How AI is connecting different stages of diagnostic workflows

Artificial intelligence is expanding its role in healthcare beyond supporting individual diagnostic tasks. Current research illustrates how AI can assist at different stages of the diagnostic pathway by extracting clinically relevant information from diverse data sources, including routine healthcare records, genomic data, and biomedical images.

Although these projects were developed independently and address different scientific questions, they reflect a common direction: using AI to make complex diagnostic data more accessible and actionable within clinical and research workflows.

Researchers at Hannover Medical School (MHH) are investigating whether routine health insurance data can be used to predict individual cancer risk at an earlier stage. As part of the ROKAVI project, the team is developing and comparing conventional statistical models with AI-based algorithms using routine health insurance data, including diagnoses, healthcare services, and sociodemographic information.

The approach aims to complement existing screening programs by identifying people who may benefit from earlier preventive measures or targeted communication. The project also examines ethical considerations and patient preferences regarding risk communication, reflecting the need to integrate predictive analytics into established care pathways responsibly. If successful, the work could demonstrate how routinely collected healthcare data can support more targeted prevention without requiring additional diagnostic procedures.

While other researchers are focusing on population-level risk assessment, scientists at the Max Planck Institute for Molecular Genetics are addressing a different challenge: interpreting the biological activity encoded in DNA sequences.

Unlike earlier sequence-based models, the AI tool Corgi (Context-aware Regulatory Genomics Inference) incorporates information about regulatory gene expression, allowing predictions that account for the cellular environment in which genes operate. The researchers report that the model can predict multiple indicators of genome activity, including gene expression and chromatin accessibility.

Although the work is primarily intended for basic research, it addresses a challenge that extends beyond genomics: deriving biologically relevant insights from large-scale molecular data. Improved interpretation of disease-relevant variants could strengthen the analytical foundation for precision diagnostics and future research into disease mechanisms.

Biomedical imaging represents another area where AI is helping researchers extract clinically relevant information from complex datasets. At the University of Duisburg-Essen and the Leibniz Institute for Analytical Sciences (ISAS), Professor Jianxu Chen develops AI methods that analyze large volumes of complex microscopy data.

Modern microscopy generates multidimensional datasets that are difficult to evaluate manually. Chen's research focuses on identifying subtle cellular patterns associated with disease and integrating AI into imaging workflows from the earliest stages of experiment design. Tailoring algorithms to specific biomedical questions helps ensure that imaging data are collected and analyzed with the intended clinical or research application in mind, reducing manual effort while revealing information that may otherwise remain unnoticed.

These three research projects address different scientific questions. Nevertheless, they demonstrate how AI is being applied across multiple domains of diagnostics, from analyzing routine healthcare records and genomic data to interpreting complex biomedical images. Together, they illustrate how AI is expanding the range of information that can be incorporated into diagnostic workflows, allowing different types of clinical and biological data to contribute to more comprehensive analysis.

Sources: Hannover Medical School, Max Planck Institute for Molecular Genetics, University of Duisburg-Essen