BACKGROUND: Intraoperative hypotension (IOH) and tachycardia are associated with perioperative myocardial injury (PMI), and thereby increased postoperative mortality. Patients undergoing vascular surgery are specifically at risk of developing cardiac complications. This study aimed to explore the association between different thresholds for IOH and tachycardia, and PMI. It also aimed to explore which threshold for IOH and tachycardia best predicts PMI.

METHODS: In this single-center prospective observational study, high-sensitivity cardiac troponin T was measured preoperatively and at 4, 24, and 48 hours after vascular surgery. Absolute and relative thresholds were used to define intraoperative systolic, mean, and diastolic arterial hypotension, measured every 15 seconds by invasive arterial pressure monitoring and heart rate using the Philips IntelliVue X3 monitor. Decision tree machine-learning (ML) models were used to explore which thresholds for IOH and tachycardia best predict PMI. Clinical utility and transparency were prioritized over maximizing the performance of the ML model and therefore a white-box model was used.

RESULTS: In all, 498 patients were included in the study. Ninety-nine patients (20%) had PMI. Significant associations were found between IOH and PMI using both absolute and relative thresholds for systolic, mean, and diastolic arterial pressure. Absolute thresholds based on diastolic arterial pressure had the strongest correlation with PMI and yielded greater statistical significance. The threshold that was most predictive of PMI was an absolute diastolic arterial pressure <44 mm Hg. The prediction model with the absolute threshold of diastolic arterial pressure <44 mm Hg had a macro average F1 score of 0.67 and a weighted average F1 score of 0.76. No association was found between tachycardia and PMI.

CONCLUSIONS: We found that an absolute, not relative, IOH threshold based on diastolic arterial pressure, and not systolic or mean arterial pressure, or tachycardia, was most predictive of PMI.

Objectives. To quantify and evaluate the dynamic response of RR intervals (RRI) and heart rate (HR) measurements of commercially available Bluetooth chest-worn HR monitors during induced rapid changes in HR.

Approach. An arbitrary function generator created synthetic electrocardiogram signals simulating the heart activity. Different scenarios of rapid changes in HR were simulated several times using: (1) step responses; (2) exercise data (EX); and (3) intermittent EX data. RRI and HR were recorded using the standard Bluetooth HR service for four wearable monitors: Garmin HRM-Dual, Movesense active, Polar H10, and Wahoo TRACKR. RRI latency, HR latency, and agreement were evaluated from the reference signal.

Main results. RRI latency (median and interquartile range) was 0.7(0.5,0.7) s for Garmin, 0.4(0.2,0.5) s for Movesense, 2.6(2.2,2.8) s for Polar, and 2.1(1.9,2.4) s for Wahoo, where results did not differ greatly between tests. HR response latency was different between devices and tests. During intermittent EX tests, HR latency was 3.3(3.0, 3.3) s for Garmin, 1.0(1.0,1.0) s for Movesense, 2.3(2.3,2.3) s for Polar, and 2.2(2.2,2.3) s for Wahoo, where all devices consistently underestimated HR peaks and overestimated HR valleys, with a greater discrepancy in HR valleys.

Significance. Most validation protocols of RRI and HR measured by wearable monitors neglect their dynamic characteristics. The present study demonstrated that manufacturers implemented different digital filters to compute the HR values, limiting the devices’ ability to capture rapid HR changes. Open documentation of the processing steps is advised, and use cases involving sharp HR changes—such as intermittent high-intensity training—should rely on beat-to-beat RRI recordings.

OBJECTIVE: To develop models to predict opportunities for improvement in trauma care and compare the performance of these models to the currently used audit filters.

DESIGN: Retrospective registry-based study.

SETTING: Single-centre, Scandinavian level one equivalent trauma centre.

PARTICIPANTS: 8220 adult trauma patients screened for opportunities for improvement between 2013 and 2022.

PRIMARY AND SECONDARY OUTCOME MEASURES: Two machine learning models (logistic regression and XGBoost) and the currently used audit filters were compared. Internal validation by an expanding window approach with annual updates was used for model evaluation. Performance measured by discrimination, calibration, sensitivity and false positive rate of opportunities for improvement prediction.

RESULTS: A total of 8220 patients, with a mean age of 45 years, were analysed; 69% were men with a mean injury severity score of 12. Opportunities for improvement were identified in 496 (6%) patients. Both the logistic regression and XGBoost models were well-calibrated, with intercalibration indices of 0.02 and 0.02, respectively. The models demonstrated higher areas under the receiver operating characteristic curve (AUCs) (logistic regression: 0.71; XGBoost: 0.74). The XGBoost model had a lower false positive rate at a similar sensitivity (false positive rate: 0.63). The audit filters had an AUC of 0.62 and a false positive rate of 0.67.

CONCLUSIONS: The logistic regression and XGBoost models outperformed audit filters in predicting opportunities for improvement among adult trauma patients and can potentially be used to improve systems for selecting patients for trauma peer review.

Intraoperative hypotension (IOH) is associated with an increased risk of heart and kidney complications. Although AI tools aim to predict IOH, their real-world reliability is often overstated due to biased data selection. This study introduces a framework to enhance reliability by: (1) including borderline blood pressure cases (65–75 mmHg, the “Gray Zone”), (2) comparing AI model to simple blood pressure threshold, and (3) validating across diverse surgical cohorts, centers and demographics. Using datasets from Karolinska University Hospital (Sweden) and VitalDB (Korea), we found AI model performs better than MAP threshold method in more ambiguous cases. In contrast, when hypotensive and non-hypotensive cases had clearly separated MAP values, both methods performed similarly well. Cross-validation revealed asymmetric generalizability: models trained on datasets containing more borderline (Gray Zone) cases generalized better to datasets with clearer class separation, whereas the reverse struggled. To ensure fair model comparison and reduce dataset-specific bias, we standardized the MAP difference between positive (hypotension) and negative (non-hypotension) samples at the time of prediction. This virtually eliminated the class separation and demonstrated that inflated performance in some datasets can be attributed to selection bias rather than true model generalizability. Age also influenced generalization: Cross-age validation revealed models trained on older patients generalized better to younger cohorts, whereas differences in ASA classification had minimal effect. These findings highlight the need for realistic validation to bridge the gap between AI research and clinical practice.

