JSJulia Sellin

Julia Sellin

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Julia Sellin has a total of 99 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Insulin Dose–ResponseAlignment for Glucose Forecasting in Type 1 Diabetes

    by Jonas Wolber, Moein Einollahzadeh Samadi, Julia Sellin, Martin Mücke, Andreas Schuppert

    Supervised models trained on observational clinical logs often learn spurious correlations. Held-out prediction error can remain low while the fitted mapping contradicts expected pharmacology, such as predicting that insulin raises glucose. A dose–response curve (DRC) defines expected physiological behaviour when a treatment dose is varied while holding other factors fixed. In a fitted model, this response is an individual conditional expectation (ICE). Existing explainable artificial intelligence methods can diagnose an unphysiological ICE but cannot repair it. We introduce DOCAS (dose–response curve alignment via synthetic augmentation). Given a target DRC, DOCAS adds synthetic do(·) instances labelled with the curve's incremental effect and trains the chosen architecture once on the combined data. Alignment error (AE) is the root-mean-square gap between each test-context ICE and the target DRC. In type 1 diabetes glucose forecasting, the established UVa/Padova simulator supplies the insulin DRC that we use as the target. Across OhioT1DM (n = 12) and AZT1D (n = 24), DOCAS reduced the 60-minute AE from 38.13 to 2.79 mg/dL and from 49.30 to 4.13 mg/dL, respectively (89–93% across horizons), at RMSE increases of 2.7–4.4%. In an in-silico decision-support evaluation, an aligned model raised the mean correction bolus from 1.09 to 2.23 U and post-decision time in range from 73.61% to 76.01% without inducing hypoglycaemia under an explicit safety guard. These results indicate that aligning the treatment response is as important as minimizing prediction error when a forecaster is used for dosing decisions.

  • Open Access

    Predicting disease outcomes from remote monitoring using machine learning: a systematic review

    by Jonas Wolber, Andreas Schuppert, Martin Mücke, Julia Sellin

    Background: Chronic conditions cause millions of deaths annually worldwide. Remote patient monitoring using wearable devices and sensors, combined with machine learning (ML), offers promising strategies for disease management. However, diverse methodological approaches and study designs impede comparability and the development of best practice guidelines. Methods: A systematic review was conducted following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines. Four scientific databases were searched for relevant prospective studies published between 2014 and 2024. Studies had to use ML to predict disease outcomes of chronic conditions in remotely monitored patients. The studies were tagged for characteristics such as health outcomes, dataset, monitored parameters, and algorithms. Results: From 6668 initially identified studies, 76 met inclusion criteria. 73.7% of studies were considered to have a high risk of bias, mainly due to methodological shortcomings in the Analysis domain. Parkinson’s disease was most frequently monitored, followed by diabetes and chronic obstructive pulmonary disease (COPD). Wearable devices were the predominant remote sensors, with accelerometer data being the most common parameter. Tree-based algorithms were most frequent, and studies using leave-one-out cross-validation showed significantly higher accuracy. Feature engineering and publication year were also significantly associated with model performance. Conclusion: This review highlights both progress and challenges in applying ML to chronic disease monitoring. While conditions like Parkinson’s, COPD, and diabetes are well-represented, others such as liver and kidney diseases are underexplored. Future research should prioritize standardization of methodologies, model interpretability, and ethical considerations including data privacy and algorithmic fairness. When properly implemented, ML-driven remote monitoring has the potential to enhance patient care, reduce complications, and deepen our understanding of chronic conditions. However, addressing challenges in reproducibility, generalizability, and clinical integration is crucial for advancing the field.


  • Kurze Wege zur Diagnose mit künstlicher Intelligenz – systematische Literaturrecherche zu „diagnostic decision support systems“

