Pre-Symptomatic Sepsis Detection Using a Wearable Biosensing System and Deep Learning Framework

Glossary

qSOFA - Quick Sequential Organ Failure Assessment; a bedside sepsis screening score using three clinical criteria, with poor discriminative accuracy

DS-TCN - Dilated Separable Temporal Convolutional Network; the project's core deep learning architecture for time-series sepsis prediction

CBAM - Convolutional Block Attention Module; attention mechanism that focuses the model on the most clinically relevant signals and time points

AUROC - Area Under the Receiver Operating Characteristic Curve; measures overall model discrimination between septic and non-septic cases

AUPRC - Area Under the Precision-Recall Curve; measures predictive performance under class imbalance

ECE - Expected Calibration Error; measures how closely predicted probabilities match real-world outcomes

OHEM - Online Hard Example Mining; training strategy that prioritizes the most difficult misclassified samples

MICE - Multiple Imputation by Chained Equations; statistical method used to reconstruct missing physiological data during augmentation

NPV - Negative Predictive Value

PLR - Positive Likelihood Ratio; how strongly a positive result updates disease probability

ECG - Electrocardiogram; measures cardiac electrical activity to capture heart rate and rhythm

PPG - Photoplethysmography; optical sensing method measuring pulse waveform and blood oxygen saturation

SpO₂ - Blood Oxygen Saturation; percentage of oxygenated hemoglobin in the blood

BLE - Bluetooth Low Energy; wireless protocol

PMIC - Power Management Integrated Circuit

SNR - Signal-to-Noise Ratio; measure of signal quality

ICU - Intensive Care Unit

The Problem

Sepsis is a cascading immune overreaction to infection that can escalate to critical organ failure. It affects 50 million patients annually, responsible for 31.5% of all global mortality and over $62 billion in economic impact (Fig. 1 and 3)(Rudd et al., 2020). Persistent organ dysfunction prolongs ICU care, overwhelming healthcare systems and driving families into financial ruin. In low-income countries, where sepsis burden is highest, the barrier to survival is rarely treatment; it is recognition. Each hour of diagnostic delay increases mortality by 7–10%.

Sepsis detection remains limited by:

Intermittent monitoring.

Low accuracy scores like qSOFA (AUC ~0.63).

Delayed lab-based confirmation

Many systems also fail in low-resource settings due to infrastructure gaps (Wong et al., 2021).

The Solution

Sepsis develops through complex, rapidly interacting changes in the body that traditional diagnostic systems simplify into rigid rules, often detecting illness after it has significantly progressed (Fig. 2). Machine learning instead learns directly from high-dimensional, time-series clinical data (Fig. 4), capturing subtle, pre-symptomatic patterns across cardiovascular, respiratory, and metabolic signals that conventional threshold systems cannot detect, enabling earlier and precise intervention within the critical treatment window (Fleuren et al., 2020).

Design Objectives

This project pursued four objectives:

Detect sepsis before symptoms emerge by continuously monitoring patient data over time using ML, replacing fixed-interval scoring tools.

Maximize clinical reliability through high sensitivity, specificity, and risk predictions.

Design a lightweight, fully integrated wearable platform for continuous biosignal acquisition.

Enable deployment across clinical and low-resource settings without cloud reliance.

Background Research

The project integrated clinical literature review, deep learning design, and wearable prototyping, focusing on physiological markers predictive of sepsis onset and limitations of ICU scoring systems(Fig. 5). Datasets were selected for temporal resolution and clinical relevance.

Clinical Dataset

Cohort Size: 60,000 ICU patients across two hospital systems

Physiological Signals: Continuously monitored heart rate, respiratory rate, systolic/diastolic blood pressure, SpO₂, temperature

Class Imbalance: 12.8:1 negative-to-positive ratio reflecting real-world prevalence

Data integrity was ensured through patient-level separation, eliminating risk of information leakage between training and evaluation sets.

Data Preprocessing

Zeros were masked, and implausible values were clipped between the 1st–99th percentiles. Temporal gaps were imputed using training-set statistics. All features were z-score normalized and resampled to a 2-minute grid. To address imbalance, cases were augmented using Gaussian noise, time warping, and Multivariate Imputation by Chained-Equations (MICE), reducing imbalance to 2.6:1, beyond which augmentation risks synthetic artifacts.

Model Architecture

Multi-Scale Temporal Modelling

Four depthwise-separable temporal convolutional blocks with parallel kernels (3, 5, 7) and increasing dilation rates capture sudden physiological shifts and gradual deterioration patterns, enabling early recognition of sepsis progression.

