Search

Determining when a patient can be discharged from a care setting is critical to optimize the utilization and delivery of timely care. Furthermore, timely discharge can lead to better clinical outcomes by effectively mitigating the prolonged length of stay in a care environment. This paper presents a novel algorithm for the prediction of likelihood of patient discharge within the next 24 or 48 hours from acute or critical care environments on a daily basis. Continuous patient monitoring and health data obtained from acute hospital at home environment (n=303 patients) and a critical care unit environment (n=9,520 patients) are retrospectively used to train, validate and test numerous machine learning models for dynamic daily predictions of patients discharge. In the acute hospital at home environment, the area under the receiver operating characteristic (AUROC) curve performance of a top XGBoost model was 0.816 ± 0.025 and 0.758 ± 0.029 for daily discharge prediction within 24 hours and 48 hours respectively. Similar independent prediction models from the critical care environment resulted in relatively a lower AUROC for likewise predicting daily patient discharge. Overall, the results demonstrate the efficacy and utility of our novel algorithm for dynamic predictions of daily patient discharge in both acute- and critical care healthcare settings.

This paper consists of two main parts. First, a brief review on two band-pass filters based on N-path passive mixers are given, for low-IF application, along with harmonics rejection techniques. This review provides a good insight about the concept of N-path filters, with focus on the ideas necessary for presenting the proposed filter. Then, a new structure is presented for bandpass filters for use in low-IF applications. The filter structure provides a significant noise-Figure performance improvement in comparison with similar works. Furthermore, the size of baseband capacitors are significantly decreased, while obtaining a high quality factor, which saves a considerable die area. Designed and simulated in a standard 90-nm CMOS technology, the filter covers the frequency range from 200 MHz to 1.11 GHz by exploiting two digital clock generators. Over the entire frequency range, the analog power consumption is only 21 mW, and the achieved quality factor is higher than 100.

This paper presents a new fully-integrated CMOS single-to-differential low-noise amplifier (LNA) topology for cognitive radios. Operating from 50 MHz to 5 GHz, the proposed wideband LNA uses two current-reuse shunt-feedback (SFB) amplifiers as the core circuit and an inverter amplifier as the active balun. The transconductance of the amplifiers is boosted by utilizing both nMOS and pMOS transistors in complementary structures. Designed in TSMC 0.18 µ m CMOS technology, the LNA achieves a voltage gain of 25.3 dB and a noise figure of 3.6 to 5.0 dB, with a power dissipation of 16.7 mW. The complementary characteristics leads to partial second-order distortion cancellation. The achieved IIP3 and IIP2 are from 0 to 9.5 dBm and > 31 dBm, respectively. The proposed inductorless balun-LNA is post-layout simulated and it occupies only 0.0094 mm2.

In this paper, a method for signal folding improvement in N-path filters is presented. By applying a new technique based on pre-filtering to a former structure, the difference in gain between the desired signal and folded harmonics increases. The proposed N-path filter also performs single-ended to differential conversion to reject harmonic selectivity around even harmonics. The new structure, which is simulated using 90nm technology, has a frequency tuning range of 800MHz that varies from 1.3GHz to 2.1GHz. The input impedance matching and NF are, respectively, less than −10dB and 4.6dB in the entire frequency range. The in-band IIP3 is better than 5dBm, and the power consumption of the non-ideal clock generator varies from 2.9mW to 4.8mW depending on the filter center frequency.

This paper discusses computational modeling of predictive risk factors for neonates undergoing a Norwood surgical procedure, a multi-stage cardiac procedure that restores functional systemic circulation in patients such as neonates with Hypoplastic Left Heart Syndrome (HLHS). In this model, we apply machine learning based binary classication to 549 cases reported by the Pediatric Heart Networks Single Ventricle Reconstruction Trial. We use neural networks classier to predict risk factors for individual patients undergoing a Norwood procedure for the repair of HLHS. Results indicate that independent risk can be calculated with 85% accuracy and 0.94 area under the receiver operating characteristics curve. This model may help physicians provide counseling for families and medically optimize patients prior to surgery by modifying individual risk factors.

