complementary information about the vascular system to the ambulatory BP, which may be useful for designing effective antihypertensive treatments.