Attention-Guided Prediction

A Convolutional-Block Attention Module (CBAM) emphasizes clinically relevant temporal features, while a multi-task head jointly predicts sepsis probability and time-to-onset for earlier, more precise detection.

Performance Enhancements

The architecture incorporates Temporal Label Smoothing, asymmetric focal loss with Online Hard-Example Mining (OHEM), and isotonic regression calibration, improving difficult near-onset classification and prediction confidence reliability(Fig. 6).

Wearable Hardware Platform

A wearable biosensing system integrated ECG, dual-wavelength PPG, dual-temperature sensing, and a 100 Hz accelerometer. A dual-core microcontroller separates sensing and processing to reduce jitter. Motion artifacts were reduced using accelerometer-based rejection. A low-power PMIC enabled duty cycling, flash storage buffered data to prevent transmission gaps, and Bluetooth Low Energy supported continuous streaming(Fig. 7-8).

Calibration and Clinical Operating Point

A core requirement for any clinical decision tool is that predicted probabilities reflect reality. Isotonic regression on a stratified hold-out split reduced error by 96.5%, ensuring predicted risks reflect real outcomes at every probability level (Fig. 10).

At decision threshold 0.42, 1,751 of 2,025 confirmed sepsis cases are correctly flagged at a false positive rate of just 2.37% across nearly 10,000 non-septic patients(Fig. 12).

The Positive Likelihood Ratio of 36.5 means a single model output constitutes strong standalone diagnostic evidence by Bayesian clinical standards, significantly improving a clinician's risk assessment.

A Negative Predictive Value of 97.20% confirms negative predictions are reliable enough to safely redirect scarce monitoring resources, a capability existing sepsis tools cannot provide continuously (Fig. 11).

Discriminative Performance

Evaluated across 11,776 held-out ICU patients at the real-world class imbalance, the model achieves AUROC 0.9806, correctly ranking a sepsis patient above a healthy patient 98% of the time, and AUPRC 0.9309. Because the AUPRC baseline is only 17.2%, achieving 0.9309 confirms the model detects real physiological sepsis signals, not statistical patterns (Fig. 9).

The system detects sepsis up to 12 hours before clinical onset from as little as four hours of continuous input, with optimal performance reached at six hours, placing its detection window over 48 hours ahead of typical clinical recognition.

Deployment Validation

The two main reasons sepsis models fail in practice are late detection and too many false alarms, both of which are penalized by the PhysioNet 2019 Challenge Utility Score.

A model that detects sepsis too late or produces alerts clinicians cannot trust will score poorly, even if its overall accuracy is high. Achieving 0.9012, at the top end of published results on this benchmark, shows the model avoids both problems and delivers real clinical value beyond strong statistical performance(Fig. 12).

Biosignal Sensing Architecture

The ECG subsystem resolves cardiac activity with a measured noise floor of 0.72 µV RMS, enabling clear identification of QRS complexes under resting conditions. The PPG channel achieves 115 dB signal-to-noise ratio, producing stable and artifact-resistant pulse waveforms with consistent morphology over extended recordings.

The dual-temperature system achieves ±0.1°C accuracy, and the differential between skin and ambient sensors remains stable, enabling reliable detection of subtle thermal gradients associated with vascular tone changes.

Accelerometer-based filtering identifies and removes motion-corrupted segments while preserving signal continuity, maintaining waveform integrity during movement without introducing distortion or aliasing artifacts.

Biosignal acquisition is designed to maintain temporal alignment across all sensor channels, ensuring that the ECG, PPG, and auxiliary signals can be meaningfully compared during model inference. Sampling stability was prioritized to minimize artifacts introduced by timing inconsistencies.

Power profiling confirms over 72 hours of continuous operation under standard duty-cycled use. The integrated PMIC gates inactive subsystems, significantly reducing energy consumption while maintaining uninterrupted sensing and wireless transmission via Bluetooth Low Energy.

Clinical Significance

Sepsis kills millions each year, yet its earliest warning signs are routinely missed until the disease has already progressed beyond easy intervention. This project directly addresses that gap. By continuously analyzing physiological signals from a wearable device, the system detects subtle biological changes that precede full septic deterioration before it becomes clinically apparent.

This early warning capability opens a critical window during which treatment can interrupt progression toward multi-organ failure, septic shock, and death.