This study presents a comparative performance analysis including delay, power and power-delay product (PDP) of ternary systems using copper and large diameter MWCNT bundle interconnects at 14 nm technology node. Comparisons are done for intermediate interconnects and repeated and unrepeated global level interconnects. It is shown that at the intermediate level, MWCNT bundle interconnects can achieve about 55% lower delay and PDP than their copper counterparts. Similar to the intermediate level, for unrepeated global level interconnects, MWCNT bundle interconnects have lower PDP as compared to Cu wires. This improvement is about 5% for unrepeated global interconnects. For repeated global interconnects copper wires have about 14% lower PDP than the repeated global level MWCNT bundle interconnects while the number of repeaters for these wires is more than the number of repeaters for MWCNT bundle interconnects.

In this paper, we investigate the effectiveness of active shielded copper and MWCNT bundle interconnects to eliminate the crosstalk induced functional failure in ternary systems. This study is done for the intermediate and global level interconnects at 22 and 14 nm technologies. The results indicate that the copper interconnects, even with the Shield lines, are not able to eliminate the crosstalk-induced functional failures in ternary logic systems while active shielded MWCNT bundle interconnects is able to omit the crosstalk induced functional failure for interconnects in global and intermediate level at the technologies under investigation. The average worst-case crosstalk noise peak in ternary logic using active shielded MWCNT bundle interconnects is reduced by 33% and 40% for the intermediate interconnects at 22nm and 14nm technologies and 37% and 42% for global level interconnects at 22nm and 14nm technology nodes compared to their Cu counterparts. In addition, active shielded MWCNT bundle interconnects have lower crosstalk delay for long wires compared to active shielded copper interconnects.

Alarm fatigue in intensive care units (ICU) is one of the top healthcare issues in the US. False alarms in ICU will decrease the quality of care and staff response time over the alarms. Normally, false alarm will cause desensitization of the clinical staff which leads to warnings and misleading, if the triggered alarm is true. In this study, we have proposed a multi-model ensemble approach to reduce the false alarm rate in monitoring systems. We have used 750 patient records from PhysioNet database. At First arrhythmia based features from electrocardiogram (ECG), arterial blood pressure (ABP) and photoplethysmogram (PPG) features were extracted from the records. Next, the dataset has been separated into two subsets on the basis of available features information. The first dataset (DS1) is the combination of ECG physiological, ABP and PPG features. Their correlation coefficient and p-values criteria have been applied for relevant alarm-wise feature-set selection, and random forest classifier was used for model development and validation. The threshold based approach was used on second dataset (DS2) which is the combination of arrhythmia, ABP and PPG features. The developed ensemble model is able to achieve sensitivity 83.33-100 % (average 95.56 %) being true alarms and suppress false alarms rate 66.67-89% (average 77.25%). The predictability of classifier shows the advantage to deal with unbalanced set of information, therefore overall model performance has reached to 83.96% accuracy.

A receiver front-end using active/passive N-path mixers techniques is presented in this paper. By applying some techniques to the former similar structure, NF is improved significantly and stability is guaranteed. NF and S11 in all over 2GHz frequency spectrum are less than 4dB and −10dB, respectively. The power consumption without considering ideal clock generator and combiner is 11.8mW and out of band IIP3 reaches to +15.9dBm.

In photovoltaic sources, operating current and voltage, due to variable environmental conditions, will change. Since the operating point can maximize output power, one approach to use a maximum power point tracking seems to be necessary. Typically MPPT algorithms are implemented on a solar system utilizing a switching power converter. The used DC-DC converter in this paper is a boost; therefore choosing an adequate topology of boost converter is very important. In this paper resonant boost converters have been compared with conventional ones.

In this work, time domain analysis is used to solve Adler's equation in order to obtain the required time, for an oscillator under external injection, reaching the steady-state condition. Then, the analysis is verified by simulations of an LF ring as well as a typical 3.65 GHz RF-LC oscillator under injection. After that, a delay-based time-domain analysis is used to reveal the dependency of circuit parameters to the locking time, for ring oscillators. Finally, a design methodology is deduced which enables designers to evaluate and minimize the timing budget required to achieve injection locking in designing a fast locking ring oscillator.