    by Julia Sellin, Jean Tori Pantel, Natalie Börsch, Rupert Conrad, Martin Mücke

    Zusammenfassung Hintergrund Seltene Erkrankungen werden oft erst spät erkannt. Ihre Diagnose ist aufgrund der Diversität, Komplexität und Heterogenität klinischer Symptome besonders anspruchsvoll. Computergestützte diagnostische Hilfen, oft als „diagnostic decision support systems“ (DDSS) bezeichnet, sind vielversprechende Tools, um die Zeit bis zur Diagnose zu verkürzen. DDSS sind trotz erster positiver Evaluationen noch nicht sehr weit verbreitet, was unter anderem auf die mangelnde Integration in existierende klinische oder Praxisinformationssysteme zurückgeführt werden kann. Ziel der Arbeit Dieser Beitrag bietet einen Einblick in aktuell existierende DDSS, die ohne Zugriff auf elektronische Patientenakten funktionieren und nur einfach zu beschaffende Informationen benötigen. Material und Methoden Im Rahmen einer systematischen Literaturrecherche wurden 8 Beiträge identifiziert, in denen DDSS untersucht werden, die bei der Diagnose seltener Erkrankungen unterstützen können und dabei keinen Zugriff auf elektronische Patientenakten oder andere Informationssysteme in Praxen und Kliniken verlangen. Die wichtigsten Vor- und Nachteile der identifizierten Systeme zur Unterstützung bei der Diagnostik seltener Erkrankungen wurden extrahiert und zusammenfassend dargestellt. Ergebnisse Symptom-Checker sowie DDSS auf Basis von Porträtfotos und Schmerzzeichnungen existieren bereits. Der Reifegrad dieser Anwendungen ist unterschiedlich. Schlussfolgerung DDSS stehen aktuell noch vor einigen Herausforderungen – so gibt es Bedenken zu Datenschutz und Genauigkeit, zudem sind die Akzeptanz und Bekanntheit noch eher gering. Dem gegenüber steht ein großes Potenzial für eine schnellere Diagnosestellung, insbesondere bei seltenen Erkrankungen, die aufgrund ihrer großen Anzahl und geringen Bekanntheit leicht übersehen werden. Der Einsatz von DDSS sollte daher von Ärztinnen und Ärzten im Einzelfall gut abgewogen werden.


  • Open Access

    Correction to: Suggestion of a new standard in measuring the mandible via MRI and an overview of reference values in young women

    by Leonie Carina Ibald, Veronica Witte, Fank Klawonn, Rupert Conrad, Martin Mücke, Julia Sellin, Marcus Teschke

  • Open Access

    Suggestion of a new standard in measuring the mandible via MRI and an overview of reference values in young women

    by Leonie Carina Ibald, Veronica Witte, Fank Klawonn, Rupert Conrad, Martin Mücke, Julia Sellin, Marcus Teschke

    Purpose: Adult idiopathic condylar resorption (AICR) mainly affects young women, but generally accepted diagnostic standards are lacking. Patients often need temporomandibular joint (TMJ) surgery, and often jaw anatomy is assessed by CT as well as MRI to observe both bone and soft tissue. This study aims to establish reference values for mandible dimensions in women from MRI only and correlate them to, e.g., laboratory parameters and lifestyle, to explore new putative parameters relevant in AICR. MRI-derived reference values could reduce preoperative effort by allowing physicians to rely on only the MRI without additional CT scan. Methods: We analyzed MRI data from a previous study (LIFE-Adult-Study, Leipzig, Germany) of 158 female participants aged 15–40 years (as AICR typically affects young women). The MR images were segmented, and standardized measuring of the mandibles was established. We correlated morphological features of the mandible with a large variety of other parameters documented in the LIFE-Adult study. Results: We established new reference values for mandible morphology in MRI, which are consistent with previous CT-based studies. Our results allow assessment of both mandible and soft tissue without radiation exposure. Correlations with BMI, lifestyle, or laboratory parameters could not be observed. Of note, correlation between SNB angle, a parameter often used for AICR assessment, and condylar volume, was also not observed, opening up the question if these parameters behave differently in AICR patients. Conclusion: These efforts constitute a first step towards establishing MRI as a viable method for condylar resorption assessment.


  • Open Access

    Cluster-Analytic Identification of Clinically Meaningful Subtypes in MCAS: The Relevance of Heat and Cold

    by Tinus Häder, Gerhard J. Molderings, Frank Klawonn, Rupert Conrad, Martin Mücke, Julia Sellin