Healthcare Impact

The benefits extend across the entire care environment:

For clinicians, the wearable platform provides continuous, real-time physiological monitoring that overcomes the limitations of periodic bedside assessments and delayed recognition.

For patients, this means a dramatically improved chance of survival and faster recovery.

For healthcare systems, it means fewer ICU admissions and significantly reduced treatment costs.

Its compact, self-contained design removes dependence on fixed hospital infrastructure, enabling deployment in resource-limited settings, from rural clinics to emergency field environments, broadening access to high-quality monitoring globally (Fig. 13-14).

Innovation and Novelty

My project introduces a novel combination of custom biosignal acquisition hardware and a deep learning model leveraging temporal pattern recognition and attention-based feature weighting. Together, these form a proactive predictive framework capable of interpreting complex, multi-parameter physiological data in real time, exceeding the diagnostic capability of conventional threshold-based clinical tools.

Broader Applications

Beyond sepsis, the framework extends to any condition with a prodromal signature, including respiratory failure and cardiac instability, establishing a generalizable methodology where clinical deterioration is systematically anticipated rather than reactively managed.

This system represents a scalable platform for continuous, real-time sepsis detection that can operate across ICUs, general wards, and resource-limited settings alike.

Future work will include:

Bioimpedance spectroscopy (preliminary development underway) to directly monitor fluid shifts and circulatory changes preceding septic shock.

Interpretability tools to identify which physiological patterns drive model predictions, helping clinicians understand and trust system alerts.

Verifying model generalizability on out-of-distribution cohorts (pediatric, post-surgical).

I would like to express my thanks to Dr. Marc De Benedetti and Mr. James Wang for their guidance and support. Their dedication to fostering a collaborative and inspiring environment has been a source of great strength and has truly made my CWSF journey both educational and rewarding. I truly appreciate all your contributions.

Allen, J. (2007). Photoplethysmography and its application in clinical physiological measurement. Physiological Measurement, 28(3), R1–R39. https://doi.org/10.1088/0967-3334/28/3/R01

Angus, D. C., Linde-Zwirble, W. T., Lidicker, J., Clermont, G., Carcillo, J., & Pinsky, M. R. (2001). Epidemiology of severe sepsis in the United States: Analysis of incidence, outcome, and associated costs of care. Critical Care Medicine, 29(7), 1303–1310. https://doi.org/10.1097/00003246-200107000-00002

Bahdanau, D., Cho, K., & Bengio, Y. (2015). Neural machine translation by jointly learning to align and translate. In Proceedings of the 3rd International Conference on Learning Representations (arXiv:1409.0473). arXiv. https://doi.org/10.48550/arXiv.1409.0473

Bai, S., Kolter, J. Z., & Koltun, V. (2018). An empirical evaluation of generic convolutional and recurrent networks for sequence modeling (arXiv:1803.01271). arXiv. https://doi.org/10.48550/arXiv.1803.01271

Bishop, C. M. (2006). Pattern recognition and machine learning. Springer. ISBN: 978-0-387-31073-2

Brealey, D., Brand, M., Hargreaves, I., Heales, S., Land, J., Smolenski, R., Davies, N. A., Cooper, C. E., & Singer, M. (2002). Association between mitochondrial dysfunction and severity and outcome of septic shock. The Lancet, 360(9328), 219–223. https://doi.org/10.1016/S0140-6736(02)09459-X

Caruana, R. (1997). Multitask learning. Machine Learning, 28(1), 41–75. https://doi.org/10.1023/A:1007379606734

Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16, 321–357. https://doi.org/10.1613/jair.953

Churpek, M. M., Snyder, A., Han, X., Sokol, S., Pettit, N., Howell, M. D., & Edelson, D. P. (2017). Quick Sepsis-related Organ Failure Assessment, Systemic Inflammatory Response Syndrome, and early warning scores for detecting clinical deterioration in infected patients outside the intensive care unit. American Journal of Respiratory and Critical Care Medicine, 195(7), 906–911. https://doi.org/10.1164/rccm.201604-0821OC

Churpek, M. M., Yuen, T. C., Winslow, C., Robicsek, A. A., Meltzer, D. O., Gibbons, R. D., & Edelson, D. P. (2014). Multicenter development and validation of a risk stratification tool for ward patients. American Journal of Respiratory and Critical Care Medicine, 190(6), 649–655. https://doi.org/10.1164/rccm.201406-1022OC