In this paper, new single supply voltage level converters (SSLCs) based on the Carbon nanotube field effect transistors (CNTFETs) are proposed for low-power design. Nowadays, in short channel devices, low power design especially in battery-powered digital VLSI systems becomes a very significant challenge. Multi-V DD design is an effective approach for lowering the power consumption without degrading the speed of the integrated circuits. In these circuits in order to convert the voltage levels between the different voltage islands, voltage level converters (LCs) are required. On the other hand, CNTFET refrains most of the limitations of the nanoscale MOSFET devices. Moreover, the threshold voltage of the CNFETs can be determined by adopting proper diameters for their nanotubes. The proposed multi-V TH CNTFET-based design method reduces the number of transistors and considerably improves the energy efficiency and robustness of the LC circuits. The results of the simulations, conducted based on the Stanford CNTFET model at 22nm technology node, demonstrate the superiority of the proposed designs in terms of power consumption, delay, PDP and sensitivity to major process variations compared to the other common and most efficient LC circuits.

Scheduling of real-time tasks in multiprocessor systems is a NP-hard problem. Recently, swarm intelligence algorithms have been efficiently applied for this problem. Real-time tasks can be classified into hard real-time tasks and soft real-time tasks. The aim of hard real-time task scheduling algorithms is to meet all tasks deadline constraints. However, slight violation is not critical, in the case of soft real-time tasks. In this paper, a new algorithm based on artificial bee colony (ABC) is proposed for scheduling of soft real-time tasks. In this method, a hybrid neighborhood search mechanism is introduced to improve the convergence of ABC. Experimental results demonstrate the effectiveness of proposed algorithm for scheduling of soft realtime tasks in heterogeneous multiprocessor systems.

A new single Miller Capacitor frequency compensation scheme for three stage amplifier is proposed. In conventional three-stage amplifier a differential feedback, from output of the second and the third stage is added to output of the first stage, which significantly improves Phase Margin and decreases compensation capacitor to less than 2% of the load capacitor value. Analyses show that the stability can be perfectly ensured. A three-stage amplifier has been simulated with and without a differential feedback in a 0.18μm CMOS. The proposed amplifier driving a 100pF capacitance load, achieved 1.39MHz gain-bandwidth product and 90° Phase Margin with a 1.8V supply. An amplifier based on conventional nested Miller compensation (NMC) can just achieve less than 0.23MHz GBW with the same load, while using more than 100pF as compensation capacitor. So this method shows an improvement of a factor of 6 in GBW and reduction of a factor of 50 in the size of compensation capacitor. It is a suitable strategy for ON-CHIP compensation in comparison to other methods.

In this paper, we present a MATLAB program for automatic symbolic simplification of analog circuits containing MOSFETs, using modified nodal analysis (MNA) and ant colony optimization (ACO). At first, all MOSFETs are replaced by the corresponding small-signal models. Then, the circuit be analyzed by applying symbolic MNA, and the exact symbolic expression of the circuit behavior is generated. The derived exact symbolic expression then be simplified using ACO. In this paper, a new criterion was introduced to minimize the mean square error (MSE) between the exact expression and the simplified one. The main advantage of proposed criterion is that final simplification error rate is controllable by user. In this way, the gain and phase MSEs across different frequencies were considered to evaluate the solutions generated by artificial ants. It is remarkable that all processing containing netlist text processing, symbolic analysis, post-processing and simplification are consecutively derived in MATLAB. Comparing the obtained numerical results with HSPICE demonstrates the efficiently of proposed tool.

Among the voltage vectors of even level inverters there is no voltage vector with zero common mode voltage. Therefore bearing currents, which may cause premature motor bearing failures, will not be eliminated due to generated common mode voltages. To eliminate these currents in even level inverters, instead of elimination of common mode voltage, the main idea is based on an extreme increment of the parasitic coupling impedance through which currents are flying. This increment of the impedance is resulted by reducing frequency of common mode voltage to zero. To achieving this aim, only voltage vectors with an equal common mode voltage are used. All calculations for switching between these vectors have been done with two independent methods. Comparison between these methods has shown that none of them has the absolute prominence.

