Operating on sophisticated natural language processing (NLP) algorithms, the NLP Viewer identifies and categorises important words and phrases in documents. Users upload a text document (in formats such as TXT, PDF, etc.) to the editor, and the algorithm then scans the content and identifies key information. For clinical content, it categorizes them into relevant categories such as health conditions, diagnoses, clinical procedures, medications, anatomy, and others. In the process of recognizing clinical content, the information is further linked to encoded clinical concepts from relevant codebooks and ontologies, such as ICD-10, KTDP, ATC, and SNOMED.
The NLP Viewer is designed to be used by various users, ranging from healthcare professionals, researchers, students to technical experts.
The key services for the NLP Viewer application are provided by the NLP server. This central component executes sophisticated natural language processing algorithms and integrates with other components of the application to ensure comprehensive document analysis. The NLP server identifies and categorizes key information, linking it to broader information sources. It adeptly processes a variety of document types, from health reports to research articles, including logs and other formats.
For linking clinical information to appropriate terminological concepts, the NLP Viewer employs SNOMED, the most extensive collection of systematically gathered and connected clinical concepts in the world. It serves as the foundation for processing healthcare texts. As explained by Robert Tovornik, the head of the Innovation team at Better, the company that developed the application, this occurs when "the NLP Viewer recognizes a specific diagnosis, health condition, or any other clinical information in the text and associates it with the corresponding term in SNOMED. This way, we achieve standardization of terms, information connectivity, and a deeper understanding of clinical content. SNOMED enables further linking of identified information with internationally recognized terms, facilitating communication and collaboration among experts on a global scale. Moreover, the use of SNOMED terms is crucial for ensuring interoperability among different health information systems".
The development of the NLP Viewer technology took place in a multidisciplinary team, bringing together experts in computer science, data science, and medicine. At Better, the development was led by the Innovation team, namely Robert Tovornik, Matic Bernik, and Emil Plesnik.
Initially, they analysed the requirements and expectations of potential users, setting the guidelines for development. Special attention was devoted to collecting and analysing Slovenian clinical terms, laying the foundation for adapting and improving natural language processing algorithms. This ensured that the application was tailored to the linguistic and professional specifics of the Slovenian healthcare environment.
"The introduction of global medical standards into the context of Slovenian medical practice was a significant challenge, which included the extensive translation of SNOMED terminology into Slovenian and the formation of clinical concepts," said Tovornik. He added that collaboration and mutual complementation among team members, content experts, and technology development experts were crucial throughout the development process.
The scope and utility of the NLP Viewer technology are vast, particularly in clinical data analytics. It represents a significant advancement in the field of Slovenian technologies, as tools for advanced understanding of Slovenian texts are rare, particularly in the clinical domain. Therefore, the vision of Better and the team involved in the development is the continuous enhancement and expansion of clinical capabilities. "Although the application is already very powerful, we see the dynamics of the healthcare sector with its specific challenges. Each domain or hospital has its characteristics in using clinical terms and practices. We see further potential for progress in close collaboration with domain experts, aiming to develop solutions that address the everyday challenges they encounter in practice," said Robert Tovornik.
A deep understanding of clinical data is essential for pioneering digital healthcare solutions. Hence, the basic potential of the application lies in advanced analysis of free texts, labeling, and summarising texts. Assistance in searching, structuring, and coding content, which is currently a very time-consuming part of the healthcare process, is the most crucial factor in the application's further development. "In the future, we see a much broader range of possibilities, but for now, our goal is to continue to explore and adapt the use of this advanced technology to best serve the needs of the healthcare sector," Tovornik added.