Du, N., Dai, H., Trivedi, R., Upadhyay, U., Gomez-Rodriguez, M., & Song, L. (2016). Recurrent marked temporal point processes: Embedding event history to vector. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 1555–1564). ACM. https://doi.org/10.1145/2939672.2939875

Evans, L., Rhodes, A., Alhazzani, W., Antonelli, M., Coopersmith, C. M., French, C., Machado, F. R., Mcintyre, L., Ostermann, M., Prescott, H. C., Schorr, C., Simpson, S., Wiersinga, W. J., Alshamsi, F., Angus, D. C., Arabi, Y., Azevedo, L., Beale, R., Beilman, G., … Levy, M. (2021). Surviving Sepsis Campaign: International guidelines for management of sepsis and septic shock 2021. Critical Care Medicine, 49(11), e1063–e1143. https://doi.org/10.1097/CCM.0000000000005337

Fleischmann, C., Scherag, A., Adhikari, N. K. J., Hartog, C. S., Tsaganos, T., Schlattmann, P., Angus, D. C., Reinhart, K., & International Forum of Acute Care Trialists. (2016). Assessment of global incidence and mortality of hospital-treated sepsis: Current estimates and limitations. American Journal of Respiratory and Critical Care Medicine, 193(3), 259–272. https://doi.org/10.1164/rccm.201504-0781OC

Fleuren, L. M., Klausch, T. L. T., Zwager, C. L., Schoonmade, L. J., Guo, T., Roggeveen, L. F., Swart, E. L., Girbes, A. R. J., Thoral, P. J., Ercole, A., Hoogendoorn, M., & Elbers, P. W. G. (2020). Machine learning for the prediction of sepsis: A systematic review and meta-analysis of diagnostic test accuracy. Intensive Care Medicine, 46(3), 383–400. https://doi.org/10.1007/s00134-019-05872-y

Gerry, S., Bonnici, T., Birks, J., Kirtley, S., Virdee, P. S., Watkinson, P. J., & Collins, G. S. (2020). Early warning scores for detecting deterioration in adult hospital patients: Systematic review and critical appraisal of methodology. BMJ, 369, Article m1501. https://doi.org/10.1136/bmj.m1501

Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press. http://www.deeplearningbook.org/

Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On calibration of modern neural networks. In D. Precup & Y. W. Teh (Eds.), Proceedings of the 34th International Conference on Machine Learning (Vol. 70, pp. 1321–1330). PMLR. https://proceedings.mlr.press/v70/guo17a.html

He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778). IEEE. https://doi.org/10.1109/CVPR.2016.90

Hinton, G., Vinyals, O., & Dean, J. (2015). Distilling the knowledge in a neural network (arXiv:1503.02531). arXiv. https://doi.org/10.48550/arXiv.1503.02531

Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735

Hotchkiss, R. S., Moldawer, L. L., Opal, S. M., Reinhart, K., Turnbull, I. R., & Vincent, J.-L. (2016). Sepsis and septic shock. Nature Reviews Disease Primers, 2, Article 16045. https://doi.org/10.1038/nrdp.2016.45

Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017). MobileNets: Efficient convolutional neural networks for mobile vision applications (arXiv:1704.04861). arXiv. https://doi.org/10.48550/arXiv.1704.04861

Ioffe, S., & Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd International Conference on Machine Learning (Vol. 37, pp. 448–456). PMLR. https://proceedings.mlr.press/v37/ioffe15.html

Jubran, A. (1999). Pulse oximetry. Critical Care, 3(2), R11–R17. https://doi.org/10.1186/cc341

Kumar, A., Roberts, D., Wood, K. E., Light, B., Parrillo, J. E., Sharma, S., Suppes, R., Feinstein, D., Zanotti, S., Taiberg, L., Gurka, D., Kumar, A., & Cheang, M. (2006). Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Critical Care Medicine, 34(6), 1589–1596. https://doi.org/10.1097/01.CCM.0000217961.75225.E9

Levi, M., & van der Poll, T. (2017). Coagulation and sepsis. Thrombosis Research, 149, 38–44. https://doi.org/10.1016/j.thromres.2016.11.007

Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal loss for dense object detection. In Proceedings of the IEEE International Conference on Computer Vision (pp. 2980–2988). IEEE. https://doi.org/10.1109/ICCV.2017.324

Lipton, Z. C., Kale, D. C., Elkan, C., & Wetzel, R. (2016). Learning to diagnose with LSTM recurrent neural networks (arXiv:1511.03677). arXiv. https://doi.org/10.48550/arXiv.1511.03677

