... IEEE Signal Processing in Medicine and Biology Symposium (SPMB). IEEE Signal Processing in Medicine and Biology Symposium最新文献

With the use of a wireless, wearable, passive knitted smart fabric device as a strain gauge sensor, the proposed algorithm can estimate biomedical feedback such as respiratory activity. Variations in physical properties of Radio Frequency Identification (RFID) signals can be used to wirelessly detect physiological processes and states. However, it is typical for ambient noise artifacts to appear in the RFID signal making it difficult to identify physiological processes. This paper introduces a new technique for finding these repetitive physiological signals and identifying them into two states, active and inactive, using k-means clustering. The algorithm detects these biomedical events without the need to completely remove the noise components using a semi-unsupervised approach, and with these results, predict the next biomedical event using these classification results. This approach enables real-time noninvasive monitoring for use with actuating medical devices for therapy. Using this approach, the algorithm predicts the onset of respiratory activity in a simulated environment within approximately one second.

Central venous stenosis is often undiagnosed in patients with hemodialysis vascular access, partly due to imaging difficulties. Noninvasive, point-of-care detection could rely on detecting regions of turbulent blood flow caused by blood velocity changes. Here we present flexible microphone arrays for time-correlated measures of blood flow sounds and a new signal processing approach to calculate time correlation between spectral features. Continuous wavelet transform was used to produce an auditory spectral flux analytic signal, which was thresholded to identify systolic start and end phases. Microphone arrays were tested on pulsatile flow phantoms with blood flow rates of 850-1,200 mL/min and simulated stenosis from 10-85%. Measured results showed an inversion in the time onset of systolic spectral content for sites proximal and distal to stenosis for hemodynamically significant stenoses (+22 ms for stenosis<50% and -20 to -38 ms for stenosis>50%). Equivalent blood velocity increases were calculated as 142-155 cm/s in stenotic phantoms, which are within the physiologic range as measured by ultrasound.

Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervous system development in embryonic mice. High-frequency ultrasound (HFU) is the only non-invasive, real-time modality available for rapid volumetric imaging of embryos in utero. However, manual segmentation of the BV from HFU volumes is tedious, time-consuming, and requires specialized expertise. In this paper, we propose a novel deep learning based BV segmentation system for whole-body HFU images of mouse embryos. Our fully automated system consists of two modules: localization and segmentation. It first applies a volumetric convolutional neural network on a 3D sliding window over the entire volume to identify a 3D bounding box containing the entire BV. It then employs a fully convolutional network to segment the detected bounding box into BV and background. The system achieves a Dice Similarity Coefficient (DSC) of 0.8956 for BV segmentation on an unseen 111 HFU volume test set surpassing the previous state-of-the-art method (DSC of 0.7119) by a margin of 25%.

The overarching framework for incorporating informatics into the Wesley College (Wesley) undergraduate curriculum was to teach emerging information technologies that prepared undergraduates for complex high-demand work environments. Federal and State support helped implement Wesley's undergraduate Informatics Certificate and Minor programs. Both programs require project-based coursework in Applied Statistics, SAS Programming, and Geo-spatial Analysis (ArcGIS). In 2015, the State of Obesity listed the obesity ranges for all 50 US States to be between 21-36%. Yet, the Center for Disease Control and Prevention (CDC) mortality records show significantly lower obesity-related death-rates for states with very high obesity-rates. This study highlights the disparities in the reported obesity-related death-rates (specified by an ICD-10 E66 diagnosis code) and the obesity-rate percentages recorded for all 50 US States. Using CDC mortality-rate data, the available obesity-rate information, and ArcGIS, we created choropleth maps for all US States. Visual and statistical analysis shows considerable disparities in the obesity-related death-rate record-keeping amongst the 50 US States. For example, in 2015, Vermont with the sixth lowest obesity-rate had the highest reported obesity-related death-rate. In contrast, Alabama had the fifth highest adult obesity-rate in the nation, yet, it had a very low age-adjusted mortality-rate. Such disparities make comparative analysis difficult.

