Anna-Lena Semmler
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Anna-Lena Semmler has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
Anouk E. Lichtenberg
Institute for Transfusion Medicine and Immunohematology, German Red Cross Blood Donation Service Baden-Wuerttemberg-Hessen, 60528 Frankfurt, Germany
1 shared publication- Cassian Afting
Centre for Organismal Studies Heidelberg (COS), Heidelberg University, Germany
1 shared publication Christian Seidl
Institute for Transfusion Medicine and Immunohematology, German Red Cross Blood Donation Service Baden-Wuerttemberg-Hessen, 60528 Frankfurt, Germany
1 shared publicationChristina Merk
Institute for Clinical Transfusion Medicine and Immunogenetics Ulm, German Red Cross Blood Transfusion Service, Baden Wuerttemberg-Hessen, Ulm, and University Hospital Ulm, 89081 Ulm, Germany
1 shared publication- Daniel Fürst
Institute of Clinical Transfusion Medicine and Immunogenetics Ulm, German Red Cross Blood Transfusion Service, Baden-Württemberg – Hessen and University Hospital Ulm, Ulm, Germany
1 shared publication - Hanns-Martin Lorenz
Division of Rheumatology, Department of Medicine V, Heidelberg University Hospital, 69120 Heidelberg, Germany
1 shared publication
Latest Publications
- Open Access
Simulation-Trained Deep Learning for Automated Cell-Based HLA Antibody Assay Interpretation in Pre-Transplant Diagnostics
Preformed and de novo antibodies against donor human leukocyte antigen (HLA) antigens remain a major cause of antibody-mediated rejection and graft loss after organ transplantation. Although solid-phase assays and virtual crossmatching have reshaped pre-transplant risk assessment, physical crossmatching, which assesses whether recipient antibodies react with donor cells, remains widely used as the final compatibility assessment before transplantation. These workflows include both complement-dependent cytotoxicity (CDC) and flow cytometry crossmatch (FCXM) assays, but their interpretation remains partly manual, operator-dependent, and, for microscopic CDC readout, semi-quantitative. Here, we present AlloViewer, a web-based software platform for automated and traceable interpretation of image-based and flow-cytometry-based HLA antibody diagnostics. For CDC microscopy, AlloViewer employs a simulation-trained deep learning (UNet) workflow that combines automated lymphocyte segmentation with experiment-specific fluorescence classification and well-level cytotoxicity scoring. To deliberately capture the technical variability encountered in routine diagnostics, we generated simulated CDC-like training images spanning differences in image resolution, acquisition conditions, staining quality, cell density, cell distribution, clustering, and background fluorescence, among others. The resulting simulation-trained model enabled robust lymphocyte detection across heterogeneous imaging conditions and outperformed a conventional rule-based image analysis pipeline under variable acquisition conditions. Automated CDC scoring achieved performance within the range of human inter-annotator variability and approached the practical reproducibility limit defined by human disagreement. To cover all modalities of pre-transplant physical crossmatching, AlloViewer further supports automated FCXM interpretation through cell population identification and population-specific immunoglobulin G (IgG) readouts. The platform integrates these workflows in an assay-specific web interface and provides application programming interface (API) access for programmatic submission of assay data and retrieval of processed results. Together, AlloViewer establishes an integrated computational framework for standardized and traceable interpretation of CDC crossmatch assays, CDC-based HLA antibody identification testing using commercial test-cell panels, and FCXM workflows. More broadly, this work demonstrates how simulation-trained artificial intelligence (AI) can facilitate robust computational analysis across technically heterogeneous laboratory environments, providing a framework for standardizing traditionally operator-dependent diagnostic workflows.
Open AccessSimulation-Trained Deep Learning for Automated Cell-Based HLA Antibody Assay Interpretation in Pre-Transplant Diagnostics
Preformed and de novo antibodies against donor human leukocyte antigen (HLA) antigens remain a major cause of antibody-mediated rejection and graft loss after organ transplantation. Although solid-phase assays and virtual crossmatching have reshaped pre-transplant risk assessment, physical crossmatching, which assesses whether recipient antibodies react with donor cells, remains widely used as the final compatibility assessment before transplantation. These workflows include both complement-dependent cytotoxicity (CDC) and flow cytometry crossmatch (FCXM) assays, but their interpretation remains partly manual, operator-dependent, and, for microscopic CDC readout, semi-quantitative. Here, we present AlloViewer, a web-based software platform for automated and traceable interpretation of image-based and flow-cytometry-based HLA antibody diagnostics. For CDC microscopy, AlloViewer employs a simulation-trained deep learning (UNet) workflow that combines automated lymphocyte segmentation with experiment-specific fluorescence classification and well-level cytotoxicity scoring. To deliberately capture the technical variability encountered in routine diagnostics, we generated simulated CDC-like training images spanning differences in image resolution, acquisition conditions, staining quality, cell density, cell distribution, clustering, and background fluorescence, among others. The resulting simulation-trained model enabled robust lymphocyte detection across heterogeneous imaging conditions and outperformed a conventional rule-based image analysis pipeline under variable acquisition conditions. Automated CDC scoring achieved performance within the range of human inter-annotator variability and approached the practical reproducibility limit defined by human disagreement. To cover all modalities of pre-transplant physical crossmatching, AlloViewer further supports automated FCXM interpretation through cell population identification and population-specific immunoglobulin G (IgG) readouts. The platform integrates these workflows in an assay-specific web interface and provides application programming interface (API) access for programmatic submission of assay data and retrieval of processed results. Together, AlloViewer establishes an integrated computational framework for standardized and traceable interpretation of CDC crossmatch assays, CDC-based HLA antibody identification testing using commercial test-cell panels, and FCXM workflows. More broadly, this work demonstrates how simulation-trained artificial intelligence (AI) can facilitate robust computational analysis across technically heterogeneous laboratory environments, providing a framework for standardizing traditionally operator-dependent diagnostic workflows.