Digital control of a series-parallel resonant converter has been analyzed in this paper. To design the controller, the reduced third-order model of the converter has been used. Utilizing this model, the influence of the time delay which is mainly due to the calculation time of the processor, is discussed. Considering the results, it can be concluded that stability and transient behavior of the system are affected by the time delay value. Although, linear prediction is the suggested method of this paper to compensate these effects, but by increasing the resonant frequency, the error of this method will increase and consequently the compensation will not be achievable.

In this paper a new structure for a multilevel inverter has been proposed which its operation is based on a combination of diode-clamped and two level full bridge inverters. Major advantage of the proposed topology is reduction in the number of switches compared to the conventional topologies. Since increasing the number of levels, may lead to probably problems in diode clamped section of the topology, a solution has been also presented.

The recognition of face images is a complicated problem. Face images are often sufferedfrom variations in brightness, head rotation, facial emotions and so on. Besides, amazing abilities of human brain in face recognition in the presence of these variations, contribute to design face recognition systems based on procedure of human brain. Surveying the recognition and perceptual system of human, shows that, this system has hierarchical and bidirectional structure. Furthermore, the performance of the system would strongly be improved by applying the information of upper layers of face recognition system in interpreting and processing the input data. In this paper, novel bidirectional architecturefor face recognition inspired by human face recognition system is presentedvia applying inversion in artificial neural networks (ANN's). In this approach, storeddata in the inverse networkis applied in the recognition system iterativelyandthen the correctness of face recognition model has been consequently improved by 8%. The proposed model is able to produce 12 various facial expressions on the output, from only one input expressionof each person, after training with AUT database images.

This paper considers the recovery of multiple sparse vectors, with partially shared supports, from a small number of noisy linear measurements of each. It has recently been shown that it is possible to lower the sample complexity of recovery, for all levels of support sharing, by using a “dirty model”: a super-position of sparse and group-sparse modeling approaches; this is based on convex optimization.

A new class of digital FIR filters with application to the decimation filter design for ΣΔ Analog to Digital Converters (ADCs) is introduced, which can be realized using an efficient multiplier-less structure. The filter coefficients are conditioned in such a way that, independent of the order, the realization structure comprises shifts (delays) and only two additions; hence the proposed filters have close relation to the first order Cascaded Integrator-Comb (CIC) filters, but offering more design parameters to overcome their limited degree of freedom. A filter Involving Shifts and Only Two Additions (abbreviated as ISOTA), has coefficients dependant to each other; this together with the non-linearity of the discrete-space of the coefficients, makes the design procedure somewhat limited. A simple design method is used to find the desired solutions, based on applying gradient search algorithm to the possible combinations of the coefficients (obtained by state tree diagram). Demonstrative examples as well as the practical application are presented.

in this paper, a new method to compensate three stage amplifier is proposed, second non-dominant pole cancellation of multi-path nested Miller compensation (MNMC) are paid. Using the feed forward signal and changing the location of the second zero, second non-dominant pole is removed and behavior of system is approached to the first order system. To verify the method, one practical example is presented in 0.18um CMOS technology. The simulation results confirm effectiveness of proposed method to compensate three-stage amplifier.

This paper provides the circuits which are needed to implement RBF Neural Networks. A novel circuit to implement an M-dimensional Euclidean distance is presented. All the blocks were designed in weak inversion region to gain ultra low power consumption. The results were done using HSPICE by level 49 parameters (BSIM3v3) in 0.35µm standard CMOS technology. Finally, an RBF neuron with five hidden layers to approximate a nonlinear function is implemented. Having confirmed that the proposed circuit is working properly, simulation results have shown that the power consumption of our circuits is about several microwatts.

RSA encryption is one of the public-key methods that has been popular in last decade. Considering increment of security requirements, size of the keys has been larger. With key length growing, delay of exponentiation computation has changed into major problem in selecting longer keys. The binary or in other words square-and-multiply method is the classical exponentiation technique that is used in RSA. In this paper a new algorithm of exponentiation in RSA is presented that works in parallel, needs fewer multiplications and so has less delay. Therefore this technique is more useful in larger key computations.