Liu, S. F., & Malik, A. B. (2006). NF-κB activation as a pathological mechanism of septic shock and inflammation. American Journal of Physiology-Lung Cellular and Molecular Physiology, 290(4), L622–L645. https://doi.org/10.1152/ajplung.00477.2005

Mei, H., & Eisner, J. M. (2017). The neural Hawkes process: A neurally self-modulating multivariate point process. In I. Guyon et al. (Eds.), Advances in Neural Information Processing Systems (Vol. 30). Curran Associates. https://proceedings.neurips.cc/paper/2017/hash/6463c88460bd63bbe256e495c63aa40b-Abstract.html

Mukkamala, R., Hahn, J.-O., Inan, O. T., Mestha, L. K., Kim, C.-S., Toreyin, H., & Kyal, S. (2015). Toward ubiquitous blood pressure monitoring via pulse transit time: Theory and practice. IEEE Transactions on Biomedical Engineering, 62(8), 1879–1901. https://doi.org/10.1109/TBME.2015.2441951

Pan, J., & Tompkins, W. J. (1985). A real-time QRS detection algorithm. IEEE Transactions on Biomedical Engineering, 32(3), 230–236. https://doi.org/10.1109/TBME.1985.325532

Paoli, C. J., Reynolds, M. A., Sinha, M., Gitlin, M., & Crouser, E. (2018). Epidemiology and costs of sepsis in the United States — An analysis based on timing of diagnosis and severity level. Critical Care Medicine, 46(12), 1889–1897. https://doi.org/10.1097/CCM.0000000000003342

Payne, R. A., Symeonides, C. N., Webb, D. J., & Maxwell, S. R. J. (2006). Pulse transit time measured from the ECG: An unreliable marker of beat-to-beat blood pressure. Journal of Applied Physiology, 100(1), 136–141. https://doi.org/10.1152/japplphysiol.00657.2005

Prescott, H. C., & Angus, D. C. (2018). Enhancing recovery from sepsis: A review. JAMA, 319(1), 62–75. https://doi.org/10.1001/jama.2017.17687

Rangayyan, R. M. (2015). Biomedical signal analysis (2nd ed.). Wiley-IEEE Press.

Reinhart, K., Daniels, R., Kissoon, N., Machado, F. R., Schachter, R. D., & Finfer, S. (2017). Recognizing sepsis as a global health priority — A WHO resolution. New England Journal of Medicine, 377(5), 414–417. https://doi.org/10.1056/NEJMp1707170

Romero, A., Ballas, N., Kahou, S. E., Chassang, A., Gatta, C., & Bengio, Y. (2015). FitNets: Hints for thin deep nets (arXiv:1412.6550). arXiv. https://doi.org/10.48550/arXiv.1412.6550

Royal College of Physicians. (2017). National Early Warning Score (NEWS) 2: Standardising the assessment of acute-illness severity in the NHS (Updated report of a working party). RCP.

Rudd, K. E., Johnson, S. C., Agesa, K. M., Shackelford, K. A., Tsoi, D., Kievlan, D. R., Colombara, D. V., Ikuta, K. S., Kissoon, N., Finfer, S., Fleischmann-Struzek, C., Machado, F. R., Reinhart, K., Rowan, K., Seymour, C. W., Watson, R. S., West, T. E., Marinho, F., Hay, S. I., … Naghavi, M. (2020). Global, regional, and national sepsis incidence and mortality, 1990–2017: Analysis for the Global Burden of Disease Study. The Lancet, 395(10219), 200–211. https://doi.org/10.1016/S0140-6736(19)32989-7

Seymour, C. W., Liu, V. X., Iwashyna, T. J., Brunkhorst, F. M., Rea, T. D., Scherag, A., Rubenfeld, G., Kahn, J. M., Shankar-Hari, M., Singer, M., Deutschman, C. S., Escobar, G. J., & Angus, D. C. (2016). Assessment of clinical criteria for sepsis: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA, 315(8), 762–774. https://doi.org/10.1001/jama.2016.0288

Shrivastava, A., Gupta, A., & Girshick, R. (2016). Training region-based object detectors with online hard example mining. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 761–769). IEEE. https://doi.org/10.1109/CVPR.2016.89

