Sarala Ghimire Subedi

• Subedi, Sarala Ghimire; Celik, Turgay; Gerdes, Martin & Omlin, Christian Walter Peter (2026). Looking Ahead When It Is Safe: An Uncertainty-Aware Paradigm for Blood Glucose Prediction with Dynamic Horizon Control. Machine Learning and Knowledge Extraction. 8(6). doi: 10.3390/make8060145. Full text in Research Archive Show summary

  Abstract Reliable time-series forecasting under rapidly changing conditions remains a critical challenge across many domains, particularly in healthcare, where physiological signals are inherently dynamic and uncertain. Blood glucose level prediction exemplifies this challenge, as accurate and timely forecasts are essential for effective diabetes management, yet traditional approaches rely on fixed prediction horizons and single-point estimates, which may yield unreliable decisions under rapidly changing physiological conditions. In this work, we propose a novel approach for adaptive horizon selection, applied to BGL prediction, employing a deep learning model. It employs evidential learning-based uncertainty quantification that decompose uncertainty into epistemic and aleatoric. Each set of models is trained to predict blood glucose levels at different future time steps, each providing both a point prediction and an associated uncertainty measure. At inference time, it dynamically balances predictive accuracy and reliability by selecting the longest horizon whose predicted uncertainty remains below a predefined threshold. This enables confidence-based horizon selection, using longer prediction horizons during stable periods and switching to shorter horizons when uncertainty signals critical glucose events requiring immediate intervention. This uncertainty-aware prediction approach promotes transparency by exposing confidence levels alongside predictions. Applicable to time-series forecasting tasks broadly, the proposed framework demonstrates encouraging potential, and when applied to BGL prediction as a representative clinical case, shows particular promise for supporting glycemic management through calibrated uncertainty estimation, offering a more transparent and interpretable alternative to fixed-horizon models toward trustworthy decision support in diabetes care.

• Subedi, Sarala Ghimire; Celik, Turgay; Gerdes, Martin & Omlin, Christian Walter Peter (2026). Personalized Blood Glucose Prediction Using Physiology- Informed Machine Learning. Machine Learning and Knowledge Extraction. 8(4). doi: 10.3390/make8040096. Full text in Research Archive
• Subedi, Sarala Ghimire; Celik, Turgay; Gerdes, Martin & Omlin, Christian Walter Peter (2024). Deep learning for blood glucose level prediction: How well do models generalize across different data sets? PLOS ONE. 19(9). doi: 10.1371/journal.pone.0310801. Full text in Research Archive Show summary

  Deep learning-based models for predicting blood glucose levels in diabetic patients can facilitate proactive measures to prevent critical events and are essential for closed-loop control therapy systems. However, selecting appropriate models from the literature may not always yield conclusive results, as the choice could be influenced by biases or misleading evaluations stemming from different methodologies, datasets, and preprocessing techniques. This study aims to compare and comprehensively analyze the performance of various deep learning models across diverse datasets to assess their applicability and generalizability across a broader spectrum of scenarios. Commonly used deep learning models for blood glucose level forecasting, such as feed-forward neural network, convolutional neural network, long short-term memory network (LSTM), temporal convolutional neural network, and self-attention network (SAN), are considered in this study. To evaluate the generalization capabilities of each model, four datasets of varying sizes, encompassing samples from different age groups and conditions, are utilized. Performance metrics include Root Mean Square Error (RMSE), Mean Absolute Difference (MAD), and Coefficient of Determination (CoD) for analytical asssessment, Clarke Error Grid (CEG) for clinical assessments, Kolmogorov-Smirnov (KS) test for statistical analysis, and generalization ability evaluations to obtain both coarse and granular insights. The experimental findings indicate that the LSTM model demonstrates superior performance with the lowest root mean square error and highest generalization capability among all other models, closely followed by SAN. The ability of LSTM and SAN to capture long-term dependencies in blood glucose data and their correlations with various influencing factors and events contribute to their enhanced performance. Despite the lower predictive performance, the FFN was able to capture patterns and trends in the data, suggesting its applicability in forecasting future direction. Moreover, this study helps in identifying the optimal model based on specific objectives, whether prioritizing generalization or accuracy.