In this paper, we develop an optimal scheme for conferencing (i.e. all-to-all communication), in communication networks. Our scheme uses the minimum sum rate. First we state a lower bound on sum rate and show how to achieve it on networks with tree topology. This result can be used on general networks considering the spanning tree which connects the conferencing nodes.

Multi-site cooperative transmission is gaining industrial support in latest standard development. One key criteria to sucessfully implement a multi-site cooperative transmission system is the need to quantize an expanded set of channels. To maintain quantization precision, codebook size needs to scale roughly exponentially with the number of total antennas and renders quantizing the expanded channel set a computationally demanding task. In this paper, we identify a connection between noncoherent communication and the phase invariant property of beamforming vectors. We leverage techniques from the noncoherent communication literature to design an efficient beamforming vector quantization scheme. Our proposed scheme has a quantization complexity that grows linearly with number of antennas. Simulation results indicate this method has slightly more quantization distortion than RVQ while being significantly more efficient.

An approach to classify disorders in autonomic control of cardiovascular system is proposed in this paper. The target of this study is to highlight main features of malfunctions in cardiovascular system due to autonomic disorder. Collecting the data from the physionet archive, we divide patients into two groups of normal and abnormal, based on having autonomic disorder in their cardiovascular system or not. Systolic blood pressure (SBP) and heart rate (HR) time series are evaluated for each patient. We then plot the diagram of SBP against HR for all patients in a single figure. Fuzzy c-means clustering (FCM) method is also applied to cluster data into two groups. A neural network is then implemented to classify and to distinguish the two groups. The network is trained with data of a normal patient and is tested with data of other normal and abnormal patients. Result show that selected features can clearly detect disorders in autonomic system.

Objective: The aim of this study is to found how well one can characterize the location and extent of relatively compact infarcts using electrocardiographic evidence. Method: Here, we address a method base on behavior of some ECG's features, which are Q, amplitudes and ST dispersion. We call these Q and ST curves. At the first step, by plotting the variability of Q, amplitude and ST dispersion for nodes which lies in lines on torso plane, these curves are obtained. The behavior of the mentioned curves for normal lines both in horizontal and vertical line differ from the abnormal ones. A threshold method is used here to determine the infarcted area. Results and Conclusion: The method is evaluated on Challenge 2007 database. The results are EPD=8, SO=0.944, and CED=1. The method achieved the best EPD and CED scores and the second place for SO and overall ranked the highest scores (first rank) in CinC/PhysioNet Challenge 2007.

Hypothesis/Objective: The aim of this study is to characterize the location and extent of moderate to large, relatively compact infarcts using ECG evidence. Method: In this paper, we proposed a method on the basis of vectorcardiography which assumes that heart vector is proportional to relevant active depolarization area(s). To examine our ideas, we used the normal VCG which includes the information of location, amplitude, and direction of heart vector at any instant. The model based comparison of cases under study and relevant normal VCGs gives region i.e. segment(s) and depth i.e. extent of myocardial infarction. Results and Conclusion: We evaluated the method on CinC/Physionet Challenge 2007 database. In our final entry the scores of EPD equal to 32 (ranked 3rd), SO equal to 0.933 (ranked 3rd) and CED equal to 1 (ranked 1st) are achieved. It also ranked the third among the other methods proposed to CinC/Physionet Challenge 2007.

In this paper a design for implementation of an ANC system with online secondary path modeling on FPGA devices is proposed. There are two adaptive units in this ANC system, the adaptive main control filter with the FxLMS algorithm and the adaptive modeling filter with the LMS algorithm for updating the adaptive coefficients array. Computer simulations show that the proposed design achieves the same performance as the existing methods.

This paper provides an overview of methods on robust model predictive control (RMPC) from 1999- 2006. First, some optimization problems in model predictive control (MPC) have been reviewed. Then, the methods based on model uncertainty and the methods, which are using disturbance uncertainties, based on optimization problem are considered. Finally, applied and simulated methods of proposed problem are stated.