Use case: Automated reading of health data

The ongoing digitalization of the healthcare sector requires efficient solutions for recording, processing and Analysis of laboratory values. Healthcare facilities and service providers need precise and timely data in order to make informed decisions and provide their patients with the best possible information and advice. Konfuzio has developed a solution for the simple provision and evaluation of laboratory values and health data that can be used in a variety of ways.

1. initial situation

A service provider in the healthcare sector wants to develop a digital and intuitive information channel for patients. Users will be able to upload their lab results as images or documents and receive an analysis of them on a provided platform. The readout, processing and evaluation of the laboratory values is to be automated and the information made available in real time. To implement this project, the healthcare provider chose the Helm & Nagel GmbH, with your Konfuzio software solution as an experienced project partner.

2. challenges in the selection of health data

The selection of laboratory values poses various challenges. These are primarily due to the special nature of the industry: The sensitivity and importance of the data are particularly high in the healthcare sector. Patients and service providers therefore have equally high demands on the quality and security of the information read out. 

Another challenge in the extraction of laboratory values is that there is currently no standardized representation of health data or laboratory results. Data extraction must therefore recognize and process many different formats. It should also produce results that are as accurate as possible, even with poorly legible user submissions.

In summary, the following points can be identified as the greatest challenges of the project.

• Variety of formats - Laboratory values can be available in different formats and layouts, which makes standardized processing difficult. In addition, the images submitted by users are often of poor quality
• High user expectations - The requirement for the completeness and correctness of data is particularly high in the healthcare sector. Errors in selection or evaluation have major consequences in the healthcare sector - they can lead to uncertainties for patients and form the basis for incorrect therapies.
• Data Privacy - Health data and laboratory results are private and sensitive data that are particularly worthy of protection. Secure processing and storage is therefore a key challenge when dealing with health data.

3. objective of the development

The main objective of the planned application is to provide patients and healthcare providers with a user-friendly platform that enables efficient reading and analysis of laboratory values. The platform should:

• Digitize laboratory findings in a uniform presentation.
• A detailed evaluation of the values in the context of standard ranges is provided.
• Provide helpful explanations and recommendations in the event of deviations from standard values.
• Provide a history of the health data
• Provide an interface that facilitates integration into existing healthcare provider systems.

4. solution & implementation

Together with Konfuzio, the challenges described were solved and the goals set were achieved. On the basis of Konfuzio Medical NER a customized solution was developed that enables the analysis and readout of laboratory values and health data quickly and precisely. Konfuzio's Medical NER technology is particularly powerful because it has been trained on a wide range of medical terms, abbreviations and specific terminology. 

Various language and extraction models were also trained and tested in order to further increase the accuracy of the extraction and solve the challenges of the diverse forms. The experts at Konfuzio extensively analyzed the strengths and weaknesses of the models and combined potentials. However, none of the models tested were able to meet the high quality requirements of the healthcare industry. 

Through manual development work by the experts at Konfuzio, close coordination with the healthcare provider's partners and the linking of various models, it was possible to develop a tailor-made model based on the Konfuzio Table Recognizer. This solution has shown significant progress in data quality and in the recognition of different formats. The following models were used in the development:

• Konfuzio RF Extraction Kit - This AI solution was deemed promising for extracting additional information and context.
• Konfuzio Table Recognizer - Use this tool to recognize and extract text from tables.
• SpaCy models - First attempts to use Natural Language Processing (NLP) for the identification of specific laboratory values.

In addition, the flexible Konfuzio API a Interface for seamless integration into the healthcare provider's existing systems. The data is thus read out in real time, transmitted and made available to the user on the service provider's platform.

To ensure a comprehensive Data Privacy Konfuzio relies on several pillars: Firstly, the data is processed and stored anonymously - it is only assigned to the user's anonymous account so that they can view the information via the healthcare provider's platform. In addition, Konfuzio always works in compliance with the GDPR, stores the data in accordance with European regulations and follows high data protection standards. By defining clear access rights, the data integrity of the laboratory values is also comprehensively ensured.

5. resultReliable extraction of health data

By implementing the platform, it has been possible to extract and analyze laboratory values efficiently and precisely. The tests prove that the solution is able to process various formats and generate high-quality, user-friendly output. The platform supports the healthcare provider in making the laboratory results of its users available quickly and comprehensibly.

By linking different models, the accuracy of the extraction could be further increased and thus also meet the high demands of the healthcare industry. The data is transferred securely and quickly via an API interface so that the evaluation and classification of the health data is available in real time and can be viewed by users.

The result is an intuitive platform for the exchange of information between healthcare providers and patients. The solution increases medical transparency and patient autonomy and is another example of how AI can have a positive impact on people's everyday lives.

6. conclusion

The development of the laboratory values platform for the healthcare provider is an important step towards increasing efficiency in the healthcare sector. Despite initial challenges, Konfuzio has developed a solution that meets the needs of the healthcare provider and its users through innovative approaches and the use of advanced AI technologies.

In the long term, the platform will help to optimize the processing of laboratory values and improve the quality of patient care. This solution has not only created clear added value for the service provider, but has also laid the foundation for future improvements in the field of digital healthcare solutions.

The development of high-precision extraction and processing of laboratory values can also be used in other applications. Konfuzio is also happy to work with your company to develop customized solutions for extracting laboratory values. In addition to integration with healthcare providers, the extraction and processing of laboratory values is also relevant for the following sectors:

• Clinical diagnostics - Automated analysis of laboratory results enables a faster and more accurate diagnosis, which means patients can be treated more quickly.
• Research & Development - In research, large volumes of laboratory results can be analyzed efficiently and valuable data can be used to gain new insights and identify trends.
• Quality control in the industry - Automated systems can consistently and accurately monitor and analyze lab results to ensure production standards are met.