LIMS IN&OUT: Automated order entry and report generation
In laboratory operations, a great deal of paperwork and administrative work arises in addition to the actual analysis. Orders must be entered, results assigned, and reports written, and data moves back and forth between different systems. Much of this is still done by hand today. This takes time, is prone to errors, and ties up specialized staff who are urgently needed elsewhere.
Two types of software play a central role in this process:
- LIMS (Laboratory Information and Management System): This software manages a laboratory’s orders, samples, test results, and reports.
- HIS (Hospital or Clinic Information System): This software manages patient data and the course of treatment, such as test results and discharge summaries.
Both systems contain valuable information, but the data often has to be laboriously entered, transferred, and converted into text form. This is where artificial intelligence, or more specifically, so-called large language models (LLMs), can help. These are AI models that have been trained on a vast amount of text and have thereby learned to understand language and compose texts on their own. Using existing reports as a model, they can draft new reports, answer questions about content, or evaluate and categorize data. The specialist remains the final authority: the system makes suggestions, but the human makes the decision.
The data from LIMS and KIS can be used for yet another purpose: Process Mining. Each of these systems logs when a specific step took place for a given order or patient—for example, sample receipt, analysis, release, and report dispatch. These records are called Event Logs. Process Mining automatically analyzes them and uses them to reconstruct how workflows actually unfold in practice—not just how they are planned. This reveals bottlenecks, wait times, detours, and deviations, allowing for targeted process improvements. In addition, Data Mining can detect anomalies and irregularities in the data.
The DATA FLOW Project: Data Automation and Transfer Algorithms for Fast Laboratory Operations Workflow
DATA FLOW automates data flow in laboratories and clinics. The focus is on LIMS IN&OUT, i.e., automated order entry and report generation. Incoming information, such as from order forms, is automatically captured and transferred to the appropriate system; reports and letters are generated from just a few details. This applies not only to LIMS but also to HIS. Using process mining on the existing data from both systems, it is also possible to analyze where processes can be improved.
The technology behind it is the same in all projects. However, each application is individually tailored to the respective circumstances and systems to ensure it functions reliably in everyday use.
The technology behind it is the same in all projects; however, each application is individually adapted to the respective circumstances and systems to ensure it functions reliably in everyday use.
All subprojects share a common foundation: the use of language models.
- “Large Language Models” are trained using existing texts. They generate new texts—such as reports—and can answer questions.
- Termination letters can be automatically generated from just a few details based on previous reports.
- Very large amounts of data are required for routine operation. However, smaller “proof-of-concept” versions can already be achieved with just a few (<500) examples.
- Chatbots that evaluate and categorize data as well as answer questions are also possible.
- Data from LIMS and HIS can be used for process mining to optimize processes and detect anomalies, for example, based on existing event logs.
Collaborative Projects with Partners
The following applications are currently being developed in collaboration with partners:
- Text recognition and information extraction from non-standardized order forms. Relevant information is automatically extracted from forms that vary in structure and layout (see also AI-SEED)
- Process mining for detecting irregularities. Extracting event logs from the individual machines on which a sample was processed and calculating the average processing times for samples through the laboratory process.
- Preparation of rehabilitation discharge letters: A particular challenge is ensuring compliance with the regulations of the German Pension Insurance.
