A5035068963- Privacy-Preserving Technologies in Data
- Artificial Intelligence in Healthcare and Education
- Bioinformatics and Genomic Networks
- Radiomics and Machine Learning in Medical Imaging
- Machine Learning in Healthcare
- Gene expression and cancer classification
- Ethics in Clinical Research
- Research Data Management Practices
- Scientific Computing and Data Management
- Metabolomics and Mass Spectrometry Studies
- Biomedical Text Mining and Ontologies
- Cell Image Analysis Techniques
- AI in cancer detection
- Health and Medical Studies
- COVID-19 diagnosis using AI
- Traditional Chinese Medicine Studies
- Breast Cancer Treatment Studies
- Gut microbiota and health
- Data Quality and Management
Universität Hamburg
2022-2024
Background Machine learning and artificial intelligence have shown promising results in many areas are driven by the increasing amount of available data. However, these data often distributed across different institutions cannot be easily shared owing to strict privacy regulations. Federated (FL) allows training machine models without sharing sensitive In addition, implementation is time-consuming requires advanced programming skills complex technical infrastructures. Objective Various tools...
Most complex diseases, including cancer and non-malignant diseases like asthma, have distinct molecular subtypes that require clinical approaches. However, existing computational patient stratification methods been benchmarked almost exclusively on omics data only perform well when mutually exclusive can be characterized by many biomarkers. Here, we contribute with a massive evaluation attempt, quantitatively exploring the power of 22 unsupervised using both, simulated real transcriptome...
Background Central collection of distributed medical patient data is problematic due to strict privacy regulations. Especially in clinical environments, such as time-to-event studies, large sample sizes are critical but usually not available at a single institution. It has been shown recently that federated learning, combined with privacy-enhancing technologies, an excellent and privacy-preserving alternative sharing. Objective This study aims develop validate privacy-preserving, survival...
Background: Patient-reported survey data are used to train prognostic models aimed at improving healthcare. However, such typically available multi-centric and, for privacy reasons, cannot easily be centralized in one repository. Models trained locally less accurate, robust, and generalizable. We present apply privacy-preserving federated machine learning techniques model building, where local never leaves the legally safe harbors of medical centers. Methods: centralized, local, on two...
Batch effects in omics data obscure true biological signals and constitute a major challenge for privacy-preserving analyses of distributed patient data. Existing batch effect correction methods either require centralization, which may easily conflict with privacy requirements, or lack support missing values automated workflows. To bridge this gap, we developed fedRBE, federated implementation limma's removeBatchEffect method. We implemented it as an app the FeatureCloud platform. Unlike its...
Standardising the representation of biomedical knowledge among all researchers is an insurmountable task, hindering effectiveness many computational methods. To facilitate harmonisation and interoperability despite this fundamental challenge, we propose to standardise framework graph creation instead. We implement standardisation in BioCypher, a FAIR (findable, accessible, interoperable, reusable) transparently build graphs while preserving provenances source data. Mapping onto ontologies...
<sec> <title>BACKGROUND</title> Central collection of distributed medical patient data is problematic due to strict privacy regulations. Especially in clinical environments, such as time-to-event studies, large sample sizes are critical but usually not available at a single institution. It has been shown recently that federated learning, combined with privacy-enhancing technologies, an excellent and privacy-preserving alternative sharing. </sec> <title>OBJECTIVE</title> This study aims...
<p class="first" id="d4449516e229">Unsupervised patient stratification based on omics data is traditionally approached by clustering methods which may be inefficient for datasets with multiple patterns overlapping in rows and columns. Biclustering that are searching submatrices a specific pattern two-dimensional sample-gene matrix represent promising alternative to conventional [1]. However, practice, the existing biclustering show limited ability robustly recover known PAM50 breast cancer...
<sec> <title>BACKGROUND</title> Machine learning and artificial intelligence have shown promising results in many areas are driven by the increasing amount of available data. However, these data often distributed across different institutions cannot be easily shared owing to strict privacy regulations. Federated (FL) allows training machine models without sharing sensitive In addition, implementation is time-consuming requires advanced programming skills complex technical infrastructures....