A storage-efficient ensemble classification using filter sharing on binarized convolutional neural networks.

This paper proposes a storage-efficient ensemble classification to overcome the low inference accuracy of binary neural networks (BNNs). When external power is enough in a dynamic powered system, classification results can be enhanced by aggregating outputs of multiple BNN classifiers. However, memory requirements for storing multiple classifiers are a significant burden in the lightweight system. The proposed scheme shares the filters from a trained convolutional neural network (CNN) model to reduce storage requirements in the binarized CNNs instead of adopting the fully independent classifier. While several filters are shared, the proposed method only trains unfrozen learnable parameters in the retraining step. We compare and analyze the performances of the proposed ensemble-based systems depending on various ensemble types and BNN structures on CIFAR datasets. Our experiments conclude that the proposed method using the filter sharing can be scalable with the number of classifiers and effective in enhancing classification accuracy. With binarized ResNet-20 and ReActNet-10 on the CIFAR-100 dataset, the proposed scheme can achieve 56.74% and 70.29% Top-1 accuracies with 10 BNN classifiers, which enhances performance by 7.6% and 3.6% compared with that using a single BNN classifier.

Predicting hypotension in the ICU using noninvasive physiological signals.

Hypotension frequently occurs in Intensive Care Units (ICU), and its early prediction can improve the outcome of patient care. Trends observed in signals related to blood pressure (BP) are critical in predicting future events. Unfortunately, the invasive measurement of BP signals is neither comfortable nor feasible in all bed settings. In this study, we investigate the performance of machine-learning techniques in predicting hypotensive events in ICU settings using physiological signals that can be obtained noninvasively. We show that noninvasive mean arterial pressure (NIMAP) can be simulated by down-sampling the invasively measured MAP. This enables us to investigate the effect of BP measurement frequency on the algorithm's performance by training and testing the algorithm on a large dataset provided by the MIMIC III database. This study shows that having NIMAP information is essential for adequate predictive performance. The proposed predictive algorithm can flag hypotension with a sensitivity of 84%, positive predictive value (PPV) of 73%, and F1-score of 78%. Furthermore, the predictive performance of the algorithm improves by increasing the frequency of BP sampling.

Supervised Machine-Learning Algorithms in Real-time Prediction of Hypotensive Events.

Hypotension is common in critically ill patients. Early prediction of hypotensive events in the Intensive Care Units (ICUs) allows clinicians to pre-emptively treat the patient and avoid possible organ damage. In this study, we investigate the performance of various supervised machine-learning classification algorithms along with a real-time labeling technique to predict acute hypotensive events in the ICU. It is shown that logistic regression and SVM yield a better combination of specificity, sensitivity and positive predictive value (PPV). Logistic regression is able to predict 85% of events within 30 minutes of their onset with 81% PPV and 96% specificity, while SVM results in 96% specificity, 83% sensitivity and 82% PPV. To further reduce the false alarm rate, we propose a high-level decision-making algorithm that filters isolated false positives identified by the machine-learning algorithms. By implementing this technique, 24% of the false alarms are filtered. This saves 21 hours of medical staff time through 2,560 hours of monitoring and significantly reduces the disturbance caused by alarming monitors.

Immunity of nanoscale magnetic tunnel junctions with perpendicular magnetic anisotropy to ionizing radiation.

Spin transfer torque magnetic random access memory (STT-MRAM) is a promising candidate for next generation memory as it is non-volatile, fast, and has unlimited endurance. Another important aspect of STT-MRAM is that its core component, the nanoscale magnetic tunneling junction (MTJ), is thought to be radiation hard, making it attractive for space and nuclear technology applications. However, studies on the effects of ionizing radiation on the STT-MRAM writing process are lacking for MTJs with perpendicular magnetic anisotropy (pMTJs) required for scalable applications. Particularly, the question of the impact of extreme total ionizing dose on perpendicular magnetic anisotropy, which plays a crucial role on thermal stability and critical writing current, remains open. Here we report measurements of the impact of high doses of gamma and neutron radiation on nanoscale pMTJs used in STT-MRAM. We characterize the tunneling magnetoresistance, the magnetic field switching, and the current-induced switching before and after irradiation. Our results demonstrate that all these key properties of nanoscale MTJs relevant to STT-MRAM applications are robust against ionizing radiation. Additionally, we perform experiments on thermally driven stochastic switching in the gamma ray environment. These results indicate that nanoscale MTJs are promising building blocks for radiation-hard non-von Neumann computing.