• Subedi, Sarala Ghimire; Martinez, Santiago; Hartvigsen, Gunnar & Gerdes, Martin (2023). Augmented Reality enhanced device usage training tool for in-home health-self-monitoring by pregnant women. Reports of the European Society for Socially Embedded Technologies. doi: 10.48340/ihc2023_p015. Full text in Research Archive Show summary

  Virtual care comprising virtual visits and monitoring via audio or video has the potential to reduce access barriers to care and has been successfully implemented in prenatal care. It reduces the frequency of in-person visits and increases self-care skills. However, the knowledge and competence in handling monitoring equipment at home directly influences satisfaction and engagement with the system and the quality of the information provided to healthcare professionals. Therefore, providing ad-hoc training to end-users would help develop confidence in using devices at home correctly. This paper proposes an augmented reality (AR) application to guide and train pregnant women to use monitoring devices. The aim is to instruct pregnant women to use the monitoring devices and follow the video guidelines for handling them correctly. It validates pregnant women's encouragement and ability to self-care for their pregnancy.

• Subedi, Sarala Ghimire; Martinez, Santiago; Hartvigsen, Gunnar & Gerdes, Martin (2023). Virtual prenatal care: A systematic review of pregnant women's and healthcare professionals' experiences, needs, and preferences for quality care. International Journal of Medical Informatics. ISSN 1386-5056. 170. doi: 10.1016/j.ijmedinf.2022.104964. Full text in Research Archive Show summary

  Introduction: Exploitation of telehealth in prenatal care has the potential to reduce the access barrier to care and empower women to participate in their own care. This review aims to assess the practical implications of virtual prenatal care and identify the needs and experiences associated with it. Methods: A systematic literature review was conducted in four electronic databases: PubMed, Web of Science, Scopus, and Cochrane. The keywords used were “pregnancy”, “virtual visit”, “prenatal”, and others. The search included all relevant studies published from 2011 to 2021 written in English. Articles mentioning virtual prenatal care incorporating synchronous communication between pregnant women and health care professionals were included. Those unrelated to prenatal care or employing asynchronous means of virtual care were excluded. The review was structured following the PRISMA guidelines. Different quality appraisal methods such as JBI, CASP, NOS, and Cochrane were used to assess the methodological quality of the literature. The data were then analyzed based on the categorization of the studies. Results: Overall, 2863 articles were identified, of which 19 met the inclusion criteria after removing duplicates, screening of abstracts, and full text—four articles identified from hand-searching were incorporated, making a total of 23 eligible articles for the review. The studies’ findings revealed the preference for implementing costeffective virtual care based on the resource set, technological literacy, and consistent accessibility. Further, no significant differences in clinical outcomes were observed between two modes of care, virtual and in-person. The higher satisfaction by pregnant women and healthcare professionals indicated the continuity of the care. In addition, the hybrid model of virtual prenatal care integrated with traditional in-person care was acceptable to both low-risk and high-risk pregnant women. Virtual prenatal care substantially reduced travel time and absences from work, drops in clinic wait time and no-show rate, limited the risk of exposure during a pandemic, and increased self-accountability. Conclusion: Virtual prenatal care offers predominant advantages over in-person when it is carefully designed with the inclusion of pregnant women and healthcare professionals’ needs. Evidence showed that providing adequate technology training, proper instruction, and guidelines for initial setup and assurance of a reliable and accessible system is vital in increasing access to care.

• Subedi, Sarala Ghimire; Martinez, Santiago & Gerdes, Martin (2022). Self-imperative Care of Pregnancy using IoT Solutions. In Henriksen, André; Gabarron, Elia & Vimarlund, Vivian (Ed.), Proceedings of the 18th Scandinavian Conference on Health Informatics. Linköping Electronic Conference Proceedings. ISSN 9789179293444. p. 64–68. doi: 10.3384/ecp187011. Full text in Research Archive Show summary

  Typically, routine prenatal care includes several in-person visits with healthcare professionals by pregnant women, where fetal and maternal assessments are performed. This paper proposes an architectural framework for prenatal care using non-invasive, simple, and low-cost internet of things (IoT) monitoring system. The aim is to design an IoT-based architecture that serves as a fundamental system for self-imperative care in regular pregnancy check-ups in the comfort of the home that offers routine prenatalscreening tests. The system provides easy access to care regardless of the location and internet availability. We implemented preliminary architecture with simulated sensor data for blood pressure monitoring.

View all works in NVA