Vascular access dysfunction is the leading cause of hospitalization for hemodialysis patients and accounts for the most medical costs in this patient population. Vascular access flow is commonly hindered by blood vessel narrowing (stenosis). Current screening methods involving imaging to detect stenosis are too costly for routine use at the point of care. Noninvasive, real-time screening of patients at risk of vascular access dysfunction could potentially identify high-risk patients and reduce the likelihood of emergency surgical interventions. Bruits (sounds produced by turbulent blood flow near stenoses) can be interpreted by skilled clinical staff using conventional stethoscopes. To improve the sensitivity of detection, digital analysis of blood flow sounds (phonoangiograms or PAGs) is a promising approach for classifying vascular access stenosis using non-invasive auditory recordings. Here, we demonstrate auditory and spectral features of PAGs which estimate both the location and degree of stenosis (DOS). Auditory recordings from nine stenosis phantoms with variable DOS and hemodynamic flow rate were obtained using a digital recording stethoscope and analyzed to extract classification features. Autoregressive modeling and discrete wavelet transforms were used for multiresolution signal decomposition to produce 14 distinct features, most of which were linearly correlated with DOS. Our initial results suggest that the widely-used auditory spectral centroid is a simple way to calculate features which can estimate both the location and severity of vascular access stenosis.

Clinical electroencephalographic (EEG) data varies significantly depending on a number of operational conditions (e.g., the type and placement of electrodes, the type of electrical grounding used). This investigation explores the statistical differences present in two different referential montages: Linked Ear (LE) and Averaged Reference (AR). Each of these accounts for approximately 45% of the data in the TUH EEG Corpus. In this study, we explore the impact this variability has on machine learning performance. We compare the statistical properties of features generated using these two montages, and explore the impact of performance on our standard Hidden Markov Model (HMM) based classification system. We show that a system trained on LE data significantly outperforms one trained only on AR data (77.2% vs. 61.4%). We also demonstrate that performance of a system trained on both data sets is somewhat compromised (71.4% vs. 77.2%). A statistical analysis of the data suggests that mean, variance and channel normalization should be considered. However, cepstral mean subtraction failed to produce an improvement in performance, suggesting that the impact of these statistical differences is subtler.

To be effective, state of the art machine learning technology needs large amounts of annotated data. There are numerous compelling applications in healthcare that can benefit from high performance automated decision support systems provided by deep learning technology, but they lack the comprehensive data resources required to apply sophisticated machine learning models. Further, for economic reasons, it is very difficult to justify the creation of large annotated corpora for these applications. Hence, automated annotation techniques become increasingly important. In this study, we investigated the effectiveness of using an active learning algorithm to automatically annotate a large EEG corpus. The algorithm is designed to annotate six types of EEG events. Two model training schemes, namely threshold-based and volume-based, are evaluated. In the threshold-based scheme the threshold of confidence scores is optimized in the initial training iteration, whereas for the volume-based scheme only a certain amount of data is preserved after each iteration. Recognition performance is improved 2% absolute and the system is capable of automatically annotating previously unlabeled data. Given that the interpretation of clinical EEG data is an exceedingly difficult task, this study provides some evidence that the proposed method is a viable alternative to expensive manual annotation.

The interpretation of electroencephalograms (EEGs) is a process that is still dependent on the subjective analysis of the examiners. Though interrater agreement on critical events such as seizures is high, it is much lower on subtler events (e.g., when there are benign variants). The process used by an expert to interpret an EEG is quite subjective and hard to replicate by machine. The performance of machine learning technology is far from human performance. We have been developing an interpretation system, AutoEEG, with a goal of exceeding human performance on this task. In this work, we are focusing on one of the early decisions made in this process - whether an EEG is normal or abnormal. We explore two baseline classification algorithms: k-Nearest Neighbor (kNN) and Random Forest Ensemble Learning (RF). A subset of the TUH EEG Corpus was used to evaluate performance. Principal Components Analysis (PCA) was used to reduce the dimensionality of the data. kNN achieved a 41.8% detection error rate while RF achieved an error rate of 31.7%. These error rates are significantly lower than those obtained by random guessing based on priors (49.5%). The majority of the errors were related to misclassification of normal EEGs.