A machine-learning approach to predicting hypotensive events in ICU settings.

BACKGROUND: Predicting hypotension well in advance provides physicians with enough time to respond with proper therapeutic measures. However, the real-time prediction of hypotension with high positive predictive value (PPV) is a challenge. This is due to the dynamic changes in patients' physiological status following drug administration, which limits the quantity of useful data available for the algorithm. METHOD: To mimic real-time monitoring, we developed a machine-learning algorithm that uses most of the available data points from patients' records to train and test the algorithm. The algorithm predicts hypotension up to 30 min in advance based on the data from only 5 min of patient physiological history. A novel evaluation method is also proposed to assess the performance of the algorithm as a function of time at every timestamp within 30 min of hypotension onset. This evaluation approach provides statistical tools to find the best possible prediction window. RESULTS: During about 181,000 min of monitoring of 400 patients, the algorithm demonstrated 94% accuracy, 85% sensitivity and 96% specificity in predicting hypotension within 30 min of the events. A high PPV of 81% was obtained, and the algorithm predicted 80% of hypotensive events 25 min prior to onset. It was shown that choosing a classification threshold that maximizes the F1 score during the training phase contributes to a high PPV and sensitivity. CONCLUSIONS: This study demonstrates the promising potential of machine-learning algorithms in the real-time prediction of hypotensive events in ICU settings based on short-term physiological history.

Hospital enterprise Architecture Framework (Study of Iranian University Hospital Organization).

BACKGROUNDS: Nowadays developing smart and fast services for patients and transforming hospitals to modern hospitals is considered a necessity. Living in the world inundated with information systems, designing services based on information technology entails a suitable architecture framework. OBJECTIVES: This paper aims to present a localized enterprise architecture framework for the Iranian university hospital. METHODS AND RESULTS: Using two dimensions of implementation and having appropriate characteristics, the best 17 enterprises frameworks were chosen. As part of this effort, five criteria were selected according to experts' inputs. According to these criteria, five frameworks which had the highest rank were chosen. Then 44 general characteristics were extracted from the existing 17 frameworks after careful studying. Then a questionnaire was written accordingly to distinguish the necessity of those characteristics using expert's opinions and Delphi method. The result showed eight important criteria. In the next step, using AHP method, TOGAF was chosen regarding having appropriate characteristics and the ability to be implemented among reference formats. In the next step, enterprise architecture framework was designed by TOGAF in a conceptual model and its layers. For determining architecture framework parts, a questionnaire with 145 questions was written based on literature review and expert's opinions. The results showed during localization of TOGAF for Iran, 111 of 145 parts were chosen and certified to be used in the hospital. CONCLUSION: The results showed that TOGAF could be suitable for use in the hospital. So, a localized Hospital Enterprise Architecture Modelling is developed by customizing TOGAF for an Iranian hospital at eight levels and 11 parts. This new model could be used to be performed in other Iranian hospitals.

Discontinuous unilateral involvement of 12 part core biopsies by adenocarcinoma predicts bilateral involvement of subsequent radical prostatectomy.

At our institution, percent tumor burden in prostate core biopsies is quantified using variations of one of two methods. Measurement by the Aggregate method reports only adenocarcinoma and omits intervening stroma and benign prostatic glands while the Discontinuous method includes the intervening stroma and benign glands between distinct foci of adenocarcinoma. In this study, we selected cases with 12-part core biopsies that were followed by a radical prostatectomy within two years. Interestingly, we found that when adenocarcinoma involved prostate 12-part core biopsies and subsequent resection unilaterally, there is no significant difference in absolute percentage of tumor using either measuring method (P = 0.4). In contrast, when adenocarcinoma involved the biopsies unilaterally and subsequent prostatectomy bilaterally, the two measurement methods had a statistically significant difference in percentage scores (P = 0.002). In the study cohort, other factors including Gleason score (P = 0.88) and total number of adenocarcinoma-involved cores (P = 0.27) did not introduce any significant correlation with bilateral involvement. In this study, we found that biopsies that discontinuously and unilaterally involve half of a prostate are much more likely to involve both lobes than those that are unilateral and present in nodular aggregates.