The Power of the Semantic Relationships in Concept Mapping for Fostering AI:

Dumdata Purify approach

https://bit.ly/429RPaG

A semantic relationship refers to the connection between concepts or entities in a domain of knowledge based on their meaning. It involves identifying and classifying these relationships to enhance data interoperability, searchability, and usability. Semantic relationships are essential because they provide a structured and meaningful way to connect disparate pieces of information. Understanding how concepts are related based on their meanings makes it possible to create connections that enhance the utility and accessibility of data. For instance, linking symptoms to diseases or treatments to outcomes through semantic relationships in healthcare can significantly improve information retrieval and decision-making processes. This structured approach allows for more accurate diagnoses, personalized treatments, and better patient outcomes. Semantic relationships facilitate interoperability across different systems and datasets, making sharing and integrating data from various sources easier. This leads to more comprehensive insights and fosters the development of AI applications that can leverage these connections to deliver innovative solutions in healthcare and other fields.

SNOMED proposes the following hierarchy: connecting the source to the target (screenshot from BioPortal: https://bit.ly/4hO0B2U).

SNOMED provides the following semantic relationship concepts.

· Broad map source to narrow map target (foundation metadata concept) (broad)

· Exact match between map source and map target (foundation metadata concept) (Exact)

· Map source not mappable to map target (foundation metadata concept) (unmatched)

· Map source to map target correlation not specified (foundation metadata concept) (conceptual)

· Partial overlap between map source and target (foundation metadata concept) (Partial)

Let’s explore how different SNOMED CT codes could match a clinical concept like “Hypertension”:

1. Exact Match: Exact match between map source and map target (foundation metadata concept)

— SNOMED Code: 38341003Description: Hypertensive disorder (this is a general code for any type of hypertension).

2. Partial Match: Partial overlap between map source and target (foundation metadata concept) (Partial)

— SNOMED Code: 67781001 Description: Essential hypertension (this is a specific subtype of hypertension, but it does not cover other forms, such as secondary hypertension).

3. Conceptual Match: Map source to map target correlation not specified (foundation metadata concept)

— SNOMED Code: 413145008 Description: Hypertension due to hyperaldosteronism (a specific cause of hypertension, which is conceptually related to hypertension but more precise).

4. Broad Match: Broad map source to narrow map target (foundation metadata concept)

— SNOMED Code: 22253000 Description: Pain (a general term for any kind of pain, much broader than "chronic pain," which is a more specific concept).

In our current project to showcase how Dumdata Purify could be used to enhance our understanding and connect between source and target concepts, we leveraged these SNOMED concepts to ensure that humans and machines alike not only understand the semantic meaning in the same way but also ensure that FAIR principle is achieved.

Use case with Dumdata Purify: map ICD10 Prolymphocytic leukemia [C91.3] into SNOMED Code. Concepts flagged in green are the Exact Match, and those in yellow refer to the Partial Match.

https://bit.ly/429RPaG

By defining and approving this relationship, we can develop a robust system for semantic match and search engines.

https://bit.ly/429RPaG

Dumdata Purify provides a trusted approach to ensure that the semantic relationship is well-defined and that the machine is readable for better interoperability.