Singer, M., Deutschman, C. S., Seymour, C. W., Shankar-Hari, M., Annane, D., Bauer, M., Bellomo, R., Bernard, G. R., Chiche, J.-D., Coopersmith, C. M., Hotchkiss, R. S., Levy, M. M., Marshall, J. C., Martin, G. S., Opal, S. M., Rubenfeld, G. D., van der Poll, T., Vincent, J.-L., & Angus, D. C. (2016). The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA, 315(8), 801–810. https://doi.org/10.1001/jama.2016.0287

Smith, G. B., Prytherch, D. R., Meredith, P., Schmidt, P. E., & Featherstone, P. I. (2013). The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation, 84(4), 465–470. https://doi.org/10.1016/j.resuscitation.2012.12.016

Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1), 1929–1958. http://jmlr.org/papers/v15/srivastava14a.html

Subbe, C. P., Kruger, M., Rutherford, P., & Gemmel, L. (2001). Validation of a modified Early Warning Score in medical admissions. QJM: An International Journal of Medicine, 94(10), 521–526. https://doi.org/10.1093/qjmed/94.10.521

Tamura, T., Maeda, Y., Sekine, M., & Yoshida, M. (2014). Wearable photoplethysmographic sensors — Past and present. Electronics, 3(2), 282–302. https://doi.org/10.3390/electronics3020282

Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. (1996). Heart rate variability: Standards of measurement, physiological interpretation and clinical use. Circulation, 93(5), 1043–1065. https://doi.org/10.1161/01.CIR.93.5.1043

Ullah, S., Higgins, H., Braem, B., Latre, B., Blondia, C., Moerman, I., Saleem, S., Rahman, Z., & Kwak, K. S. (2012). A comprehensive survey of wireless body area networks. Journal of Medical Systems, 36(3), 1065–1094. https://doi.org/10.1007/s10916-010-9571-3

van Buuren, S., & Groothuis-Oudshoorn, K. (2011). mice: Multivariate imputation by chained equations in R. Journal of Statistical Software, 45(3), 1–67. https://doi.org/10.18637/jss.v045.i03

van den Oord, A., Dieleman, S., Zen, H., Simonyan, K., Vinyals, O., Graves, A., Kalchbrenner, N., Senior, A., & Kavukcuoglu, K. (2016). WaveNet: A generative model for raw audio (arXiv:1609.03499). arXiv. https://doi.org/10.48550/arXiv.1609.03499

van der Poll, T., van de Veerdonk, F. L., Scicluna, B. P., & Netea, M. G. (2017). The immunopathology of sepsis and potential therapeutic targets. Nature Reviews Immunology, 17(7), 407–420. https://doi.org/10.1038/nri.2017.36

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. In I. Guyon et al. (Eds.), Advances in Neural Information Processing Systems (Vol. 30). Curran Associates. https://proceedings.neurips.cc/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html

Vincent, J. L., Moreno, R., Takala, J., Willatts, S., De Mendonça, A., Bruining, H., Reinhart, C. K., Suter, P. M., & Thijs, L. G. (1996). The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. Intensive Care Medicine, 22(7), 707–710. https://doi.org/10.1007/BF01709751

Webster, J. G. (Ed.). (2010). Medical instrumentation: Application and design (4th ed.). Wiley.

Wong, A., Otles, E., Donnelly, J. P., Krumm, A., McCullough, J., DeTroyer-Cooley, O., Pestrue, J., Phillips, M., Koneshloo, Y., Kamdar, N., Shah, N., & Singh, K. (2021). External validation of a widely implemented proprietary sepsis prediction model in hospitalized patients. JAMA Internal Medicine, 181(8), 1065–1070. https://doi.org/10.1001/jamainternmed.2021.2626

Woo, S., Park, J., Lee, J.-Y., & Kweon, I. S. (2018). CBAM: Convolutional block attention module. In V. Ferrari et al. (Eds.), Proceedings of the European Conference on Computer Vision (pp. 3–19). Springer. https://doi.org/10.1007/978-3-030-01234-2_1

Yu, F., & Koltun, V. (2016). Multi-scale context aggregation by dilated convolutions (arXiv:1511.07122). arXiv. https://doi.org/10.48550/arXiv.1511.07122

Zadrozny, B., & Elkan, C. (2002). Transforming classifier scores into accurate multiclass probability estimates. In Proceedings of the 8th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 694–699). ACM. https://doi.org/10.1145/775047.775151

Zhang, Y., & Yang, Q. (2022). A survey on multi-task learning. IEEE Transactions on Knowledge and Data Engineering, 34(12), 5586–5609. https://doi.org/10.1109/TKDE.2021.3070203