    Background Mast cell activation syndrome (MCAS) is a clinically heterogeneous disease with allergy-like symptoms and abdominal complaints. Its etiology is only partially understood and it is often overlooked. Aims The aim of this study was to identify subgroups of MCAS patients to facilitate diagnosis and allow a personalized therapy. Methods Based on data from 250 MCAS patients, hierarchical and two-step cluster analyses as well as association analyses were performed. The data used included data from a MCAS checklist asking about symptoms and triggers and a set of diagnostically relevant laboratory parameters. Results Using a two-step cluster analysis, MCAS patients could be divided into three clusters. Physical trigger factors were particularly decisive for the classification as they showed remarkable differences between the three clusters. Cluster 1, labeled high responders, showed high values for the triggers heat and cold, whereas cluster 2, labeled intermediate responders, presented with high values for the trigger heat and low values for cold. The third cluster, labeled low responders, did not react to thermal triggers. The first two clusters showed more divers clinical symptoms especially with regard to dermatological and cardiological complaints. Subsequent association analyses revealed relationships between triggers and clinical complaints: Abdominal discomfort is mainly triggered by histamine consumption, dermatological discomfort by exercise, and neurological symptoms are related to physical exertion and periods of starvation. The reasons for the occurrence of cardiological complaints are manifold and triggers for respiratory complaints still need better identification. Conclusion Our study identified three distinct clusters on the basis of physical triggers, which also differ significantly in their clinical symptoms. A trigger-related classification can be helpful in clinical practice for diagnosis and therapy. Longitudinal studies should be conducted to further understand the relationship between triggers and symptoms.


  • Open Access

    A diagnostic support system based on pain drawings: binary and k-disease classification of EDS, GBS, FSHD, PROMM, and a control group with Pain2D

    by D. Emmert, N. Szczypien, Tim T. A. Bender, L. Grigull, A. Gass, C. Link, F. Klawonn, R. Conrad, M. Mücke, J. Sellin

    Background and objective: The diagnosis of rare diseases (RDs) is often challenging due to their rarity, variability and the high number of individual RDs, resulting in a delay in diagnosis with adverse effects for patients and healthcare systems. The development of computer assisted diagnostic decision support systems could help to improve these problems by supporting differential diagnosis and by prompting physicians to initiate the right diagnostic tests. Towards this end, we developed, trained and tested a machine learning model implemented as part of the software called Pain2D to classify four rare diseases (EDS, GBS, FSHD and PROMM), as well as a control group of unspecific chronic pain, from pen-and-paper pain drawings filled in by patients. Methods: Pain drawings (PDs) were collected from patients suffering from one of the four RDs, or from unspecific chronic pain. The latter PDs were used as an outgroup in order to test how Pain2D handles more common pain causes. A total of 262 (59 EDS, 29 GBS, 35 FSHD, 89 PROMM, 50 unspecific chronic pain) PDs were collected and used to generate disease specific pain profiles. PDs were then classified by Pain2D in a leave-one-out-cross-validation approach. Results: Pain2D was able to classify the four rare diseases with an accuracy of 61–77% with its binary classifier. EDS, GBS and FSHD were classified correctly by the Pain2D k-disease classifier with sensitivities between 63 and 86% and specificities between 81 and 89%. For PROMM, the k-disease classifier achieved a sensitivity of 51% and specificity of 90%. Conclusions: Pain2D is a scalable, open-source tool that could potentially be trained for all diseases presenting with pain.


  • Seltene Erkrankungen in der Differenzialdiagnose bei Myalgien

    by D. Emmert, T. Rasche, J. Sellin, R. Brunkhorst, T. T. A. Bender, N. Weinstock, N. Börsch, L. Grigull, R. Conrad, M. Mücke

    Zusammenfassung Als Myalgien bezeichnet man Schmerzempfindungen in der Muskulatur. Gemäß der 2020 aktualisierten Leitlinie (AWMF-Registernummer: 030/051) sollten neben Anamnese und klinischer Untersuchung eine elektrophysiologische Untersuchung und ein Standardlabor erfolgen. Weitere Spezialuntersuchungen wie molekulargenetische Untersuchungen, erweiterte Labordiagnostik, Bildgebung oder eine Muskelbiopsie sind nur in Sonderfällen nötig. Der Fokus dieses Artikels liegt auf seltenen neuromuskulären Erkrankungen, die klassischerweise mit Myalgien einhergehen. Hierbei wurden ätiologisch verschiedene Erkrankungen berücksichtigt, wobei einigen genetisch bedingten Krankheiten (fazioskapulohumerale Muskeldystrophie [FSHD], myotone Dystrophie, Morbus McArdle, Morbus Pompe, Glieder-Gürtel-Dystrophie) (auto-)immunbedingte Pathogenesen (Stiff-person-Syndrom, Isaacs-Syndrom) gegenüberstehen. Insbesondere die diagnoserelevanten Aspekte werden hierbei hervorgehoben. Therapeutische Aspekte der Erkrankungen werden nicht näher behandelt.