Video is the most used tool for sport performance analysis as it provides a common reference point for the coach and the athlete. The problem with video is that it is a subjective tool. To overcome this, motion capture systems can used to get an objective 3D model of a per- son’s posture and motion, but only in laboratory settings. Unfortunately, many activities, such as most outdoor sports, cannot be captured in a lab without compromising the activity. In this paper, we propose to use an aerial drone system equipped with depth cameras, AI-based marker- less motion capture software to perform automatic skeleton tracking and real-time sports performance analysis of athletes. We experiment with off-the-shelf drone systems, miniaturized depth cameras, and commer- cially available skeleton tracking software to build a system for analyzing sports-related performance of athletes in their real settings. To make this a fully working system, we have conducted a few initial experiments and identified many issues that still needs to be addressed.

Background Trauma is one of the leading causes of morbidity and mortality worldwide. Morbidity and mortality review of selected patient cases is used to improve the quality of trauma care by identifying opportunities for improvement (OFI). The aim of this study was to assess how patient and process factors are associated with OFI in trauma care.

Methods We conducted a registry-based study using all patients between 2017 and 2021 from the Karolinska University Hospital who had been reviewed regarding the presence of OFI as defined by a morbidity and mortality conference. We used bi- and multivariable logistic regression to assess the associations between the following patient and process factors and OFI: age, sex, respiratory rate, systolic blood pressure, Glasgow Coma Scale (GCS), Injury Severity Score (ISS), survival at 30 days, highest hospital care level, arrival on working hours, arrival on weekends, intubation status and time to first computed tomography (CT).

Results OFI was identified in 300 (5.8%) out of 5182 patients. Age, missing Glasgow Coma Scale, time to first CT, highest hospital care level and ISS were statistically significantly associated with OFI.

Conclusion Several patient and process factors were found to be associated with OFI, indicating that patients with moderate to severe trauma and those with delays to first CT are at the highest odds of OFI.

In modern hospitals, monitoring patients’ vital signs and other biomedical signals is standard practice. With the advent of data-driven healthcare, Internet of medical things, wearable technologies, and machine learning, we expect this to accelerate and to be used in new and promising ways, including early warning systems and precision diagnostics. Hence, we see an ever-increasing need for retrieving, storing, and managing the large amount of biomedical signal data generated. The popularity of standards, such as HL7 FHIR for interoperability and data transfer, have also resulted in their use as a data storage model, which is inefficient. This article raises concern about the inefficiency of using FHIR for storage of biomedical signals and instead highlights the possibility of a sustainable storage based on data compression. Most reported efforts have focused on ECG signals; however, many other typical biomedical signals are understudied. In this article, we are considering arterial blood pressure, photoplethysmography, and respiration. We focus on simple lossless compression with low implementation complexity, low compression delay, and good compression ratios suitable for wide adoption. Our results show that it is easy to obtain a compression ratio of 2.7:1 for arterial blood pressure, 2.9:1 for photoplethysmography, and 4.1:1 for respiration.

Low-Power Wide Area Networks have contributed in several parts of the Internet of Things ecosystem during the last years by enabling long range, robust and low power communication. Machine Learning for embedded systems has also assisted the advancement of the Internet of Things by identifying patterns and increasing the accuracy of predicting events and behaviours. At the same time, wearable and mobile systems are less obtrusive, consuming less energy and have more computing resources. In this paper we combine these three components and propose a low cost wearable system based on a regular shoe and off-the-shelf electronics which is able to recognize foot gestures and transmit messages over long range, in cases of emergency. The evaluation considers an application scenario where the user performs specific foot gestures to trigger the transmission of an emergency message, during other activities (e.g., walking). The proposed wearable system would benefit a user who is in danger and attempts to notify her/his emergency contacts in a discreet manner. Results show that the proposed system is able to identify the intended foot gestures with 98% accuracy.

This work focuses on predicting near-term onset of hypotension prior to onset using convolutional neural networks. Based solely on the arterial blood pressure curve, our initial attempt can predict an onset with 60% sensitivity and 80% specificity 5-15 minutes before onset.

Clinical relevance Hypotension is common during large surgery. By identifying and treating hypotensive episodes early, preferably even before onset, hypotension and its associate post- surgery complications are reduced. Even a prediction with 80% sensitivity/specificity is valuable for the anesthesiologist. 

The advent of Low-Power Wide Area Networks has enabled significant developments of the IoT ecosystem. Long range communication using low power is now feasible and offers connectivity to remote areas where cellular network is not available. Therefore, new application scenarios have emerged, such as smart cities, smart metering and more, which are attracting a lot of attention from both research and industry. Beside the aforementioned popular scenarios, Low-Power Wide Area Networks have started to be used in wearable systems scenarios as well. In this position paper, we pose some questions regarding the Human Computer Interaction aspects of Low-Power Wide Area Networks which will help them integrate in Ubiquitous Computing applications. We illustrate by a wearable system, which is based on an foot gesture interface, a Low-Power Wide Area Network, and an Neural Network classifier. The discussion is based on the state of art of foot interfaces and highlights open issues and challenges. Copyright